[This Transcript is Unedited]
The Public Ledger Building
150 S. Independence Mall West
Philadelphia, Pennsylvania
VICKIE M. MAYS, PhD, MPH, Department of Psychology, University of California, Los Angeles, Los Angeles, California
KATHRYN L. COLTIN, MPH, Director, External Quality and Data Initiatives, Harvard Pilgrim Health Care, Wellesley, MA
EUGENE J. LENGERICH, VMD, Penn State University, Hershey, Pennsylvania
PAUL NEWACHECK, DrPH, Professor of Health Policy Studies, School of Medicine, University of California, San Francisco CA
BARBARA STARFIELD, MD, MPH, Distinguished University Professor and Professor of Health Policy and Pediatrics, School of Hygiene/Public Health, The Johns Hopkins University, Baltimore, Maryland
AUDREY BURWELL, OMH
SUSAN QUEEN, PhD, HRSA
LULA BEATTY, NCMHHD, NIH
LESLIE COOPER, PhD, NCI, NIH
SUZANNE HEURTIN-ROBERTS, PhD, MS, NCI, NIH
DALE HITCHCOCK, ASPE
CILLE KENNEDY, PhD, ASPE
JACQUELINE LUCAS, NCHS
EDNA PAISANO, IHS
Agenda Item: Call to Order, Introductions, Background and Overview.
DR. MAYS: Good morning. I think we are live on the internet. I would like to call to order the meeting of the Subcommittee on Populations of the National Committee on Vital and Health Statistics.
The subject of today's meeting is that we have asked several of our colleagues, who are sitting around the table and in the audience to talk with us about some of the state issues, and regional surveys, particularly as it pertains to the collection of data for racial and ethnic groups, and collecting data for racial and ethnic groups in which it will help us to better understand health disparities. As many of you know, the elimination of health disparities in racial and ethnic groups is a high priority of the administration.
What we are going to do is to start by introducing ourselves. After that, I will give a little background and we will then proceed to having some welcoming remarks. Maybe we will start here with Dale.
MR. HITCHCOCK: Good morning. My name is Dale Hitchcock. I am with the Data Policy Office at ASPE and staff person for this subcommittee on population.
DR. LENGERICH: Gene Lengerich from Penn State University, a member of the committee and subcommittee.
DR. FRIEDMAN: Dan Friedman, Massachusetts Department of Public Health and a member of the committee.
MR. ATKINSON: I am Delton Atkinson, with the National Center for Health Statistics. I am one of the speakers this morning.
DR. PAXMAN: Good morning. I am Dalton Paxman. I am the regional health administrator for the Department of Health and Human Services here in Region Three, and I welcome you to Philadelphia.
DR. COHEN: I am Bruce Cohen. I am from the Massachusetts Department of Public Health, and I am a speaker this morning.
DR. ABBOTT: Peter Abbott with the California Department of Health Services.
DR. ONAKA: I am Alvin Onaka. I am with the Hawaii State Department of Health, and I am one of the speakers this morning.
MS. BREEN: Good morning. I am Nancy Breen from the National Cancer Institute. I am in the Division of Cancer Control and Population Sciences there.
DR. NEWACHECK: I am Paul Newacheck from the University of California at San Francisco, and a member of the committee.
MS. BURWELL: Audrey Burwell from the Office of Minority Health, and lead staff for the subcommittee.
DR. MAYS: I am Vickie Mays, the chair of the Subcommittee on Populations. Maybe we can introduce those individuals who are here in the audience.
MS. ROGERS: My name is Jane Rogers. I am from the Office for Civil Rights for the Department of Health and Human Services.
MR. HO: My name is John Ho(?). I am from CMS.
MS. KELLY: Good morning. My name is Dorothy Kelly, regional minority health coordinator for this region, Philadelphia.
MR. BELMONT: Good morning. Louis Belmont is my name. I am Dr. Paxman's deputy. Welcome to the regional office.
MS. GROOT: Sherry Groot. I am the writer for the populations committee.
MS. WEEKS: Christina Weeks. I am from the Center for Health Equity Research and Promotion.
MS. JACKSON: Debbie Jackson, National Center for Health Statistics, staff to the national committee.
MS. HARSHBERGER: I am Dorothy Harshberger, from the Alabama Department of Public Health, and I am a speaker today.
MS. POTOBOVSKY: I am Trish Potobovsky from the Pennsylvania Department of Health.
MS. VENTURA: I am Stephanie Ventura from the National Center for Health Statistics.
MR. MELTZER: Mel Meltzer, Quality Resource Systems, a presenter this afternoon.
MS. WHITE: I am Gracie White, NCHS staff.
DR. MAYS: Great. I think everyone has introduced themselves. I want to briefly talk about why we are here today and what has brought us to the table, and the request to have several of you who have introduced yourselves as speakers.
The Subcommittee on Populations has a long history of being interested in the issue of minority health statistics. Part of our role has been to be advisory to HHS around the collection of data on racial and ethnic groups as it applies to health. As we have done our work over time, one of the things that we have looked at has been the issue of the collection of data for racial and ethnic groups.
In light of the OMB-15 directive, the issue still remains on the table for us, about how the data is being collected, whether the data is being collected, how useful is the data that is collected.
We have had a series of hearings which started last February. In addition to hearings, we have had individuals come into our subcommittee when we have had full committee meetings, to discuss with us many of the issues that are relevant to the collection of such data. In the continuation of that, what has become clear to us -- and many of you have been the individuals to share with us -- is that some of the issues around the collection of this data for the state is a bit different than at the federal level.
There is a relationship, of course, between the states and the feds and the collection of much of this data, but some of the issues of concern to you may differ from that which we sometimes pay attention to when data is being collected, either at the national level in a population-based survey, or that is being collected for other federal uses.
Here, what we are interested in today is exploring with you a number of questions. For those of you who are out there in internet land, I am going to actually share with you many of the questions that we have asked our state presenters today to respond to.
To begin with, would you please briefly discuss the demographic composition of the state in terms of racial and ethnic subgroups and where these groups are concentrated. Other issues we are interested in, in addition to information on racial identity, does your state routinely collect detailed data on ethnicity and national origin in its ongoing surveillance data sets.
Are these data collected only for Hispanics or for other groups as well? On what data sets are these data collected? Will all ongoing data sets maintained by your state health department use the same race and ethnicity standards for data collection, tabulation and reporting? If so, is this standard based on the 2000 Census standard, the OMB preferred standard, or some combination of both?
Do you believe that there is a significant problem of misclassification into racial and ethnic categories in the state. If so, we have asked you to tell us a little bit about what data set, what racial and ethnic groups and what steps, such as possible partnerships with advocacy groups, are in place for studying these issues and for making the corrections.
These also ask whether or not the state plans to adopt the race and ethnicity items on the new NCHS standard birth certificate. Do the race and ethnic guidelines for OMB have adequate utility for the state, or does the state frequently find the need to collect information using other or expanded categories? Are there inconsistencies in ethnic and racial data collection methodology between various instruments that the state uses to collect data, either solely for state purposes or to provide to the federal government.
Does the state have adequate guidance, for example, to do the bridging and tabulation from the older standards on the collection of race, to the newer standards on the collection of race. Does the state collect racial and ethnic data in their Medicaid managed care system? Also, if there is a statewide hospital discharge data system, does it collect racial and ethnic data?
Finally, are racial and ethnic data routinely reported in state health-related publications and reports? So, for those of you on the internet, that will give you an idea of what will be making up the substance of our hearing today.
Without further ado, what I would like to do is actually turn it over to Dalton Paxman, who is the regional health administrator for Region III, which is where we are, the Department of Health and Human Services. Thank you for having us.
Agenda Item: Welcoming Remarks.
DR. PAXMAN: Thank you, Vickie. Welcome to all of you and, on behalf of the assistant secretary for health, I do want to welcome you to Philadelphia, the city of brotherly love, and the home of the Declaration of Independence, where it was written. For those of you who are first time visitors to Philadelphia, Independence Hall is just across the street, and I encourage you to walk around. Our regional office is just around the corner. For those of you who would like to stop by, we welcome you to do so. Dorothy Kelly introduced herself, but I also encourage you to speak to Dorothy. She is our minority health person in the community. She works with the community, and she will keep you on your toes, as she keeps me on my toes on a daily basis, to remember the community.
Many years ago, when I worked in the Office of Disease Prevention and Health Promotion on Healthy People 2010, we recognized the importance of these kinds of data on racial and ethnic minorities, population data. That is why we chose one of the two goals for Healthy People 2010, is to reduce and eliminate health disparities.
Your data, this meeting is so important for us to achieve that goal, and I think it is very important. While I worked in Washington, D.C. and on Healthy People, these data were always something that were kind of abstract. You know, they were data but you never got a feel for them.
Coming out to the regional office, where we work regularly with state and local health departments and the staff, and working with the community, you understand why these data are so important, and why the work you are going to do today, and why the work you do on the subcommittee is so important. You use those data. You use those to target your interventions for public health.
When I came here three years ago, we knew there was a national epidemic of asthma, that the rates of asthma were getting worse, and that was through our Healthy People 2000 project. It was the racial and ethnic data we collected on it in cities that suggested that the health disparities for asthma were even more of a problem. In Washington, D.C., for example, African American boys were five to seven times more likely to die from an asthma attack than their white counterparts. Here, in the city of Philadelphia, the prevalence of asthma is about 10 percent. In some of the communities, the minority communities, the rates of asthma go up to 20, 25, even 30 percent. I think it is those data that helped us focus our efforts on asthma initially. We are now doing the same things with obesity, where the health disparities are also becoming clear. So, your work today is very important and I do just want to welcome you to the city and encourage you to have a wonderful and productive day. I look forward to hearing from you.
Actually, Dorothy will be here most of the day. I unfortunately have to leave for part of the day, but I will come back. Dorothy will keep me informed and I assure you, this is very important. So, have a wonderful day. I don't want to take up any more time because you have such a full agenda. So, thank you.
DR. MAYS: Dalton, thank you very much. Dorothy, thank you also, and we are happy to hear that you will be with us. Dorothy has been an asset in making sure that the word got out that we were here in Philadelphia. So, we appreciate both of you in your efforts helping us with the meeting today.
What I would like to do is actually to get us started in terms of our presenters that we have. The way in which we typically do this is that, we will have a set of presentations, we will take questions from the committee, and then we will open it up to anyone who is in the audience. So, let's get started. I would like to start with Bruce Cohen from the Massachusetts Department of Health. Bruce is the director of the Division of Research and Epidemiology, and will start in terms of our state presentations today.
While he is getting set up, I will just make a comment or two about him. He is known as an avid manager and researcher in the areas of chronic disease, minority health surveillance, social demography, perinatal issues and health services research. This is an area that I know, from having had discussions to him, is near and dear to his heart, in terms of some of the issues of collection and use of racial and ethnic data. Massachusetts has been one of our states that has been quite active also in contributing to the scientific literature in this area. Bruce?
Agenda Item: Massachusetts Department of Health.
DR. COHEN: Thank you very much, and thank you for inviting me to testify before this subcommittee on a matter of great personal and professional interest to me, and a matter of vital importance to our public health department, and I think for public health throughout the United States.
I will cover three topics today. First, I will discuss subtle and not so subtle differences in the states' needs and uses of race ethnicity data compared to the federal government, which I think the key to understanding this issue from the state perspective.
Secondly, I will present what we have been doing in Massachusetts at the Department of Public Health and, third, I will highlight some key issues for state departments of health in general.
The first two questions that I would like to address are, one, do the revised OMB standards work for states and, two, what are the inconsistencies between the state and federal methods and needs.
As you know, there are five race categories in the 1997 revised OMB standards, as compared to four in the 1977 OMB-15 directive.
At the state level, neither four nor five categories are enough. There is more heterogeneity within each category than between categories. For states, I think ethnicity may be more relevant than the concept of race.
By ethnicity, I mean national origin, ancestry and heritage. For example, OMB uses the broad race category, Asian. The more meaningful ethnicity groups would be Japanese, Chinese, Cambodian, Hmung, Filipino, Asian Indian, et cetera.
When individuals self identify ethnicity, these descriptors are less likely to change over time. They also reflect important linguistic and cultural variation, masked by the broader race categories.
The census calls some of these subgroups, that I call ethnicity groups, race groups on their collection form, and I will discuss that issue a little later.
OMB also has declined to identify Cape Verdians as a separate race group, keeping Cape Verdians with blacks in their tabulations.
This is a huge issue for Massachusetts, where Cape Verdians represent a large ethnic subgroup, particularly since less than 25 percent of Cape Verdians self-identify themselves as black.
Seventy-nine percent of Cape Verdian mothers, on our birth certificates in the year 2000, identified their race as other.
In the 2000 census, the Hispanic ethnicity question precedes the race question in an attempt to get better reporting on the race question.
This really didn't work. Almost one half of Hispanics report their race as other, accounting for 90 percent of the other respondents nationwide.
In Massachusetts, approximately 53 percent of Hispanics identified themselves as their race being other.
Notice that the census lists more than five race categories, as I mentioned. That is because the census lists selected Asian and Pacific ethnicity groups that can be re-aggregated into the minimum race categories.
The census designation of these particular check box categories for the implementation of the revised standards may have unintended and, in my opinion, negative consequences for state and local public health programs.
If the state and local agencies are compelled to adhere to these check box categories, we may end up being required to collect data that are less relevant to our public health program needs, rather than focusing on categories that are more meaningful for targeting our interventions.
Let me explain why. Many surveillance and programmatic data collection instruments, at the state level have severe space constraints and are crafted to minimize extraneous data collection.
In Massachusetts, less than one in 10,000 persons are Hawaiian, Guamanian or Samoan. Why require that these categories be included in data collection and data tabulation. It will collect no meaningful data, while creating resistance among substantive program staff and data base designers, and use valuable limited space on our hard copy forms.
Let me step back from the data details for a moment to provide a broader consideration of the context of some of these issues.
What do states really need? First, states need detailed ethnicity data, rather than broad race data, as I have been discussing.
Also, many states have unique populations that we wish to track, for instance, Cape Verdians and Dominicans in Massachusetts.
Second, states need intracensal small area race denominators for race calculation. Most important, states need consistent guidance from all federal agencies, for collection issues such as dealing with missing data, for using an other race category, for converting multiple responses to single responses, for bridging and trend analysis.
Indeed, some federal agencies include an other race category. Others do not. There are differences between CDC's common data element definitions and the OMB approach, and there is no DHHS consensus for tabulation.
I think both HRSA and SAMHSA are using a single multiracial category for tabulation that obscures the intent of allowing individuals to select more than one race.
So, there remains substantial variation and inconsistency across federal agencies and national organizations.
There are several major reasons why state needs may differ from federal approaches. First, a much higher proportion of federal data are collected in face-to-face situations using trained interviewers to elicit self reports.
States use more secondary data sources, observer and proxy reporters, medical records, and less trained intake workers to determine race and ethnicity.
Second, major federal interests aren't producing national and state level data, whereas small area data are needed by states and local health agencies.
Third is probable differences in approaching toward bridging, which I will talk about shortly.
To summarize my response to the first two key questions, the OMB standards don't have adequate utility for the states, and guidance from the federal government has been inconsistent.
What are my recommendations? Well, states need uniform guidance from federal agencies, both for data collection and tabulation.
We need, as states, to develop partnerships with federal agencies to address technical issues around bridging and small area estimation.
I feel states should emphasize ethnicity rather than race, and we should focus on self report.
I would like to switch gears now and present the Massachusetts Department of Public Health experience.
The first question I wish to address is what ethnicity data do we collect in Massachusetts.
In January, 2001, we did a department-wide review of 76 data sets, six population surveillance data sets, 18 disease-specific surveillance sets, 32 programmatic data set, and 20 other reporting systems.
Eighty-five percent of those data systems collected either race or ethnicity. Fifty-five percent collected the item separately. Seventy-four percent had an Hispanic identifier, with 27 percent having more detailed Hispanic subgroup information. Approximately 36 percent had information on other ethnicity subgroups.
This survey represents the data sets prior to the implementation of the revised standards. I have details that I will share with the subcommittee of the results of this survey.
The fundamental data collection questions that all states need to answer are, how will we use the OMB and census categories, and will we adopt the NCHS recommendation for our birth and death certificates.
In Massachusetts, we are planning on going beyond the OMB and census approaches, as well as implementing our own strategy for vital statistics, race and ethnicity data collection.
We are focusing on two possible options that will meet the OMB requirements and the NCHS minimum recommendations, while providing our state health department maximum flexibility to collect the data that we need.
The two options that are under consideration are a three question check box approach, and the second option is a modified text approach.
Question one of the three question check box approach is an expanded version of the census question, to identify persons of Hispanic ethnicity.
The starred categories on this slide indicate categories that are used by the U.S. Census and recommended by NCHS on a standard certificate.
Question two is really where we deviate the most from the current OMB, NCHS and census approach. It provides data that meets our state needs.
It is a check box question collecting detailed ethnic group information for ancestry categories important for public health in Massachusetts.
Question three follows the exact recommendations of the OMB minimum standards for identifying race. It is a selection of race allowing respondents to select more than one.
We are also considering an option based on an approach taken by California and Hawaii and others for the collection of Hispanic and race data.
This approach essentially is a free text approach for collecting ethnicity data. We would record text responses for question one on Hispanic ethnicity and, on question two, for other ethnicity groups.
Question one meets the OMB revised standard guideline, while question two provides us with the detailed race and ethnicity information we are interested in.
Question three would be the same as in option one, a check box category to identify race.
Other possible questions that staff are interested in, in our health department, are a question about language preference, a question about country of origin, and a question about migration patterns to the United States.
So, Massachusetts' current plans are not only to meet our federal obligations to collect race and Hispanic ancestry, consistent with revised the OMB standards and NCHS recommendations, but to use this change as an opportunity to implement the data collection strategy that focuses on collecting more detailed ethnic groups.
The next question is, do states have adequate federal guidance on bridging. The simple answer is no. I don't feel that states have been given enough support, but we have tried to implement our own strategy in Massachusetts. Let me describe that briefly.
In order to address the problem of bridging, the Massachusetts Department of Public Health has developed our own population estimates for 1999 and 2000 that convert the new census rate categories into denominators consistent with the numerator race data collection systems still in place throughout our department.
We have reallocated persons of some other race, and two or more races, based on the empirical proportion of the single race groups and each race combination. This is called fractional assignment by OMB.
The race denominators were calculated for our four modified OMB-15 race categories, white, black, Asian and Native Americans.
Reallocation was done using the proportion of the races at the town level in Massachusetts. Thus, our final population file for each year consists of 101,088 new cells, the 351 towns times the 18 five-year age group, times the two sexes, times the eight race ethnicity categories.
Also, I brought for the subcommittee a detailed explanation of our methodology, for your review.
Do states have sufficient guidance for tabulation of data once they are collected? There is reasonable information suggested by OMB, but I think it is important -- I think it is actually imperative -- that a unified federal statement be made supporting this OMB recommendation.
Essentially, the approach that we are recommending in Massachusetts, consistent with OMB, is the tabulation and reporting of all single race categories, plus all multiple race combinations with at least one percent of observations in a data set. Then we will collapse the less frequently reported combinations into one category.
Of course, we will strongly encourage the presentation of the detailed ethnicity data that hopefully our department will be collected.
Are racial and ethnic data routinely reported in state health-related publications and reports. I think we are extremely proud and committed to the publication of race and ethnicity data in Massachusetts, both in hard copy format and on line directly through Masschip, our online information service, where persons can access data from 50 DPH and other related public health data sets, that has race information and detailed ethnicity information on many of those data sets.
Here is a list of some of the publications that we generate. They are all available on the web. Certainly, our annual surveillance reports, and we focused on special minority birth reports that look at the heterogeneity of subgroups within Asians and Hispanics. Next month, we will be publishing our first report on blacks. I brought copies of several of those reports to share with the committee as well.
We generate routinely health status indicators by race and ethnicity, and in fact, we are very proud of working on a publication on the web that will examine the leading health indicators, tracking the Healthy People 2010 indicators by race.
In addition, we provide these revised modified denominators in our population data.
To summarize the Massachusetts response to the specific questions the subcommittee posed, MDH will continue to focus on a detailed array of ethnicity groups while possible.
We will standardize our approach throughout the department to be consistent with the revised OMB guidelines, while developing our expanded approach for implementing our vital statistics certificates.
Differing federal mandates, coupled with specific program needs and data collection modes pose challenges for consistency at the state level.
We will continue to use and expand the use of race and ethnicity data in our surveillance reports.
I would like to leave you with several concluding thoughts. First, I think it is crucial that we consider how race and ethnicity data are being used.
We need to link these data to other measures of SES, linguistic and demographic indicators.
State responses to the revised standard are made difficult by the multiplicity of state data sets and data collection modalities that the states use.
The state perspective and needs do differ substantially from the federal perspective. We need to focus, at the state and local level, on the heterogeneity of ethnicity and small area data for service delivery and surveillance.
The states also are in desperate need of consistent federal agency guidance, and we need to be able to apply that flexibly to meet our own needs.
The final point I wish to reinforce is that much of our discussion about race data collection grows out of our consideration of the OMB directives.
The development of these standards stems from responsibilities to enforce civil rights laws. Data were needed to monitor equal access in housing, education and employment, for populations that have historically experienced discrimination.
While this approach may serve well for those functions, as social scientists and public health researchers, we need to strongly consider whether there are better approaches for collecting race and ethnicity data, for health surveillance activities, and for targeting interventions to end health disparities. Thank you very much.
DR. MAYS: Thank you, Bruce. Our next speakers are from California. I don't know exactly how we are proceeding.
Peter, are you willing to switch and start with the California health interview survey and then, as we determine how to get Jane on line, we can have her follow you.
Okay, Peter Abbott is going to talk about an example of the collection of such data by the California Health Interview Survey.
Peter is the chief of the Office of County Health Services, the California Department of Health Services. He has been in public health for a number of years.
He has been the administrator of a variety of programs supporting county health services, was actually part of the negotiations with the legislature over the use of the Proposition 99 tobacco surtax, that we happen to have in California.
Agenda Item: California Department of Health.
MR. ABBOTT: Thank you. It is also a pleasure and honor to be here and to address you. I agree with what Bruce had to say. California, as you know, is a very heterogeneous population. As of the 2000 Census, there was no majority population in California.
Jane McKendry, of the Center for Health Statistics, her presentation was designed to actually address the questions that the committee asked of the presenters.
Our experience is very similar to that of Massachusetts. So, we support and emphasize the recommendations of Bruce Cohen.
What I am here today principally to talk to you about is the California health interview survey. The California health interview survey is something that was first done in 2001, and represents a significant effort on the part of several organizations and multiple funders at both the federal, state as well as philanthropic level, to generate health survey information for California.
As the slide indicates, our need for these type of data was for policy analysis, development and advocacy, service and program planning, as well as research.
The three major partners were the UCLA Center for Health Policy Research, Dr. Rick Brown and his colleagues, the California Department of Health Services, and the Public Health Institute in California, located in Berkeley.
We had a significant planning process extending over three years, involving multiple organizations and many people. We have an advisory board, as well as technical advisory committees.
The funding partners were several. The California Department of Health Services put up approximately $4 million. Our Prop 10, families and first commission, put up $2 million. The National Cancer Institute, Nancy Breen's organization, was very generous in putting in more than $2 million. The Centers for Disease Control and Prevention was a major contributor. The Indian Health Service was a collaborator, not so much in terms of money, but especially in terms of helping us develop an over-sample for Alaska Natives and American Indians in California. Lastly, the California Endowment, a philanthropic foundation.
We basically had, in CHIS, a significant survey design to try to develop county-level data, as well as data on our minority subpopulations.
We had an adult and adolescent, as well as a child, the most knowledgeable adult on behalf of a child, in a household.
So, what I am going to quickly do is just demonstrate to you what the topics of these questionnaires were. Starting first with the adult questionnaire, extensive demographics, health insurance coverage, public program eligibility and participation, utilization of health services, access to services, including barriers, health status, activity limitations, chronic health conditions, health behaviors, mental health and dental health.
The adolescent survey was actually done with the adolescent after obtaining parental permission. Again, just quickly, health status, demographics, behavior, perceived parental knowledge of behavior, adult supervision outside of school.
Then, in terms of child, just quickly, chronic condition, management of asthma as well as attention deficit disorder, health behaviors, child care arrangements, early childhood development, and parenting.
As you can tell, this was a significant attempt to gather health information and, obviously, one of the critical factors in California was data on race and ethnicity.
As I mentioned, we had dual objectives here. One was to develop county-level data. The other was to develop it for the major ethnic groups, and several subgroups within those populations.
In order to do that, it was a massive survey, over 55,000 households. We have 41 separate strata for sampling. These included 33 of California's 58 counties with their own individual strata, and then, in others, we grouped counties together.
As a result, we have a very robust sample, because of California's diversity, in terms of representation of the listed racial and ethnic groups.
CHIS was a random digit dial sample. Obviously, therefore, we had to do some things differently in terms of talking to people rather than having them fill out forms.
They ended up with slightly over 55,000 adults, nearly 6,000 adolescents, and then some 12,600 children, and this was the most knowledgeable adult.
Again, it does reflect California's diversity. We did do some over-sampling of certain subpopulations to ensure that they were adequately represented in the sample.
One of the issues is, we did attempt, in the California health interview survey, to be relatively consistent with what was done in the Census and ultimately with the OMB guidelines.
Therefore, we did have two initial questions, one about Hispanic or Latino descent, and then proceeded after that into the race and ethnicity.
One of the issues, of course, is how to get back to a single variable on race and ethnicity. What I am attempting to do here -- by the way, this is somebody else's presentation that I am just trying to give in a semi-coherent fashion, so please, bear with me.
UCLA basically wanted to create a single variable which combines the Latino and race and other ethnicities. Basically, in terms of the questions that were asked, you can see what was asked in CHIS on the left compared to the census on the right.
Again, moving through the questions, you can see that there is a rough parallel in terms of what was done in the census.
Then, what we have done here, the results, I believe, were, the first breakdown, are you of Latino descent or are you not, this was the breakdown in California.
Then, when we added the additional questions, this is what we got, in terms of the Latino-non-Latino. Then they were also asked about their race. Both the Latino was asked, as well as the non-Latino.
What you can see here, we did have a significant portion of the sample say they were of two races, approximately 4.5 percent.
You can also see that, amongst the Latino group, and emphasizing what Bruce said, most Latinos said, I already gave you my race and then, when pressed, said, well, I am an other.
For adolescents, similar numbers. Here, we have a higher proportion of Latino adolescents, again, reflecting California's population.
Again, if we break that down, you can see that, amongst the Latino, most of them reported other race. There was a number of two or more races. Again, you can see the distribution and the diversity, again, reflecting California's population.
For children, similarly, in the Latino category, reflecting California's demographics and birth data, a significant number of parents reporting Latino, and then non-Latino 60 percent.
Similarly, if you break it down, you will see, again, the heterogeneity of the population and again the two or more race reporting is now up to 8.5 percent.
What we tried to do in terms of constructing a single variable for race and ethnicity within the CHIS data base, this is where we were trying to move.
Basically, we also had a question as to which of these races or ethnicities do you most identify with. Again, we also had a question asking country of birth.
As far as constructing the single race ethnicity variable, what we have is these decision rules here, which is that if there is a non-Latino individual of a single race, they were assigned to the reported race.
If it was a non-Latino with multiple race assigned, they were assigned to the one they most identified with.
If they reported that they were Latino and a race, they were assigned to the most identified with category of race and, if they were Latino and a single other race, it was as described here.
As far as American Indian Alaskan Natives, we also asked information about tribal identification and then the decision rules, again, as listed here.
So, this is the results, I believe, which is that initially, there were 11,840 adults who said that they were of Latino descent. Then, amongst those who were non-Latino, that was the breakdown.
Then, in terms of how we distributed the Latino population, according to the decision rules and also dealt with the other and two plus, this was the following breakdown on the lower level.
Similarly, with adolescents, these are the numbers. Similarly, with children, this is how the results, using this methodology, came out.
One of the things that was very important to us was that we had to adapt the CHIS questionnaires in terms of cultural adaptation.
We also administered CHIS in six languages that are listed there on the chart. To demonstrate the importance of this, we show here the number of interviews that were conducted in a language other than English.
Again, in adults, the number is significant at 12 percent. Then, in terms of the child interviews, the most knowledgeable adult, usually the mother, 20 percent of the child interviews were done in a language other than English. In adolescents, as one would expect with the assimilation and acculturation, it was still a significant percentage, but less than the other two.
This shows that, in terms of target groups where we oversampled, again, it shows the RDD sample, the over-sample, where we went to over-sample the populations and, again, the percentage that were done in the language of that particular target group.
I will talk a little bit about CHIS. Actually right there, next to Bruce, are my bookmarks on CHIS. We have had a major effort in terms of data dissemination involving development of public use files, development of confidential files for our local health departments, as well as the state health department.
We are producing a variety of reports. Then, we have a very nifty internet query-based system called Ask CHIS. So, that information is on those little pieces of paper, the bookmarks, that are being passed around.
We also plan to do a bunch of focused reports on asthma and health insurance coverage. Food insecurity, a policy brief, is coming out on that. This is, I understand, the politically correct way of talking about hunger and not having adequate food.
Forthcoming reports will be on diabetes, cancer, and health and development of young children. This is a particular interest of our Proposition 10 commission.
Ask CHIS, as I mentioned, is really an excellent way of accessing the CHIS data base. These are population estimates for the jurisdiction of interest. You are going to get county level data, statewide data, multiple county data, that basically allows you to construct a table as to your choosing.
To subset that, it is very user friendly, and hopefully will become even more so. We are now in the second generation of this.
So, please do, if you have the time and the inclination, go to AskCHIS.edu, and we are very much interested in your feedback in terms of the system and the data base and things like that.
We are planning another cycle of the California Health Interview Survey in 2003. Again, it should be of the same magnitude in terms of 55,000 households and, hopefully, we will continue this on a biennial basis.
Again, a little bit more electronic data files are available, public use files. There is a data access center at the UCLA Center for Health Policy Research.
The Department of Health Services, our Center for Health Statistics, is our broker for data sets for state agencies and persons within the state health departments, as well as our local health departments.
We currently are doing workshops throughout California, both oriented toward researchers as well as other types of users.
We have been holding some invitational conferences. We held one on African Americans in May, one on Latinos in August, and Asian American and Pacific Islander in December, upcoming, and then another one in January on American Indians and Alaska Natives.
These are the main forces behind CHIS, Dr. Richard Brown, Dr. Charles Desogra. Inoza Ponce(?) is the individual who has really helped us a great deal in terms of our cultural adaptation, our collection of race and ethnicity information, as well as the design and development of our single race ethnicity identifier.
So, that is, I believe, the end of my presentation, and I just want to personally say that doing the California Health Interview Survey represented a 15-year quest for me.
It has been very difficult to secure the funding, to work through the technical and procedural aspects of doing something of this magnitude.
Again, given California's diversity, that is why we have federal partners. We have numerous states that are utilizing the data base, and at some point in time, we hope to do small area estimates and additional capacity, so that these data may be utilized in other jurisdictions throughout the country.
I would be very pleased to answer any questions. I guess I wait until discussion. Okay. Have we managed to connect to Jane yet
DR. MAYS: Are we close? If not, we will move on to Hawaii, while you all continue to connect. While you do that, why don't we move on to Hawaii. You have still got to plug and dial.
MR. ABBOTT: One other thing that I did pass out is a 1994 report that we did. It just emphasizes two things, really, that Bruce mentioned.
One is the heterogeneity within the major racial classification, particularly in the Asian American, Asian Pacific Islander populations.
This presents the data from 1990 and shows considerable diversity as you disaggregate those categories, whether it is Latino Hispanic or Asian American category.
We are actually repeating this study based on 2000 data, and we can demonstrate that, as well as the importance of denominator data. Without denominator data, it really frustrates our capacity to do these kinds of analyses.
DR. MAYS: I think what we have done is pass it around. I think that our partner here, Dorothy, made additional copies. We will make sure that everyone, before they leave, has a copy of that information.
I tell you what, we are going to do Hawaii, while you all do technical stuff. Alvin, can I call upon you to start us off with your presentation? Thank you.
We are happy to have Alvin Onaka. Welcome, Alvin. Alvin is the state registrar of vital statistics and chief of the Office of Health Status Monitoring for the Hawaii State Department of Health. He has been in this position since 1989.
I also know that he is currently on a committee of the Institute of Medicine, in which they are looking at some of these very issues, of trying to determine whether there is adequate data collection on race and ethnicity in the federal data sets.
I think he brings us a state perspective and also a broader perspective. Thank you, Alvin, for being here today.
Agenda Item: Hawaii Department of Health.
DR. ONAKA: Thank you very much for the opportunity to share with you some of the things we are doing in Hawaii.
What I would like to do, first of all, is just give a brief overview of the race and ethnic collection of data in Hawaii.
First of all, the first slide indicates that the 2000 census indicated we had about 1.2 million residents in Hawaii.
They were distributed among the seven inhabited islands in Hawaii, from the northwest being Nihal, then the island of Kuai, the island of Oahu, which is where Honolulu is, the largest island, then Molokai, Lanai and the island of Maui, then the big island, which is the island of Hawaii, which is the second-largest island.
What I wanted to get into what at least the mainlanders use the word ethnicity for, and that is Hispanic origin. In the 2000 census, we had about seven percent of the people indicated they were of Hispanic origin. Again, 93 percent indicated they were of non-Hispanic origin.
What I think is different, maybe, than on the mainland United States is that, when we had race in combination, Asians made up 58 percent, next being white, then the third being Native Hawaiian or other Pacific Islander. Black was fourth, and American Indian Alaska Native followed.
I wanted to show, again, a little of the geographic distribution. Here, again, around the distribution of the islands, Oahu had the largest concentration who were Asian alone.
Again, on Molokai, Lanai, Maui and the big island were where there was the biggest groups that were white alone.
These are the individuals distributed over the islands who were Native Hawaiian or other Pacific Islander alone. The black population in Hawaii is concentrated on the island of Oahu, where much of our military residents reside.
Interesting enough, in terms of the percentage of those who indicated that they were of American Indian or Alaska Native, they were on the neighbor islands.
What I think is very significant in our demography is that, while in the 2000 census there were, in the United States, 2.4 percent were of two or more races, in Hawaii, 21.4 percent indicated that they were of two or more races.
Again, I am using the terminology as I guess it is used on the mainland of race and ethnicity. I agree with Bruce and Peter in terms of, again, the concept of ethnicity which includes culture, language and immigrant status, in many ways, is a better indication of what we are trying to measure.
Again, this is a distribution among the islands, of those who are of two or more races.
What I also wanted to show was, again, in the concept of ethnicity, that our foreign-born population was about 20 percent. I think it is an important aspect to know the migration status. It is an important element here in looking at race and ethnicity.
Again, this is a distribution of those who are foreign born, the highest concentration being on Oahu, which has the largest population, and the capital of Honolulu.
What I wanted to do next was give some idea -- before I go into that, I think there was one slide missing that I wanted to bring in, and that is an important component, I think, of race and ethnic data collection is language, and language spoken at home.
In the 2000 census, 27 percent of households in Hawaii indicated that their primary language at home was other than English. Of that 27 percent of households, 88 percent of the other than English are Asian and Pacific languages.
Again, I think this is very important, especially when we look at access to care and various other components. Again, we have done a lot of studies in this area.
It is not enough to know that the language is Asian, but the various dialects. I would like to use the Filipino language as an example here.
The census indicated there were more Tagalog speaking Filipinos in Hawaii than Ilokano. However, when we have these help lines where people come in to ask for language assistance in health care, Ilokano, the Filipino dialect, was the most requested.
It really does reflect the pattern of migration and other aspects of our historical origins here in Hawaii. That is, the Ilokanos was the region in the Philippines where our earliest immigrants came, and they were the less educated individuals.
This also, I think, is a reflection of the migration patterns to Hawaii. The Tagalog speaking ones, as you know, in the Philippines, they have eight years of English.
It was really different from when the people from the Ilokanos came to Hawaii. These are the individuals in the older generation who are seeking health care services.
Even our data, sometimes we have to dig deeper in terms of understanding languages. Again, we are very envious of California being able to do surveys where they have the resources to ask these questions in many languages and dialects.
I would like to move on to the questions we were asked with regard to what data sets do we have. The Department of Health, in particular, has been collecting multi-race data in Hawaii for over 100 years.
I have records that go back to 1840, our vital records, with individuals that are indicating that they are of multi-race.
However, prior to the year 2000, to satisfy federal requirements, we had to really aggregate our multiple races into the requirements that the National Center for Health Statistics asked us to do, in putting it into one race.
Again, the office I work in was created in 1896, and even from then we collected ethnicities, the multiple ethnicities.
We have a long-standing evaluation program to assess the quality of the information that we have been gathering for, again, over 100 years.
The first is that, in cooperation with the federal program, we have a birth and infant death record linkage. We have used that to see -- we hope that the race of a child would not change within the first year of life.
We were especially interested, especially when the article by Hahn came out in JAMA that indicated, at least on the mainland, that many of the children were born Asian but died white.
So, we really checked our information and we have been checking it all along. The concept of one's ethnicity, as we do annually, of the 124 infant deaths that we had in the year 2001, we couldn't match only four of those.
We found out that the reason was that the child was adopted, by parents who were of different race than what the child was born as.
Secondly, we have an annual health survey. I guess before I get to that, maybe I should emphasize that, although it indicated that we had only 22 percent of the individuals in the 2000 census that were of two or more races, when we look at our birth data, that is, the core that just have been born in 2001, over 50 percent of these children indicate two or more races of the parents.
So, you will see, in the subsequent censuses, that 22 percent is going to go up. It is actually much, much higher.
The second way that we look at the quality of data, or the question that the committee asked us was misclassification, we have been collecting since 1968, we have been doing an annual health and demographic survey.
Again, I will further describe this because, in some ways, it is similar to what California has been doing. We have been doing this continuously for the past 34 years.
One aspect of it is looking at whether races are being misclassified, if you may. Interestingly, we were asked recently by the Indian Health Service to look at how many American Indians and Alaska Natives there were in Hawaii.
Our health survey had consistently come up with about 22,000 of our residents being of American Indian Alaska Native ancestry.
In the 1990 census or the censuses prior to that, they had indicated there were about 6,000 American Indians in Hawaii. Again, our survey showed that there were 22,000.
Interesting enough, when the 2000 census came out, there were about 24,000 individuals who had indicated that they were of American Indian ancestry.
In many ways, it validates that when you open it up to having people indicate what races they were, if you may, that there were many, many individuals who were American Indians who were also, say, white, or Native Hawaiian, and had to choose one of the two.
When they were given the opportunity in the 2000 census to indicate their multiple races, we were able to get a figure that our surveys were telling us all along.
Besides vital statistics, the government race and ethnic data have been collected in almost all the surveys that we do, at least by the state governments. It is an essential variable, a very important one, especially in Hawaii to the agency, social services and managed care, and also with regard to the hospital discharge data set, does include race and ethnicity.
I think one way that it may differ from the surveys that I will be talking about is the quality of the information and how it is gathered is a little different.
For instance, the managed care, the Medicaid data, as well as hospital discharge, the information on race and ethnicity are gathered at time of enrollment instead of at each encounter.
The method in which it is collected, because it is not collected by the health department, is that at times it may be by the observer and at times by self identification. So, there is still much improvement that needs to be done as far as how the data are collected by agencies in Hawaii.
I would like to move on to one of our major surveys, as I said, which is the Hawaii health survey. That survey is modeled after the National Health Interview Survey.
For historical purposes, Peter Hurley, from the National Center for Health Statistics, spent a sabbatical in 1968, I think, at the University of Hawaii, and helped us, at that time, develop the Hawaii health survey, which is better.
It was first conducted in 1968. In 1996, we changed over to it being a telephone survey. As opposed to 55,000 households in California, we are proud of 5,000 for our population. It is a fairly large sample.
We collect health and demographic variables of all persons in the state except for those who are in group housing, as well as it has been changed to a telephone survey, and we do not have households who do not have telephones, including the homeless.
We have information on gender, on age, race and ethnicity. We ask that question in many ways, not just one way.
As in our birth certificates, I should say, too, it is an open ended text question. That is, we ask what races they are. They can list, I believe for each parent in the birth certificate, we limit it to eight entries. So, including the census six and whatever else they want to indicate that they are.
This is important in Hawaii, because there is a proportion of individuals, even in the 2000 census, who said they were of all six race categories, including other.
We also asked what the race and ethnicities are of the respondent's parents, so we get another look at their parents, as well as asking the question, what race do they most closely identify with.
We have worked with the National Center for Health Statistics in trying to keep our questions similar, so that we can compare it nationally.
We ask, again, educational attainment, that again being a very important variable in any race and ethnic collection of data, especially as I think I indicated, closely linked to language.
You know, you can have the best sort of material in the dialect and language of the individual, but at the level of education, it may not be understandable, especially using more technical and medical terms to describe what their health status is.
We also ask for marital status, and questions that we have been asking for some time with regard to health insurance status and health conditions.
Also, as Peter pointed out, food insecurity is a hot topic that we have been dealing with in the health survey.
The most common races that we collect data on -- again, I want to indicate that they are open ended. So, people can put down whatever they do.
Interesting, under other, Portuguese is one that many people put down. They do not consider themselves white. They are not the Cape Verdians as in Massachusetts. Again, it is a reflection of the culture and history of the plantation laborers that were brought into Hawaii to work the sugar and pineapple plantations.
What I wanted to show here is race in combination here. It really is a reflection and we need to break down the Asian Pacific Islanders more specifically here.
It has to do with the migration status, the place of birth and, really, language here. You can see for the first, the Japanese are the less, in many ways, mixture. It does reflect the way that individuals came.
I don't know if you know the concept of picture bride, but many of the Japanese who migrated to Hawaii then went back and had picture brides. That is all they saw, was a picture of the individual, and married.
In contrast, the Chinese were male laborers who came over that did not bring wives. They may have had a wife that they left in China, but they also married in Hawaii to the native population.
Again, it is reflected when we ask for the multiple race, that you can see this. Again, maybe going all the way to the right of the table, American Indian Alaska Native, you can see what I was talking about there, that although we have very few who say they are only American Indian Alaska Native, our population has many who are of American Indian, Alaska Native and another race.
The health conditions that we collect, among others, are those that we do not have registries for, and it is to supplement. That is why we don't have cancer questions, for example, because we are a SEER state and we have a tumor registry, but we have asthma, arthritis, high blood pressure, high cholesterol, the SF12 questions on physical well being and obesity.
What I wanted to do was sort of illustrate why one needs to collect this type of information. In looking at American Indian and Alaska Native alone, they appear to be much more obese than those who indicate they are American Indian or Alaska Native only, and some other race.
True for blacks. Interesting, the Native Hawaiian population, those in combination, are more obese than the Filipino, the white, the Japanese and the Chinese.
In terms of high blood pressure, the Japanese have very high blood pressure. It appears that those who are Japanese and another race have lower rates of pressure. The same with the Chinese, the Native Americans, blacks, Filipinos.
In terms of the question that the subcommittee asked with regard to did we routinely disseminate and report our data, yes, we do. We have for decades.
Most of our dissemination now is electronic. At our web site, you will see all the studies that we do, we have race and ethnicity embedded in them.
One of our more recent reports on food insecurity has it by race, to indicate the differences.
What I wanted to do was focus a little bit more somewhat on methodology, as well as the challenges that we do have with regard to the collection of race and ethnicity in Hawaii.
I think one of the largest challenges that we have relates to the comparability to the race numerator to race denominators.
That is, we are collecting information, and will be using a modified census OMB data collection instrument but, as I said, we want to continue to collect that diversity. So, we are going to leave it as a text where people can indicate.
For those agencies that want to use a check box, we are recommending that, again, also for bridging purposes, that they can use the box, but indicate numbers, if you may, of what they more closely identify with.
If they are Chinese, Japanese, Filipino -- and we have many -- which one do they more closely identify with, so we have some type of bridging aspect to do.
The real problem that we have in terms of having data for state purposes, we do want small area. The census now is collecting information for multi-race only every 10 years. States need to have information for intercensal periods.
That is why we have been collecting and have been doing our health interview survey annually. That is how we hope to get the denominators to be comparable with the numerators that we collect.
I think that is a real challenge and we need a partnership. We need also to work with the federal government here, because of the issues of bridging, of comparability, and of getting much more detail of the subgroups or the ethnic groups, as we would prefer that the data reflect.
I think I have used most of my time here, so I will end here. Again, I am very appreciative to get to share information of what we are doing in Hawaii.
DR. MAYS: Thank you very much. I think we may have California back on the line. Jane, can you hear me?
MS. MC KENDRY: Yes.
DR. MAYS: We have to apologize. We had a little technical difficulty. It is a long way to California, but we got you and we are very happy to have you. I appreciate the fact that you have been able to wait.
Let me introduce Jane. She is the chief of the vital statistics section in the Center for Health Statistics in the California Department of Health Services.
She has been in that position for about eight years and one of her special areas of interest is actually the issue of race ethnicity data collection and report methodology. So, I think we are in good hands. Jane, are you ready? Thank you.
Agenda Item: California Department of Health.
MS. MC KENDRY: So, folks out there have the handout. So, I will just pretend the handouts are slides, and just go through them as though they were slides, and you all can follow along that way.
First, I am going to talk a little bit about California demographics and the data needs that we have because of that, and then touch on some guidelines that we have been developing to try to help various health programs standardize their data collection practices, which are now very much all over the map.
Then, I have some graphs to kind of illustrate our collection and tabulation issues.
First of all, with the 2000 census, California does not have a racial or ethnic majority. The way we have historically lumped and split the Hispanic and race issue is to have mutually exclusive categories where we put Hispanics of all races into one category, and then distribute everyone else into race categories.
So, we have a single dimension and we call that race ethnicity, and that is what this first pie chart shows, in terms of what our population is like that way.
So, we have got 32 percent Hispanic and then a pretty good chunk of some of the other race groups. We have got 2.2 percent of the census folks who were not Hispanic who reported two or more races. We also have some Hispanics who reported two or more races. That, with our tabulation methodology, they are just subsumed in the Hispanic group.
Our Hispanic group, going on to the next slide, is overwhelmingly of Mexican origin. It is almost equivalent, in most people's minds, that Hispanic means Mexican in California, which isn't quite true, but enough so that we really want to take that Mexican group and break it down more.
There are huge differences in health outcomes and culture between Mexicans -- well, the easiest way to analyze it is by Mexicans who were born in Mexico and those who were not, as kind of a proxy of acculturation.
For example, mothers of Mexican birth have much better health outcomes for their babies with utilization of much less prenatal care. So, you have definitely got different program issues.
It is a denominator issue for us as well. We don't have the ability to look at rates for Mexican born versus native born. So, that is one problem we have got.
Moving on to the next slide, our Asian subgroups are just hugely diverse. We have got about a quarter Chinese and a quarter Filipino and, as you can see, a wide variety of other groups.
Then, just huge differences among those groups. We will have maybe Chinese and Japanese who are third and fourth generation and then mom who is still not speaking English and has just been here a few decades and have just very different health needs.
There is a book that came out that looks at the Hmong culture and its interaction with the California health care system, and it is just a very sad book. I can't remember the name of it off the top of my head. It was a very good look at some of the health problems that we have got, given some of our diverse immigrant populations. Oh, The Spirit Catches You When you Fall Down, that is the name of it. It is great. I recommend it.
The next slide, although we just had 2.7 percent multiracial, we are expecting that to really increase in the population. As Alvin was saying, I don't know how we will track that. I guess that is up to our demographic research unit of our Department of Finance, who massages population data into a form that works better for us.
We have, as the health department, the official requirement to use data that comes out of our department of finance.
They have been most helpful in keeping us apprised of what is going on at the national level with census data, and I guess have not been too helpful to the Census Bureau by submitting a Freedom of Information request, and that has sort of results that are politically difficult.
On this slide you can see that, in the decade of the 1990s, we increased from nine to 14 percent of births to parents of different races. That is just looking at the minimum race categories.
So, if you look at perhaps folks of two different Asian backgrounds, which is just as much a potential cultural conflict interest as other race groups, you can see that we have got a lot of diverse populations that are going to be coming up and getting to be a much more complex and interesting problem than they are now.
So, we have got this very complex population situation and an even more complex situation with regard to how different programs are collecting and utilizing health data.
For preparing this presentation, I did a little survey that I wish I had done years ago, about how different programs collect health data. It is pretty easy to figure out how they report it.
They typically report it in collapsed categories for small sample size and simplicity and comparability. In terms of the data that is collected, it is very much all over the map.
We typically have programs that collect a single race ethnicity question, for which Hispanic is an option. Now, that is how we typically report it anyway, but it limits the -- it is typically not broken down into origin. So, there is a single Hispanic category, and then a wide variation in how many other racial categories are collected.
The Office of AIDS simply collects Hispanic, white, black, I think perhaps Asian, and others. That is about as undetailed as it gets.
Our genetic disease program has an extremely long list of race possibilities, including Middle Eastern, which is the only program that collects that category separate from white.
They also have been collecting multiple race data for a number of years. So, that is very handy. They will have some data they can use with population denominators.
Many programs collect place of birth or place of origin, even within race categories, or separate from race categories, as a separate issue. Some programs collect principal language.
I have really not paid a whole lot of attention to different federal standards, but I kept hearing issues come up from different folks I talked to about the difficulty in dealing with different standards coming down from different branches of the federal government for different purposes.
The health data folks are pretty much on the OMB census model, but other folks are much more concerned about HIPAA, for example, and are not interested in retooling their race ethnicity data sets to OMB census-type guidelines, given that HIPAA is a much more salient fact of their life. That is kind of a problem that many folks are dealing with.
Then, of course, we have had late release of any population data we could use at all. We still don't have official data released at the state level that we can use for vital statistics data.
We are kind of the gold standard. We say, we are only going to use Department of Finance data when they have released it.
One problem with that is, as Alvin was saying, we have counties that need data not only much sooner than we have been able to provide it, but even within a county, you will have groups that just don't show up in the data collection.
For example, Sacramento has a large Russian immigrant community, which, unless a program specifically collects that data, they have no way of knowing what the size of their group is going to be, that they are going to be serving.
Detail, detail, detail is what California all about, both racially and geographically.
Also, because of the collection and reporting being kind of all over the map, we need standards and guidance as well as the detailed population data.
So, in an attempt to address the need for standards, a work group was formed with the family health outcomes project of the University of California at San Francisco, in collaboration with us at the Center for Health Statistics, and participants from counties and other agencies within the health department.
There were original guidelines that were officially approved by the health department director in 1997 or 1998, I believe it was 1998.
We have been revising those. I had hoped to distribute a copy of those, but the committee work group hasn't finally approved those, so I thought it would be politically preferable to hold off on that. I would be happy to send a copy of that to the committee, as soon as the work group has finalized them.
The intent of those guidelines was to help programs bring their race and ethnicity data collection into comparability with other programs and into compliance with state and federal requirements, to provide detailed information that they needed to help the programs update their race and ethnicity data collection and reporting.
As you know, there has been so much happen in the last few years, that a lot of programs are in need of very detailed guidance, not just these are the general standards, but how do you implement this, how do you work this into your computer reporting system and your tabulation programs, and just the real nuts and bolts detail that I spend most of my life working on, instead of this high level stuff.
The guidelines did not address problems of multiple federal standards. We just went with the census standards and insisted that, whatever categories people look at, are collapsible into minimum OMB standards, that the data collection and reporting fulfill California statutory and federal contractual requirements, that programs preferably be comparable across programs and data sets.
Then, within the comparability of the general categories, that other programs could collect more detail if they needed.
Yet, in spite of all that, we understood that programs need to have their practical constraints acknowledged.
So, the center for health statistics, which primarily is concerned with vital statistics data, was very instrumental in all of that, of course.
We are really the first and main part of the health department to reflect the FHOP standards. We implemented what I will call a semi-census model of data collection model on January 1, 2000. So, we have the same categories that the census data have, approximately.
We are collecting up to three races on birth and death certificates, in somewhat more detail than the census does.
We do all the right things in terms of separating the Hispanic question, and it precedes the race question on the certificate, and place of origin in detail, again, more detail than the census provides as part of our question.
We prefer self reporting. That is a little hard on the death certificate, but the informant is close enough, we hope.
Then, of course, our data can be aggregated to OMB categories, although we do it a little differently, in that we don't typically do -- well, we have changed our table so that we have Hispanic and not Hispanic and then, under not Hispanic, we list the other categories.
So, it is the same data, but we have organized it a little differently on our new vital statistics reports, to make it a little bit easier to compare our numbers with federal numbers.
Then, what we do is, anyone who mentions more than one race, we put in a multiple race category.
Several years ago, I was committing to try to present detail within that as often as possible, but we have had really not as much reporting of multiple races on birth and death certificates than we expected. So, that has been a bit of a problem.
The main way that we differ from the census model is that we are using text boxes on birth and death certificates, instead of the check box format that the U.S. standard certificate includes.
This is how we have been doing it for a while. If your handout looks like mine, the first slide on the next page shows an appropriately completed race and ethnicity portion of a birth certificate.
The mother or informant will report up to three races, which can be entered on the certificate in the text format, which typically is what the folks want to see. They want to really see it in the form that they want to have it put.
So, on the one hand, it is very much a write-in model. On the other hand, we do have a data collection work sheet that does use check boxes. That is how we tell birth clerks and funeral directors and other data collecting folks how to collect the data, but they don't really all the time.
Ideally, it would almost exactly parallel the census data collection but, in fact, it is a little sloppy, should I say.
We have write in blanks on the work sheet. We have a similar write in situation to the census, but rather than doing a hot deck or cold deck allocation, we have a very long pick list that the coders use to allocate write ins to a race.
We allocate Mexicans to white. So, that is probably the biggest difference that we have from the census.
We are planning to maintain this format on our revised certificates. We have completed the death certificate revision, and that will have check boxes for race and ethnicity.
We are just beginning our birth certificate revision, and we expect to have that implemented on January 1, 2005.
So, that is not set in stone yet, but we are committed to having a single page birth certificate. So, that really necessitates using the text boxes because they take up so much less space.
When the FHOP guidelines are available to you, it will have incredible detail on how the birth system actually makes this work.
It is quite elegant and we are very happy with it. It is flexible and permits us to do all kinds of interesting things as public perception of race changes, including looking at what actual terms do people want to put in there.
Should I take questions, if people have them, on the birth certificate and death certificate process? I know that is of interest to you guys.
DR. MAYS: Not yet, if you could go through the rest of the presentation, and the we will have questions.
MS. MC KENDRY: Okay, so the rest of these slides are an attempt to make a point visually. This next one, death by race ethnicity, we have 0.2 percent of our death certificates reporting two or more races.
Even allowing for the age distribution of the multiple race folks in the population, which tends to be very young, that is not what we should be getting.
One of my collaborators in the maternal and child health branch, who also works with CDC, has been looking at age adjusted rates using the new Mars(?) file for the multiple race groups.
The next slide shows that, if you compare the left-most bar with the right-most bar, those are age adjusted mortality rates for everyone who mentioned more than one race on the death certificate compared with, on the right hand, all single race people.
Then she broke it down by some of the major combinations that we have in California. Total two-plus races was 791. So, it was a fairly decent sample size for statistical purposes, but not what we would expect, and so much underreported relative to the census, that you obviously can't tell there what is going on healthwise, because it is just way too low.
Hopefully, that will change as consciousness increases on the ability to report multiple races, but I would be real surprised if it ever gets up to a level that will be comparable with the census. So, that is a real big problem.
We are constantly trying to work with the funeral directors to improve this. Hopefully the implementation of the new certificates will be yet another chance to talk to them about this.
As we move toward implementation of our electronic death registration system, which is expected to happen around January 2005, we are going to be working on further upgrading the consciousness of the importance of collecting these data appropriately on the part of funeral directors.
The next slide shows a little trend line of age adjusted mortality rates by major race ethnic groups. In 2000 and 2001, the numerator I used for this was the first listed race on the death certificate.
Jim Weed and his staff did an evaluation of our first listed race on birth certificates, and I believe they are also working on death certificates, to look at the first listed rate as a bridging technique, so they could use that to combine California data with the rest of the country.
They compared the way the single and multiple races fall out using the first listed race as preferred race, with national health interview survey multiple race and preferred race.
Statistically, it looked really good. You had about the same proportion of multiple race folks falling into each of the single race groups with vital statistics, as you did with the NHIS.
Assuming that we are going to have to use single race denominators for mortality rates, we have got a bridging technique that probably works as well as anything else.
It kind of assumes that our race work sheets are not used, which is problematic, because we want them to be used but, if you don't use the work sheet, then the order of the races is less biased. So, we have got a real conflict there, in how we want to do our data collection.
The next slide shows the preponderance of births in California that are to Hispanic moms. Sixty-five percent of those mothers were born outside the United States. So, we have got a real nice split there, that it would be very important to analyze separately. We have not yet done that.
Fortunately, with births, we can use percents of utilization of prenatal care and things like that, but it would be really good to be able to do infant mortality rates by born in and outside the country, but that is a problem.
Also, with births, we have an underreporting of multiple races, 1.2 percent rather than the 2.7 percent in the general population.
We expected that to be much higher, because multiple race populations tend to be younger, child birth age and younger. So, that seems to be a reporting problem also.
The next slide I will kind of skip over. It just shows birth rates of Hispanic mothers by origin. We have got quite a lot of variability in that in the state.
The next slide, birth rates by race ethnicity, just another illustration of how multiple races are underreported, than what we would expect.
It is not likely that multiple rate mothers have a lower birth rate than any of the other race ethnic groups, so we have got a problem there.
The next slide I had to put in because it is my favorite slide that shows the importance of looking at Asian and Pacific Islander subgroups in terms of program needs.
The huge variability in utilization of prenatal care, the low bar is high utilization. So, it is kind of reversed, but it is the percentage of mothers of that group who obtained prenatal care no earlier than their third trimester, or did not have any prenatal care at all.
You can see that, over to the right, we have some groups that are really in need of being reached out to for prenatal care issues.
So, in conclusion, we have got some tabulation issues, including how to balance the detail that programs need with sample size quality, denominator quality, and enough simplicity that it makes sense, some aspects of the public in terms of data use.
We have had availability of denominator problems. For our 2000 vital statistics report, we just didn't present any race specific rates at all. I am sure we will get some flak about that.
We have this thing about whether to use first listed race as a bridging method. We had hoped that, since we have multiple race numerators, we would be able to use that as a category, but that has been problematic because of reporting issues. That basically is the numerator denominator incongruence issue. We really don't know how that is going to change over time.
In summary, collection of race and ethnic detail, even more than the census model, is crucial for California. OMB minimum categories are fine for some basic stuff, but really need to be broken down for us.
We have major subpopulation group differences and we need to look at those differences in order to work on disparity alleviation.
We need consistent and appropriate standards from the feds. Help, help, help. If the HIPAA and public health segments could be talking to each other, that would be very good.
We need timely, detailed and high quality population data, which we have not received to date. So, I am done.
DR. MAYS: Thank you. Is it possible for you to stay with us a little bit longer? We are going to open it up for questions and answers and discussion now.
MS. MC KENDRY: Absolutely.
DR. MAYS: Okay, the presentations have been very enlightening. For those of us who have sat through several hearings, I think it has been very good to hear a lot of the detail that you are presenting.
Some of this is data that will be new, in the sense that we haven't actually had the ability to look at some of the multiple race information. So, I really appreciate the fact that some of you went to the trouble of actually analyzing data that you had. So, it will be very useful to us.
Okay, let me start with questions, and I will start here at the table first. Paul?
DR. NEWACHECK: Thanks, Vickie. And thanks for the speakers, too. These were very informative presentations.
I have a couple of questions for Dr. Onaka and Dr. Abbott about telephone surveys. First, I think it is terrific that you have, for California and Hawaii, telephone surveys of the population.
In terms of our issues that this subcommittee is dealing with in terms of race and ethnicity, I wondered about underrepresentation of racial minorities and ethnic minorities in telephone surveys, given that the lower end of the SES distribution typically is represented in a telephone survey.
What adjustments are made in the California health interview survey and the Hawaii survey for potential representation?
The second question concerns whether or not you are aware of any survey mode effects -- that is, telephone versus in-household personal surveys versus self-administered surveys -- on reporting of race and ethnicity?
DR. ABBOTT: I will respond first, and then also make a disclaimer, that I am not the technical expert.
Clearly, in doing a telephone survey, you are going to miss some significant elements of the population. Clearly, people who don't have a telephone would be missed, the homeless population in particular and, as you suggest, the people that you are most likely to miss are people who have the least in terms of economic resources and capabilities.
There are ways that the CHIS data and the results have been adjusted to reflect those, but I am not capable of really telling you how.
Again, in doing CHIS, we found that having the language of the respondent available was really critical. I think I made that point several ways on the slides.
DR. ONAKA: I think that the points that you brought up are important. We used to, prior to 1996, have a personal household interview. However, resources were the constraint of why we changed.
Again, we have looked at CPS, which is still a personal household survey, and know how many households in Hawaii do not have phones. So, that is an adjustment factor.
With regard to language, we don't have the resources to have all the dialects available at the time. What we do is a call back, whereby we can get the native speaker on the line at an appropriate or prearranged time.
Again, we have done comparisons with our personal household interview and our telephone interviews, and we don't see a real major difference.
Again, being that it is much more cost effective, for very practical reasons, that is the only opportunity that we have.
DR. NEWACHECK: Do you use the CPS data to adjust the weights in your survey or do you just assume that, since there are not large differences --
DR. ONAKA: No, that is how we know the number of individuals that do not have phones.
DR. NEWACHECK: It sounds like it is very similar distribution. Do you get similar population characteristics from both surveys?
DR. ONAKA: Yes.
DR. NEWACHECK: So, you don't make any adjustments to the telephone survey, then?
DR. ONAKA: We do, for the number of households that do not have phones. So, yes, there is an adjustment.
DR. MAYS: I want to follow upon Paul's question, just to see if we can get a little more detail. Who is it that we are less likely to hear from in the population when we use telephone surveys? I am going to address this to all of you in the states, because I think California and Hawaii have very specific telephone surveys, but I also want to know in terms of Massachusetts, and Jane, if you are still on the line.
In terms of how we are collecting data, in terms of racial and ethnic groups, who is it that we are least likely to have their voice represented, either because we are doing telephone surveys or, to some extent, in other modes when we are collecting data.
I know in some states we have difficulties with individuals who are immigrants. Maybe each of you can discuss that, so we can have a sense of, as we do data collection, where do we need to worry about not hearing from?
DR. LENGERICH: Could I add something onto that as well, please? That is also the geographic representation, particularly with the advent of mobile phones. I think individuals are less likely to necessarily be a resident of a particular area, where they have their cell phone registered. So, to what degree do we miss them by geography as well.
DR. ABBOTT: Are cell phones even part of a roster of numbers that we would be dialing from? I don't think they are.
DR. COHEN: I will start. I think there are a variety of questions that need to be addressed. One, the basic question is, what is the future of telephone surveys for generalizing to our populations.
We could spend days on that, but I think you are bringing up some of the emerging limitations. In Massachusetts, between cell phone use and call screening, our survey response, particularly for our behavioral risk factor surveillance survey, has declined markedly, and it is a huge issue.
DR. MAYS: May I just ask if you know whether or not it is the screening devices? As a researcher, I have some ideas about this, but I want to hear, do you know whether those screening devices are more likely to be used by racial ethnic groups?
DR. COHEN: I don't have any of those data from Massachusetts. I just know, in discussion with folks who use telephone survey methodology, I think we need to begin much more creative ideas about reaching hard-to-reach populations.
Simple telephone frames might not work 10 to 15 years from now. We have to think of other ways to reach people who are isolated because of their telephone use, and also, linguistically isolated telephone surveys are less likely to reach those who are linguistically isolated.
We are trying one very small experiment in Massachusetts, focusing on persons of Chinese ethnicity. We have actually developed a sample frame using our vital statistics ethnicity data to identify Chinese surnames.
Using that surname list, we are able to try to reach people by mail, as opposed to by phone, and try to reach them in essentially Mandarin in addition to English.
We have just collected the data, and we are trying to analyze the data to see what the relative response rates are and how things differ.
I think you are absolutely right. If we are focusing on using survey methodology to collect basic health surveillance and health status information, the telephone might not be the best mode as we go into the future.
I think it is geographic, it is linguistically isolated, it is homeless and it is migrants.
DR. MAYS: Thanks. Anybody else want to comment on it?
DR. ABBOTT: Just what I was saying earlier. The higher end of the economic spectrum with all the high tech screening equipment, as well as the folks who don't have the telephones.
DR. COHEN: That is right. I have a 20-year-old son and they don't have a phone. They use cell phones. So, there is no permanent phone at his address. I think we are missing younger folks, too, who live in apartment and rental situations, who might be lower income and might not have phone connections that way, as well.
Of course, there is always the problem with the elderly responding to phones. I think there needs to be serious review of telephone methodologies for our data collection.
DR. ONAKA: I would like to maybe give a devil's advocate thing here. We are very much aware of the limitations of telephone survey interviewing.
However, one, I think, plus that may be is the anonymity of a respondent. In our personal household interviews prior, that is, with visual contact, there is much in Asian culture of being a shame culture. That is, if they see you, they are not going to be providing you with some of the more detailed information on a whole scale of health-related issues.
We found that we have been able to get much more response from those of that particular ethnicity through the phone rather than through personal interviews.
Again, notwithstanding that, I think the days of telephone interviewing are numbered, and we are rapidly trying to get a better way of collecting information, but it does have some benefits.
DR. ABBOTT: I agree with both Bruce and Alvin. We tried to address this by doing an extensive outreach campaign with community-based organizations, non-English media sources.
We sent out letters where we could associate the randomly drawn telephone number with an address. We sent out letters ahead of time in the six languages.
We concur, that changes in patterns of use of telephones, that certain populations may not have telephones, that this is a significant issue.
The response rate for CHIS was 38 percent. There are many who are concerned about that in terms of interpretation and use of the data.
DR. MAYS: Jane, can you tell me, for California, given its diversity, your sense of the answer to the question of who might we not be hearing from in terms of our health data?
MS. MC KENDRY: My knowledge of health data beyond vital statistics is extremely limited. Basically, I can just speak to vital statistics.
We know we have under-reporting of Native Americans, particularly, on death data, probably underreporting of Asians for reasons that I couldn't begin to guess, and underreporting of Hispanics, because we suspect that folks who still have strong ties with their Mexican towns of origin, go home to die. Beyond that, I am afraid I can't be very helpful.
DR. MAYS: No, thank you, that is helpful. Other questions?
MS. BREEN: I just wanted to follow up with a little bit more detail on what we have been pursuing. That is, I wonder if -- Jane just addressed the question of Native Americans, but I wondered if any of the surveys, where there are fairly large proportions of Native Americans, they are collecting information on tribal detail and those sorts of things, which can be important in categorizing Native Americans appropriately.
I also wondered about some of the other measures of health disparity, which can be important when looking at racial ethnic populations, one of which can be the class we call socioeconomic status, and if that information is being collected.
Then, in terms of kind of on the list of health disparities, sexual orientation is another one, geography, of course, and disability. I would like to know if you are collecting information on that.
Then, finally, access to health services, things like whether people have health insurance, maybe what type of insurance they have, usual sources of care, those sorts of things.
DR. ABBOTT: Within the California health interview survey, I am pleased to say that we attempted to address every one of the issues that you raised.
In collaboration with the Indian Health Service, as well as some Indian health advocates, we do collect tribal membership information. We also, through a listed sample, attempted to increase the sample of urban American Indians and Alaska Natives in the sample.
As far as sexual orientation, we did have a question on that. It is one that we need to improve. The way the question came out was, are you gay or bisexual and certain people thought, well, if I am not gay, I must be bisexual. Therefore, that data is probably of limited utility. So, in the next generation of CHIS, we will be attacking that in a much different fashion.
Disability, yes, we obtained some markers of disability in CHIS 2001. CHIS 2003 will have even more questions trying to delve into that.
DR. MAYS: Can you say what your SES measure is, how you did that, because that is actually what everyone is struggling with.
DR. ABBOTT: I would really commend you to go to the CHIS web site and look at the questionnaire or to contact the CHIS staff at UCLA. I am quite sure I would not do good service in responding to this question.
I know that this is a question of great concern in California amongst the staff. Nancy has actually been a collaborator in CHIS, as have a lot of staff from the National Center for Health Statistics and CDC and the National Cancer Institute.
This is, in fact, something we are trying to address. We did get information insurance coverage and status, eligibility for public programs. We collected a lot of information on income, probably too much information on income and assets and things like that, because we are, in fact, trying to figure out if people were eligible for Title XIX or the CHIP program.
MS. BREEN: Access to care and geography.
DR. ABBOTT: Geography is a factor just in terms of the way the samples were drawn. Yes, there are questions in there in terms of usual source of care, there are questions in there in terms of problems of accessing care and things of that nature.
MS. BREEN: And CHIS will have information on every county?
DR. ABBOTT: Not every county. For confidentiality purposes, any county of less than 100,000 population was aggregated with other counties so that the sum of the population was greater than 100,000. We do have county level strata and data for 33 of the 58 counties of California.
DR. COHEN: I am very envious of my colleagues who have state level health interview surveys. I think that is atypical for most states.
Although I would dearly love to develop one for Massachusetts, I think most states rely on the behavioral risk factor surveillance system for survey data and the kinds of information that the Hawaiian and California health interview surveys contain. So, I will respond from the BRFSS perspective.
We are beginning -- we would like to consider disability as a demographic variable, rather than a separate category, and routinely collect that information.
We have added it this year and in the past several years and, hopefully, we will have enough state funds to continue using this as one of the basic demographic descriptors.
We include income and education information as SES indicators on BRFSS, as well as including insurance coverage information there.
I think you brought up a crucial issue about sampling a large enough sample so that we can produce substate and geographic specific estimates.
In Massachusetts, when we had money from our tobacco control program, we were able to add funds so that we could over-sample special areas in Massachusetts.
We focused on cities that had high ethnic and poor populations. So, we have been able to generate over-sample data for six different areas in Massachusetts.
We also work to combine data to produce not county-level estimates -- counties are irrelevant in Massachusetts -- but we produce information for what we call CHENAS, community health network areas, which are 27 mutually exclusive sub-state areas in Massachusetts for the BRFSS data, by combining five years worth of data.
I also wanted to address your question about Native American data. We found, working in Massachusetts, that we have been more effective, rather than trying to collect general data, we found it more effective to work with specific tribal groups to collect specific information on those tribes.
I think it is more difficult to aggregate and generalize about Native Americans within Massachusetts, but for the comfort level, organizationally, I think we are progressing much more rapidly by working with tribes and tribal groups.
We have engaged several different groups to try to do BRFSS-type surveys within their population. Still, I must admit there is a lot of concern and distrust working with, hi, I am with the government and we want to work together.
I think it is really crucial that we go slowly and meet the community and tribal needs first, rather than focusing on what broad state surveillance needs might be, and that is the way we have had the most success.
DR. ONAKA: With regard to your first question on American Indians, yes, when we do the survey, we ask for tribal membership. So, we go down to much more detail.
Maybe this gives me the opportunity, not only do I do the BRFSS, I do the health survey also, and I also draw the sampling frame for the SAMHSA substance abuse. In that sense, we have a fairly robust -- we haven't totally taken advantage of that.
Particularly, what we were trying to do was not have a household in Hawaii sampled in any given year by those three surveys. In fact, we are able to coordinate it where we don't provide an over-burden.
Another point about the survey, we work very closely with the telephone companies to provide us with a screen of numbers. Again, we are trying to refine the way we do it, including mobile phones, et cetera.
I mean, there are a lot of limitations. Because of 34 years of experience, I think I have addressed, I know, all the questions that you have brought up with regard to even insurance. We go down to even the plan name.
We work closely with -- say, for the Medicaid, we know actually the number of people we cover, and then we survey and we work at looking at comparability of quality of data.
Disabilities, we, again -- it is a two-edged sword. We try to be comparable with the mainland, as I call it. That is, we look at NCHS questions on the national health interview survey and, when possible, we try to use the same questions.
We use SF12, for those who are familiar with that. Yet, some of them, they are not terms that we use in the islands. We try to adjust it a little.
Again, with behavioral risk factors surveillance surveys, as you know, is only of adults. So, we supplement that with our health interview survey, which is the entire household. That was even an age issue.
I think the only one that maybe we haven't had a lot of survey research work on were with regard to gender, I guess, or sexual orientation. However, I know that we have had some questions, but it is not as extensive.
MS. BREEN: How about geography?
DR. ONAKA: I think I was maybe the one that emphasized geography a lot more than any of the presenters, showing our islands. Yes, we go down to the sub-area. We have to.
There are pockets where, say, the Marshallese, because of our free association, Hawaii is one of the places where the Marshallese, the Pilauans -- again, the telephone surveys, the surveys that we do, in the beginning, you then use the public health nurses and so on, to pinpoint the community level information.
MS. BREEN: How about the highly populated island of Oahu or the island of Hawaii, which also has a large population? Do you have some geographic information?
DR. ONAKA: We go down to the telephone prefix, because of the sample size being as large as it is there. As I said, if we combine BRFSS with the substance abuse and the Hawaiian Health Survey, I think we have upwards of 15,000, 20,000 households. Compared to our population, we do a much larger sample, per capita, than the state of California does.
MR. HITCHCOCK: Just to clarify, if I could, because we haven't talked much about this, when you are talking about the SAMHSA survey, you are talking about the states needs assessment as part of the substance abuse and treatment block grant?
DR. ONAKA: I believe so. Again, the only thing I do is the draw the sample for it. I am not involved in the actual survey. It is because of the fact that we did not want to be over-surveyed.
Just another point, for all those listed numbers, a letter does go out for BRFSS as well as the other, to the household. We send it within a two-week period before the call is made, to the household saying, you may be called.
So, it is a great effort, but for only listed numbers. Again, as you may know, with technology and so forth, so many people are not listing their numbers that it is becoming more of a problem.
We send a letter that is signed by the director of health, and it goes to that household who we have selected in a sampling frame.
DR. MAYS: Could I ask what your sample response rate is?
DR. ONAKA: In the CASRO, it was something like 50 or some. I know that the PRAMS is very high. That is really working. PRAMS is an offshoot from our -- pregnancy risk assessment monitoring system.
Again, the reason that I am involved in it is that the frame is from all births. So, I draw that for them. It is something, a phenomenal 80 or something percent, but they work the numbers really hard.
It is a history, I think, of being comfortable that we are in a partnership with the population to get at these, and a long history that the data are not abused.
DR. MAYS: What I Want to do before we bring this part to a close, is to actually turn and see if we have any questions in the audience.
MR. THOMAS: Just for clarification, can I make a 30-second statement?
DR. MAYS: I think it would be best if we could get your question and then your statement later. We are really short on time.
MR. THOMAS: For identification purposes, my name is Stephen Thomas. I direct the center for minority health at the University of Pittsburgh.
I am here on behalf of the Society for Public Health Education, which represents 4,000 public health educators around the country.
My question, in the context of eliminating ethnic and racial disparities in health, and all the limitations that you are describing, and the ever-greater need to be precise, to what end, is my question?
In other words, the fact that California has no majority group, does that mean that California has not group that has the power, that still benefits from what it meant in the past to be a majority?
It is not just about numerical proportions. The imprecision in the language, by going to minority group, could cause us to not realize that, in the context of health disparities, it is about who has power, it is about the experience of discrimination, it is about not having access.
So, as you get ever more precise in the measuring, what is the context, and to what end?
Are we also looking at the extent to which these populations experience discrimination systematically, or because of their geographic isolation?
I only ask that because of the cost involved in what you are doing. We, on the practice end, who are trying to intervene in areas of eliminating disparities, need to make sure that we have now muddied the water with multiple race categories and all the things you are grappling with, that we don't lose sight of why we are doing this.
My question, in terms of this body is, to what extent will you also either address or help provide some guidance and context for what these numbers mean, in the context of health disparities.
DR. ABBOTT: Trying to respond to that, first of all, it is one thing to collect data and make it available. It is another thing to then have that data be utilized by a variety of people and organizations to address health issues, including health disparities.
We did attempt, in CHIS 2001, to seek information about discrimination in the receipt of health services. The question turned out to have some design problems associated with it.
So, they are working with our multi-cultural technical advisory committee to develop a new set of questions that would be more specific and much more informative in terms of discrimination in accessing and the receipt of health services by the whole panoply of California's population.
The state is, as I mentioned, a very heterogeneous population now. We do recognize that we have significant health disparities amongst our population and subpopulations.
Again, the intent of CHIS is to be able to identify, to inform and to guide us in our interventions and development of health policies and programs.
DR. MAYS: I guess I will comment to what end specific on behalf of the committee. That is, one of the reasons we are having the hearing is to be able to talk about whether or not there is sufficient data available for the department to be able to meet its own targets when it talks about reduction and elimination of health disparities.
We are trying to do this in a timely fashion so that, before Healthy People 2010 gets to its mid-year review, we will be able to comment about whether or not it is on track, in its ability to be able to accomplish it.
I think, on another level, the department, the Congress has directed IOM also to look at these same issues.
I think right now there is a significant amount of bodies who are really trying to determine what is the meaning of these data, are they sufficient.
Then, I think there is a very big body of user groups -- and I think that is where Peter is going, which is your group and others -- that, once there is some sense of what is there and what is good about it and what is maybe not so good about it, has the ability to be able to, in a very public way, talk about health disparities and hopefully use it in an applied fashion.
I think that the contributions of the states, it is the states even more so than, I think, at the federal level are the ones who are really asking the same questions that you are asking.
DR. FRIEDMAN: I have several different roles here, one of which is representing a state health department, and I have a couple of responses to that question.
You know, first of all, generally, I would guess that all of the people representing state health departments would say, and occasionally quite painfully say, what they do is very much public health practice, and sometimes much more public health practice than any of us would like.
Second, the ethnic group specific data, which Hawaii collects and analyzes, which Massachusetts does, which California does, at a great level of detail, both in terms of groups as well as geographic area, are used programatically to target and evaluate programs on a regular basis, as well as to increase the understanding within the agency, within community agencies, of the substantial variations between groups.
It is clear, based upon our data and I think one of the things that certainly these three states have looked at, is trying to make clear, at the community level, that looking at what we in this country call race groups, generically adds little, if anything, to an understanding of public health.
In order to understand public health, in order to understand infant mortality, in order to understand health status generally, and so forth and so on, we need to go down to a much greater level of detail.
The data that have been discussed and presented this morning enable us to do that, both in terms of education of the public, education of community public health practitioners, as well as program targeting and evaluation.
MR. THOMAS: Just in closing, I ask the committee to consider what it will mean when, very shortly, we will be talking about whites as a minority group, in terms of these power relationships. That doesn't mean the same thing.
DR. JOE: My name is Dr. Shawn Joe. I am a research assistant professor here at the School of Social Work here at the University of Pennsylvania.
My question is really, in terms of the mental health information that is collected, is there any data collected on suicide attempts, given the great numbers of suicides that claim lives in the United States?
I am just trying to get a better sense of your mental health information and also more information about suicide attempts and suicide injury and whether there are any attempts to address that issue.
DR. ONAKA: I would like to at least share what we have done in Hawaii. We have to working with the mental health offices in our department, to get at suicide ideation.
Again, I am a data collector, so I would want to maybe defer the substantive area, but yes, we are aware that our mortality data are not sufficient to deal with the problem.
We know that the police records or the help lines were not enough. So, we were supplementing with our surveys that indicated if they even ever thought about suicide.
Again, you being in the mental health area, would know what I am talking about. There is a body of literature on suicide ideation. So, we tried to use that body of literature to get at survey questions, to address the population.
Again, I also wanted to bring out the cultural aspects. It may be stereotyping, but suicide, in some societies, is not as looked down upon. I am not articulating myself well, but I think we were more cultural neutral in the way we were asking this question.
Yes, that was a major component of one of our rounds of surveys in the Hawaii health survey.
DR. JOE: Let me clarify that point. I just want to speak much more to registries rather than surveys, in terms of suicide injuries, from the departments that receive high numbers of injury due to suicide attempts and also our poison control units in terms of the number of suicide attempts that they see.
DR. COHEN: I think sources of good information on injuries related to suicide come from emergency department registries. Many states have implemented emergency department reporting systems that collect detailed information on injuries related to suicide.
There is also hospital discharge data that has information on suicides that are serious enough to require overnight hospitalization. They are available in many states and many local areas.
MS. ROBBINS: I am Jessica Robbins from the Philadelphia Department of Public Health, and welcome to Philadelphia.
I wanted to ask, you have talked more this morning about data collection than about data presentation. One question I wanted to ask was, we collect ethnicity and race as if they were separate variables.
I think a couple of the speakers referred to the fact that people are confused by that, because they are not, really.
I was thinking about it and I can't recall ever having seen anyone make use of the data that way. I can't ever recall seeing analyses that say, of Hispanics by race, or of different races by Hispanic ethnicity.
If we don't use it, and it isn't really a real distinction, because it sort of implies that ethnicity is something cultural and race is something else, maybe it is biological, so why are we presenting it that way, as often as we do.
DR. COHEN: Thank you for asking the question. I agree with you, I think we need to focus more on presenting ethnicity regardless of race.
I provided a couple of reports to the subcommittee that focus on presenting ethnicity, not race data. In our annual surveillance of births and deaths, our reports not only have tables by race, but we have separate tables looking at detailed ethnicities.
We don't try to recode ethnic groups into race categories. I think that is a distinction that I have come to professionally in my work, of focusing on ethnicity regardless of race.
That is where I would like to see us move, rather than trying to recode ethnic groups into some broader construct of race. Just focus on presenting the data by detailed ethnicity group.
MS. MC NEIL: This is in response to the comment you just made. If you, in fact, do that, won't that present problems? You mentioned with the telephone surveys and even with encounter data, that sometimes it is not self determination but someone else who is recording what the race or ethnicity is.
Sometimes it is not accurately recorded because race is determined oftentimes based on what you see. So, I may not know someone's ethnicity, but I will look at them and assume that they are a particular race.
So, if we are going to eliminate race and just look at ethnicity, isn't that going to present problems with accuracy?
DR. MAYS: Could you just say who you are?
MS. MC NEIL: I am Dorticia McNeil with the Office for Civil Rights.
DR. COHEN: I think you are absolutely right. For me, the issue is, if I had recommendations for the federal government, certainly one of them would be to provide much more in terms of resources for training for collection of these data, at the hospital intake level, at the WIC and welfare office encounter level, for funeral directors.
I think one of the huge deficits in this system is the lack of consistent training for ascertainment of these issues, if we are not using self reporting.
When I talk to folks in the field, that is the 900-pound gorilla on everybody's back. It is, how do we get these data. If intake workers are not trained to take them well, we do not get good data.
If the subcommittee hears anything, I guess it would be my recommendation that the federal government, that has more resources than the states in general, put money into training for better ascertainment of both race and ethnicity.
DR. MAYS: Okay, I think that we have heard a lot of really good things today. I think that our presenters have done an excellent job of bringing out what the issues are for the committee.
That is why it is important for you to not be shy about telling us what you think should be recommended, so that it can be considered. So, I thank you for not being shy people this morning and for your presentations.
I think at this point what we should do is take a short break. We believe in good physical health, so we do need a break, even though we are behind times.
So, let's try and keep our break to about 10 minutes. Then we are going to come back and have another presentation before we adjourn for lunch. Again, let's give a hand to our presenters. I really want to thank them.
[Applause.]
Jane, that includes you, if you are still with us.
[Brief recess.]
DR. MAYS: Okay, everybody, we are going to get started again, so that we can have some time to adjourn for lunch. So, can I ask you to take your places, so that I can introduce our next speaker.
One of the things that has emerged as a significant issue when we talk about states -- and you have heard some comments on it -- is the issue of vital statistics.
We thought that that was a significant enough issue that we are going to spend some time actually looking at a project, at a discussion that talks a little bit about the re-engineering that NCHS plans to do.
As I think has been pointed out to you, some of the issues of how race and ethnicity is determined in vital statistics, at the level of birth and death, do raise some issues for racial and ethnic minority groups.
We are fortunate to have Delton Atkinson, who is actually the project director for this re-engineering project that is coming out of the Office of Vital Statistics at the National Center for Health Statistics.
He has a masters of public health and policy administration, and also a masters of public health in biostatistics from the University of North Carolina at Chapel Hill.
Currently, he is serving as a -- if I have it correctly -- as a consultant to NCHS on this particular project. So, Delton, maybe you can share with us a bit more about this re-engineering effort that is taking place, by NCHS. Thank you.
Agenda Item: Vital Statistics Re-engineering Project.
MR. ATKINSON: Good morning. Let me say I am sincerely appreciative of the opportunity to come, just to talk with you a little bit about the collection of race and ethnicity data within the vital statistics system.
As Vickie has mentioned, I am working as a consultant within the National Center for Health Statistics.
One of the roles I have there is to work with this reengineering partnership that has been set up between the National Center for Health Statistics, the Social Security Administration, and the National Association of Public Health Statistics and Information Systems.
That partnership is going to be focused on trying to look at ways to improve our decentralized system of vital statistics.
While this partnership expects that our vital statistics system will continue to be decentralized -- and we have talked about sort of the state federal relationship -- we expect it to continue to be a decentralized system. The goal is to have this decentralized system that conforms to more the national standards.
Now, as your September 25 meeting, Dr. Jim Weed, who is the deputy director of the division of vital statistics, provided an overview of the collection and analysis of race and ethnicity data with the nation's vital statistics system.
He talked with you about multi-race data collection vis-a-vis the OMB, policy of race and ethnicity. He described where states are on this particular issue, and talked about some of the issues around the analysis of multi-race data within vital statistics.
Then, he also talked a little bit about what is NCHS' short-term strategy to deal with the analysis of multi-race data and information. In essence, Dr. Weed prompted this message that we would like to give on vital statistics.
What I would like to do today is provide you with what I call part two. How will the collection of race and ethnicity data be implemented long term in vital statistics, and what will be necessary to get us there.
Now, this two-part message revolves around re-engineering of the vital statistics system, and new partnerships.
Let me be clear about one thing up front. The need to re-engineer the vital statistics system was not due solely to the issue of race and ethnicity, but rather, it was one of the several factors that contributed to the need for us to go ahead and move forward.
When the need first arose about changing the vital statistics system to be able to collect multi-race data several years ago, a survey of the states was conducted, to get an estimate of how much it would cost just to change the information systems and the forms and all the sundry training and so forth, to collect multi-race information.
Now, of those states able to make a change in their information systems supporting vital statistics, the total nationwide estimate came up to something around $10 million.
Now, this figure was the sum of what the respective states had gotten from consulting with their vendors and their IT staffs, and obviously, other estimates that they provided.
As we examined this figure more closely, we began to understand some of the reasons. Even though some progress has been made in the past five years, we found that a predominance of states still had old information systems supporting vital statistics.
We found several states that vendors charged a significant amount to make any change to their systems. Many states still operated a dos-based electronic birth system, for example, at their collection point, and old main frame systems in cobol for processing these.
Making changes in even what seems to be minor changes as now collecting race and ethnicity in a different way, making changes to this is precarious, time consuming and expensive.
As you know, collection of death certificate information is still basically a paper-based process in this country.
Remember, I started off this statement by saying, of those states able to make the change. There were a few states using privately developed electronic software where the vendor had actually gone out of business. Making changes to those systems is virtually impossible to do, at any price.
So, just with these examples, I hope you begin to see a picture here. The infrastructure to support a public health major data system is old and outdated. Making even minor changes to it is arduous and costly.
If there is nothing else that you get from my message today, I hope this is the point I want you to remember here.
Now, given these problems, what must be done to make the vital and health statistics system more amenable to the administrative, statistical and customer service challenges of the 21st Century, including the capture of multi-race information?
After much internal debate within the vital statistics community, we now firmly believe that re-engineering our business practices, applying new technology against those revised business practices, and working collaboratively at both the state and the national level, is the answer that we need to pursue.
It is important to understand that, currently, vital statistics is a decentralized system that is based on state laws, rules and practices.
We envision the continuance of this decentralized system both now and in the future. We don't envision that changing, this decentralized system we have here.
However, we envision a system that conforms to national standards, standards defining how the predominant practice of vital registration should be executed across states and territories using newer technologies.
Now, the second thing that this partnership will need to do to assist states, it will need to explore an ongoing independent process for evaluating vendors' base-model products against a prototype model, enabling states to have an objective assessment of the quality of any of this base software that is developed.
Such an assessment, we believe, would add to the value of the state's decision-making process. So, in essence, think of this, when you think about how car crash tests are done, where you evaluate the ability of a car to withstand those tests, having an independent source to evaluate the base model products that are being developed against the prototype models that have been proposed.
Third, the partnership will need to look at better ways to integrate the vital statistics system with other data systems, reducing person-intensive processes and improving overall quality.
The national center, for example, will be developing our national coding algorithms, such as for race and ethnicity, occupation and industry, so that states do not necessarily have to develop and maintain those algorithms themselves.
Another example, SSA and NAPSIS are developing a national network for the electronic verification and certification of birth and death information.
So, states do not necessarily have to hand verify these requests in the future. Again, I think there are other examples.
Let me say we need to, within the vital statistics system, as we are thinking about re-engineering, look at those things that need to be done, that will minimize the person-intensive processes that have to go on within the vital statistics area.
The third example, NCHS is working with NAPSIS to establish a national standard for the electronic exchange of vital certificate information, that will facilitate the interstate exchange of vital certificates.
Again, this is a manual process that goes on currently, and an effort to see if we can minimize the amount of staff time involved in those particular efforts.
NCHS, as I mentioned before, has already developed new edit specifications for electronic birth, death and fetal death systems, that we strongly recommend that states use with their re-engineered systems to improve the quality of the data.
The new standard birth, death and fetal death certificates -- and some of those have been mentioned here this morning -- and the collection of multi-race information will be central, we believe has to be central, to the re-engineered system.
The integration of vital certificate information with other public health and non-public health data systems, such as newborn and metabolic screening, Medicaid, are areas that need to be explored, with the goal of reducing reporting burden on behalf of the data providers, and improving data quality and utility of the information for other public health programs.
These are just a few of the process changes and improvements either envisioned, contemplated or are current, when we think about reengineering.
Finally, this partnership -- this is a very important point -- the partnership will need to work with states in finding monies to re-engineer their systems.
Vital statistics is a state-based system, which depends heavily on state funding and receipts. Not all states and territories will be able to find the resources to totally re-engineer their vital statistics system.
That is a point I really want to stress, not all states will be able to find the resources to totally re-engineer their vital statistics systems.
If you think about that nationally, having 50 percent of the states, for example, with re-engineered systems, and having 50 percent of the states without re-engineered systems, is not an adequate solution.
New and creative funding strategies will have to be explored, so that all states and territories can re-engineer. That has got to be the goal of whatever we do and whatever combination of mechanisms we do.
Until all states have re-engineered, we cannot have the national vital statistics system that we envision and that we need for the 21st Century. That will not happen until we get all 57 states and jurisdictions there.
Re-engineering the vital statistics infrastructure collaboratively is our national strategy. To be successful, however, re-engineering must go beyond the individual states and agencies currently sitting on this oversight committee.
Each of the states and territories will have to be active participants in a state-specific re-engineering process.
State vital statistics offices will need to reach out to other public health programs, and other users of the records and statistics, to understand how best a re-engineered system will impact their programs.
In essence, this re-engineering needs to be about looking at how the data that we collect, how the collection of that in a more timely fashion, and of a higher quality, will impact another program, and hope they impact it in a positive way.
Public health programs and others will need to understand the cost benefit of more timely and accurate information to them, and be willing to use some of their funds to help implement and/or maintain a re-engineered system.
So, in essence here, as we look at re-engineering, this is not going to be something in every state that we are going to be able to do, where the monies come from one sole source.
We are going to have to look at partners in terms of making this happen. In essence, public health in every state will need to declare vital statistics re-engineering a priority, and be willing to invest both fiscally and physically in improving the vital statistics infrastructure.
It is paramount that each of our state health officers understand this. Further -- I would say we need to go further -- other federal agencies beyond simply NCHS and Social Security Administration, will need to become active partners in re-engineering, partners who are also willing to participate in the funding of states to assist them in paying for these re-engineered systems.
There are many users of vital records and vital statistics. I have been amazed, over my career, how much I have found of other programs, other departments, other agencies who need vital records and vital statistics information for some purpose.
We need to be able to somehow engage those individuals and those organizations into this process. Connecting those business functions of those other organizations with re-engineering is critically needed, as the cost benefit ratios and the fraud reduction possibilities, I think, will be significant.
Only when the vital statistics infrastructure improvement becomes a priority at the state and national level, will re-engineering ultimately be successful.
At that point, we will be in the best position to make progress toward a timely, high quality vital statistics system, one that enables a better understanding of the new or rapidly growing populations or their health disparities, one that will enable us to better understand the various health issues associated with birth and death events, and one that will enable us to better understand the potentially rich population analysis and administrative services benefits through the linkage of vital statistics and other data systems.
Let me close with two points. First of these, without some assistance, not all states will be able to re-engineer their vital statistics system.
The result of this will be a patchwork, inefficient, non-national vital statistics system, delivering data of questionable quality for many years to come, including a patchwork ability for racial analysis. This outcome is something I think we must avoid at all costs.
Second, vital statistics infrastructure improvement must be a state and national priority now and for the foreseeable future.
The central theme of this national policy must be coordination, collaboration and a comprehensive vision.
There is a book that I have been reading recently by a guy named Dr. James Martin entitled, The Great Transition.
In this book, he describes the seven principles of engineering to align people, technology and strategy. He makes a particular point, which I think is very, very relevant when we think of the national vital statistics system, and let me read this point to you.
He says that, if each part of a system, considered separately, is made to operate as efficiently as possible, the system as a whole will not operate as effectively as possible.
The performance of a system depends on how its parts interact, than on how well they work independently of one another.
So, if we are going to have an effective national vital statistics system in this country, it is going to be important that we work together in a collaborative way and with a comprehensive vision.
I would like to just make one final comment. The whole vital statistics -- the vital records and vital statistics -- area was raised -- has a particular history.
In the early 1940s, we had the President -- in fact, it was back in 1942 -- who commissioned that a study of the vital statistics program for this country be done, and he commissioned that the budget bureau do this study and come back with recommendations about how you improve that.
He made this because there was a number of things that were happening in the 1930s and 1940s. You had the passage of the Social Security Administration Act, which now says that people are now to get social security benefits and, in order to do that, you need a birth certificate as well as death certificates for deaths.
You had passage of a number of pension laws and so forth in that particular process, which also required the use of the certificates.
You had World War II, where soldiers now needed birth certificates to prove that they were citizens.
Now, in 1940, there were 55 million Americans, about 42 percent of the population, who were native born Americans, who did not have a birth certificate.
We had a patchwork network in terms of a system out there to really rectify this particular problem. So, this led to this study, and there were a number of recommendations which kind of led to the ways in which we do some things today.
It was interesting that that budget bureau said, at the closing of this report, that it cannot be assumed that the need for adequate vital records will disappear after the war emergency is ended.
On the contrary, the course of social evolution points to continually increasing needs for official records of the existence, identity and status of individuals, and for statistics based on such records.
If you think about how we have changed in those 60 years, there has been a tremendous increase in the need for the information in those records, and for those to be timely, and there has been a tremendous change in the technology that allows us to be able to do some things.
So, here today, we are facing our 1940s dilemma. How do we make our vital records and vital statistics system better for the future.
DR. MAYS: Thank you. Let's take any questions. Dan?
DR. FRIEDMAN: Delton is an old friend of mine and I admire him greatly, and we have worked together. So, I can ask him what hopefully is not -- well, it can't be an embarrassing question since you can't claim responsibility for NCHS.
The question is, we have heard this morning from three states about the congruence and lack of congruence between essentially the model certificate collection of data on race and ethnicity and the state-specific needs.
My question is, do you think that a re-engineered certificate is going to be able to -- will be flexible enough to help support state needs around greater detail in ethnicity data, greater flexibility in ethnicity data and, thirdly, what we are learning now, even from the 2000 census and what we have learned in the past about how Hispanics view themselves, and the inadequacy of the current race question for Hispanics.
MR. ATKINSON: Dan, I think that is a very good question. There are a number of things that I think, as I listen to discussions here and discussions in other places, that affect us.
One, obviously, I think, that this committee is going to have to make some recommendations to the department as to where the federal government should be going with respect to race and ethnicity.
That is going to be very key to it. Obviously, the federal government agencies are somewhat bound by the policies of OMB. So, you have got a policy issue that has got to be wrestled with.
The second is that the technology allows you to do some things that it probably makes it difficult to do on a paper certificate.
So, in essence, I think there will be some opportunities for you to explore different ways of capturing some information, and more detailed information, than maybe what is needed at the federal level, to help address your particular needs.
I think you can do that through technology. I think you have a hard time doing it in a paper certificate, where you are confined to a certain amount of space that you are trying to get all the information onto.
MS. BREEN: Thank you very much, for your presentation. I wanted a little more detail, though, from sort of an overview.
I know vital statistics are very important, but I don't know a lot of detail about what the states provide and are required to provide and which states provide them.
I did, though -- this was years ago -- I was involved in a study group with some people from NCHS. At that time, I learned that marriage and divorce information is not provided by all states.
I want to know whether birth and death records are provided by all states. Then, I would like to know, in your assessment, do you have a sense of how many states or which states are going to need particular help in getting up and running with the vital statistics, and the new programs that you are suggesting be implemented?
I think it is mainly new technology that you would like to see implemented, and maybe how much that would cost?
MR. ATKINSON: In terms of your first question of what states are providing, all the states and territories are providing death and birth information to the NCHS.
MS. BREEN: So, there it is a question of accuracy rather than information.
MR. ATKINSON: Right. At this point, there is not the provision of marriage or divorce data. I know there has been some discussion in some other organizations about the possibility of doing that, but right now, there is nothing in the resources to do that particular piece.
Now, your second question about how many states, it is difficult to know, at this particular point, in terms of where exactly each of the states are.
We are trying to get a good handle on that right now, in terms of states that are going to be able to implement in 2003, which is going to be just a handful, those that think they will be ready in 2004, those who will be ready in 2005, 2006 and so forth.
You have to realize that, as states are in those various pots, we have now got some incongruencies because, as they are implementing the re-engineered system, and also most of or all of or whatever of the new standard certificates, we will have some states sending us information that is not congruent to what other states are sending, because they are still using their old certificates.
So, we are going to have a juggling act there for a couple of years with the vital statistics community.
Now, cost is real difficult. If you look at electronic death certificate systems, and you can look at some of the bids that are coming back from the vendors, you are talking about $1.5 million to $3 million per state, just for that system.
Now, you add births and so forth, so you are looking at well over, for both a birth and death system, you are looking at well over $100 million.
MS. BREEN: Is this to tailor the system? I thought you were talking about a general system which would collect 85 percent of the data that is necessary for any state.
MR. ATKINSON: That is the goal that we are trying to do. Under the current scenario, what is happening is that each state develops their own RFP, using whatever guidelines and issues they are concerned with. They submit that out, and then a vendor responds.
In that practice, we are looking at well over $100 million, to do both electronic births and electronic deaths. Obviously, we don't think we have that kind of money to make that happen.
MS. BREEN: One other issue that Jane brought up earlier was that the funeral directors give a lot of the information on deaths. Are there plans to train or do anything additional?
I always thought that was -- I can see where that would happen, but it is an unusual place to get that information because they don't really have any incentives to give good information on this, do they?
MR. ATKINSON: Let me have two of the state folks -- Dorothy or Trish?
MS. HARSHBERGER: One of the things that I think that needs to be said about vital records is that it is a legal system.
The data are not collected for statistical purposes, for the most part. Birth is a legal record that you are collecting for a legal purpose, death is a legal record.
So, when the information is provided, it is provided in those contexts. The birth has an additional section that is collected for statistical purposes. The collection mechanism varies greatly among the hospitals that collect that.
In some cases, they do get information from medical records, in other they ask the mother, in others they may put down whatever they think, however they collect it within the hospitals, but it varies greatly.
When we looked at some of those records, we found that that was the case. So, this is the system that is used, and you collect it in the way that it is used.
Deaths, the funeral director provides the information based on how he collects it in his particular funeral home.
Those that deal with specific groups generally deal with those groups, and they will put down what those groups are without asking. Others that deal with a variety of groups may ask the individual who is providing the information, which is generally the next of kin.
I will say that, for death certificates, that is not a field that we get requests to amend or correct. So, what appears on those certificates seems to be what the family or the next of kin want on there. It may not be what the individual would put down himself, but it is what the family or next of kin wants on there.
DR. MAYS: Let me see if there are any questions. I think we have time for only one from our audience.
Okay, I want to thank you for your presentation. It has been, again, very useful to the committee and will help to illustrate some of the things we need to think about relative to our recommendations, and thanks for bringing us up to date on what is going on, in terms of this re-engineering project.
I get to make an announcement about lunch, so that you all know where to go. It is funny to make this in Philadelphia because it is known for its food, to the point where I guess the mayor had people engage in fitness in order to try to reduce weight. So, I am a little nervous about going and eating here, because your food is too good, I think.
There is a place that is in the building called Willie and Duffy's Philly Grill. Then there are two places across the street, El Azteca, and the Las Vegas Lounge.
We are going to start at -- unfortunately, we are a little bit behind here, but we are going to start at 1:15 because we have people who are catching trains, who are coming in and need to get back out. So, unfortunately, I can't give you much more time.
You are free to bring your food back here, if you want to, if you want to be with us, but we are going to start at 1:15. Thank you, and we are going to adjourn for lunch now.
[Whereupon, at 12:25 p.m., the meeting was recessed, to reconvene at 1:15 p.m., that same day.]
DR. MAYS: Good afternoon. We appreciate your tuning back in with us. We are going to get started this afternoon, because we also have a full afternoon, like we did this morning. So, I want to make sure that we have adequate time for all of our presentations, as well as discussion.
For those of you who are in internetland, nothing happened to the technology. I realized I was silent too long. I have to remember we are on the internet. Here, people can see my face. I was actually looking for a bio to introduce our next presenter.
So, I am going to tell you a little bit about her. I know that Kate Brett is with the National Center for Health Statistics.
She has been an active researcher in the area of women's health for some time, and her presentation today is going to help us to learn a little bit more about some of the resources for states, in terms of using states' data, that can be utilized by individuals through this program that they have pulled together. Kay? If you need to say something else you can. That is what I knew.
Agenda Item: Healthy Women: State Trends in Health and Mortality.
DR. BRETT: I like that introduction because I don't have to feel embarrassed or anything. I am just me. Yes, I am here to talk to you a little bit about a project called Healthy Women Statistics. It is state level data on women's health.
Healthy Women Statistics is a user friendly set of state level tables of state level data on health and well being.
It was developed by DHHS with support from the DHHS Office on Women's Health. Before I go any further, I want to add to what it is.
It is, in fact, a set of tables on all people. It is not just women. We call it Healthy Women Statistics because it is being sponsored by the Office on Women's Health, but we have all sexes, as well as men, in our tables.
The purpose of the project was to allow the viewing of previously published data by specific population groups within states. It is primarily to give information on the state level by sex, and race and ethnicity and age, currently.
We are looking at mortality, morbidity, health risk factors, potentially other issues such as natality.
The data are released as electronic interactive data tables. This is one of my principle decisions early on in the project, which was to not put out another set of printed tables. That doesn't really move anything forward.
The way we are releasing the data, many tables can be collapsed into one. Usually if you look at a print table, you look at state by race, state by sex, state by age as three separate tables.
In the format we are releasing this data, that would all be collapsed into one data table.
Because they are interactive, users can format the tables or the data, actually, in whatever format they need. They can pull out a single set of data, they can graph the data how they choose, or they can present the data in a specific way that makes sense.
Finally, because it is electronic, we aren't going through the whole publication process. That take time. Once the data are analyzed and put into the software, it can be released.
Currently, the data are available on CD ROM, and we are planning on publishing a new CD ROM early this winter, as well as on our web site, which I will give many times, so that you don't lose it.
Currently, we have added to the focus of this project of measuring Healthy People 2010 objectives. That wasn't the original focus of the project. So, there are some tables out there that have no bearing on Healthy People objectives.
We are currently in the process of taking the tables we have, and re-analyzing them in a way that divides things up so that the objectives can be measured through them.
A very quick set of examples of topics I listed here, all kinds of mortality, 41 specific causes of death. Now, these were causes of death that were chosen by the Office on Women's Health. They are definitely not NCHS' list of causes of death, but they were what women's health wanted.
Then some other things such as mammography use, tobacco use, weight for height, diabetes. In total, there are currently five mortality tables and 20 health risk and health behavior tables.
We now have data available on mortality from 1997 through 1999. Let me also add right here, that we are not looking at trends. That is not the focus of our data. So, we started fairly recently. We are not -- the point is to look at current information, as current as we can get.
The 1996-2000 health behavior risk factor surveillance system, BRFSS. We are currently going through the clearance process to have our 2001 BRFSS data approved, so that can be released. It should be released within the next month.
2000 mortality will be available soon, and we are now in the process of negotiation to put on the youth risk behavior surveillance system and AIDS incidence.
As I said, all the data is concurrently available by state, sex, race and Hispanic origin, and age. That is what we are striving for.
There may be some data systems in the future where we can't do all of those processes at the same time because of approval from the data system itself. At that point, we will be crossing state, sex and race and then state sex and age, but we would never put out a table that doesn't cross at least three out of four of those.
In order to do all these crosses, we are putting out data as three year moving averages. That allows us to increase the reliability for estimates with infrequent events or small sample sizes. It decreases confidentiality issues.
Even still, there are going to be some estimates with low precision, and those estimates are suppressed, so those data that are out there are usable.
Our suppression criteria are based on the relative standard error with a formula that is shown there on the slide.
Mortality suppression is based on estimates with less than 20 deaths, which is roughly a relative standard error greater than 23 percent.
BRFSS and hopefully all future survey data will use this suppression criteria or a relative standard error of greater than 30 percent.
Let me talk a little bit about our race data. Clearly, because we are using three-year averages, we need three years of data with new categories, in order to present the new categories, and that is based on the data as it is collected by the different systems.
Mortality will be by state and whenever the state adopts the new standard death certificates.
Among BRFSS, multiple arrays in the separate Asian Native Alaska and other Pacific Islander categories were added in the year 2001. That means that it will take probably to 2004 before we can show the data, because that is when we will have three years of data to average. As we add other data sets, we will look at what we can get.
Until the data is in the new format, for mortality, we are just going to follow what the Division of Vital Statistics is going to do with their data, who have information in several different manners, and I think that is still under negotiation within DBS. So, we will wait and see what they do.
It is our strong belief that NCHS should do the exact same thing on all the different levels, and I am not in any way going to make any decisions on that.
BRFSS, we are currently using -- there is a question on preferred race. So, for the people who list multiple races in 2001, we will then categorize them by preferred race until we have three years of data, and we are combining the categories back to Asian Pacific Islander, again, until we have three years.
Currently, our Hispanic origin is available on all data tables. Race for people of Hispanic origin is not presented now and, in our mortality tables, Hispanic origin is not pulled out of the different race groups.
So, for future releases, we plan to cross tabulate race and Hispanic origin, with the criteria that, if the data isn't collected that way, we can't, and there will be a lot more suppression among how much information is available looking at different races within Hispanic origin, is questionable.
I am just going to answer the questions that were given to the presenters. I looked at the questions and there were very few of them that I could actually answer, because most of them were addressed to state respondents.
So, I pulled three. The number three was, will all ongoing data sets use the same race ethnicity standards. The answer is, we will use the OMB standards once the data is available for three years, and we can only do this when the data system collects the data appropriately.
Number four was, do you believe there are significant problems with misclassification into race and ethnicity groups.
Yes, you have heard about the fact that there are known race and ethnicity classification problems in mortality data.
I actually asked the BRFSS people what their response to that question was, and got no response over the past week, which was all the time they had to answer me.
In any case, it is not in our purview to change the data. We do, in our technical notes, note problems, and we actually put in information into the technical notes about the misclassification issues in the mortality data. That is the best we are able to do.
Finally, question 11 was, are racial and ethnic group data routinely reported in publications and reports. Yes, in fact, that is a primary purpose of this project, is to report data by subpopulations.
Now, let me talk very quickly about the project itself, and give you a little example of how it works.
We are using software called, Beyond 2020, which is a data dissemination software developed in order to allow people to interact the presentation of data.
It allows the retrieval of data in a number of formats, or it opens up as tables, but they can be quickly formatted as maps and graphs, and a variety of different kinds of graphs.
In order to do that, a reader software is distributed with the data, both on the CD ROMs and on the web site. It is very similar to Adobe Acrobat reader software that is freely available.
Information on how to use the software is available on the CD ROM inserts, on the web site, as well as, we have enough technical support that if you have problems -- and lots of people call us -- we will walk you through it.
Let me show you -- I have opened Beyond 2020, and what normally pops up is this finder. It finds data based currently based on data type and topic within the data type.
We are also in the process of thinking through, rather than classifying by data type, classifying by category, like reproductive health, cancer, and also we are going to have a list of Healthy People 2010 objectives, that you can just click through to the objective number that you are looking for.
For any of these titles, there is a title and then a short description. If you are interested in more information, there is what they call summaries, but it is kind of background information for the files, that talk about the question that was used to collect the data come from. There is the question and how it was categorized, any problems with the data. So, in this case, some states included the question and some states didn't. Age adjustment.
Within all the data, there is a lot of contextual information. I am going to open up the smoking table, because it is probably -- it doesn't have small sample sizes.
As you can see, it opens up as a table. All of our BRFSS tables open up by state and by age. Say you weren't interested in age. Say you were interested in looking at state by race.
Up here is the race bar and right now it says all races. You can then scroll through -- and I now have been displaying white non-Hispanic, black non-Hispanic, Hispanic and so forth, and that is one way of getting all the data.
If I am not interested in age at all, I can pull this down and now I have the percent of current smokers by state and race.
Now, say I was interested in both age and race, but I was only interested in the state of Florida, because I know it is there and the race data is not suppressed.
Again, if I select that attribute and bring it over here, I have now formatted it as the state of Florida by race and age.
It is very easy, it is very quickly, it is very user friendly. That is the whole point. The whole point is there is very little learning curve, and the data that are there should be good.
Our data that is coming out in December -- well, it should be coming out even earlier than that -- not only are we going to have the data points there, but we will have confidence intervals, so that someone can look and see, how reliable are assessments, what are the standard errors around it.
That was not something women's health wanted, and I just decided, I am the one who is using the money and I will put it out however I want to. You didn't hear that from me.
Now, say I wanted to graph this information. I have now highlighted the various data that I am going to graph button, push it, and I have a graph.
So, anything I have here, I can scroll through and look at state and just scroll through the various states.
It is not publication level graphs, but it is certainly graphs that people can use.
One last thing I am going to show is the mapping feature, because mapping is something that usually takes a lot of effort.
So, there it is, and it is very hard to read, but it is all ages, all sexes, white, non-Hispanic, proportion who are current smokers.
Again, this mapping feature works just like the graphing. I can scroll through and now look at the different races and see how the map changes.
It is very hard to see. It is way up on the top. Can you see it? The problem -- and I will point out this very big problem. The very big problem of this mapping feature, which we have been harping on Beyond 2020 to fix this, it looks like there is a zero here. This is because they are missing data. They are not zero at all. It is a problem, and it is not in my purview to change, but we are working on them. Okay, let me get back to my presentation, which has now gone way beyond changing, graphing, mapping.
I just want to quickly tell you how to find all this data. If you go to the NCHS home page, which is right here, and scroll down the page, there is something called tabulating state data. Under that, you will see the arrow that says, how to build your own data, Healthy Women state trends in mortality. Click that, and you will get to our home page.
Without any other -- this is where you can get technical support for both the software -- the woman who builds our tables is in North Carolina, and she is expert. I ask her all the questions.
If you are interested in the project or the data itself, Joanna Scolargianis(?) is the administrator of this project, or myself, the PI. That is it. Are there any questions?
DR. MAYS: We will do questions after the presentations for this section. Thank you. Our next presenter is also going to be talking about a system for the provision of data, in terms of looking at minority and women's indicator data base project.
This also comes out of the Office of Women's Health. Kate's project came out of the Office of Women's Health. This one is also coming out of the Office of women's Health. Alfred Meltzer is with Quality Resource Systems, Incorporated. I think they are at a much earlier stage of the project.
MR. MELTZER: Can I have just a minute?
DR. MAYS: Yes. We will actually use the minute, if someone has a question for Kate.
DR. LENGERICH: Can you scroll through risk factor by risk factor?
DR. BRETT: No. The problem with putting the risk factors, in mortality, all the various causes of death are in the same table.
The problem with putting the risk factors into the same table is, healthy people have different age categories they are interested in. So, colonoscopy, we had to separate age at 50, same as mammography, whereas, other things we separated by ages a little more evenly. So, we put the risk factors in separate tables.
There is a way to output this data into a table -- you could combine tables and use them that way, and you can output this data into ASCI files or Lotus files and analyze them more on your own as well.
DR. NEWACHECK: Who is the targeted audience for this?
DR. BRETT: The targeted audience was the state women's health coordinators and the regional coordinators.
DR. NEWACHECK: Have they been able to use it effectively?
DR. BRETT: I really don't know. I have been at two different conferences with them and done some trainings, but they haven't necessarily been required trainings.
So, the people who come are the people who are probably more computer literate and able to do the data. I don't know what they intend to do.
Now, it certainly would be appropriate, if women's health is interested, to have someone in their office to be the resource person for the states to call and say, can you tell me how to look at this.
I think we are right now at the point where the system works and they are trying to figure out how to promote it.
DR. MAYS: Great, I think our next presenter is ready. Thank you.
Agenda Item: National Women's and Minority Indicators Database Project.
MR. MELTZER: Thank you, and good afternoon, and thanks for inviting me. I would like to talk about a project that has recently started. So, we are well behind where most of everything you have heard about today has concentrated, and I feel a little bit humble in the fact that all of you have spent a lot of time.
The project that I am going to talk about started a number of years ago, three or four years ago, when the regional office in Denver of the Office of Women's Health in the regional office, a newly appointed person called the state coordinators together and said, okay, let's plan for the future. Bring all your data and we will sit down.
It turns out that none of the six states in that region had any useful data that they could bring to the table.
So, the regional coordinator decided that this couldn't happen, this couldn't be the way of the future. So, she put out a solicitation and a contract was awarded to bring together all the data, essentially a big wish list.
Each state coordinator put together a list of the variables that they would like to see and have available to them.
The other kicker was that they wanted it all at the county or below level. So, a solicitation was put out and, at first, what they thought they would want was just a status report on the region, what was the health status of women in region eight.
Well, we assembled the data base for that purpose and put out a report that looked across all the variables that were thought to be of interest, and I will get into the details in a few minutes.
It turns out that the underlying data base was of more interest than was the health status report, because the health status report only was a sort of synthesis, whereas, people wanted access to the data that drove it.
As a result of that project, the data base was spun off as a separate entity and made available within the region.
Again, we have a county level data base for all 400-odd counties in the region that show all the variables that were of interest, that could be gotten.
As a result of that project, there was some amount of willingness to go forward and implement this in a fuller basis.
By that, I mean, when we were talking to people in one state, people from other states would come across the border and say, we would like this for ours, how much does it cost.
As it turns out, there was a sufficient groundswell that the central Office of Women's Health decided to make this a national program, and that is the program I am going to talk about, but the program is predicated on the model that was developed and tested in region eight.
The new program has a lot of bells and whistles, and let me get them out of the way, and then we can put them aside and get to the data.
There is user friendly software that allows people with literally no computer skills to sort of gain access to a gargantuan data base, and a point and click approach.
It is going to have GIS. It is going to have internet-based mapping. It is going to have an internet tool. The data base is ultimately going to reside on a data base.
It is going to have CD ROM with self-installing capabilities, so that people that go into the field and don't have access to the internet, can sort of carry a laptop along, slap a CD in, and do the work that they need to do.
So, those are the bells and whistles. We will put that to the side. The real issue is that we are going to populate a data base that is quite extensive.
If you think about the combination of all the races, all the ethnicities, all the genders -- there aren't that many -- the genders, and all the counties and all the variables, you have a pretty powerfully large data base.
Now, clearly, a lot of the cells within that matrix are going to be empty, or they are going to be small values, and we will get into the data issues later.
That is ultimately what we are trying to do. We are going to try and create this data base, populate it, and make it accessible to people in the region.
What I want to do is talk very quickly, since time is short, about the background, which I sort of indicated already, how this project came about.
I want to talk a little bit about the implementation plans and the products that will result from it. Then I will get into the data, the data sources, and the issues associated with the data.
It is important to know what the data base is going to be used for and for whom it is intended. The target audience, the sponsorship is by the Office of Women's Health.
While you will see that the variables cross cut a lot of things, a lot of issue areas, it is primarily focused on women's health issues.
However, sort of a partner in this process is the central Office on Minority Health. So, there has been a lot of focus on minority health issues, on top of the women's health issues.
The ultimate target audience is not researchers. It is people who have operational responsibilities, like PCAs, like state health offices that have special programmatic responsibilities, like low birthweight births or teenage pregnancies.
So, it is set up for public health professionals and not necessarily clinicians or researchers.
On the other hand, we have put together a clinical panel covering the medical specialties of interest, like oncology and cardiology, internal medicine, pediatrics, psychiatry, to sort of guide us and to make sure that the text is meaningful, not only to a semi-professional audience, but a clinical and health provider audience as well, so that it has meaning.
Again, the intent is not to be a clinical data base or a research data base, but an operational data base. So, that is what it is intended to do and not intended to do.
I mentioned that the background for this project was started some while ago, and I have sort of covered these things. Let me talk very briefly about the products of this project, and let me sort of talk about this complicated thing.
We are going to be phasing in this project region by region over the next 12 to 14 months. All this really means is that there is a data base product that will be delivered to each region, there will be a set of health status reports on women and minority populations in each region, and ultimately the project will go up on the internet.
On a region specific basis, what happens is that, to engage all the participants, we are going to start off with a kick off meeting in which the Women's Health Coordinator, Minority Health Coordinator, will talk to us about the regional issues.
We then engage the state women's and minority health coordinators with telephone conference calls. What is really asked of these people are, in terms of the regional people, they are asked to help us identify regional resources.
In terms of the state people, we are going to ask them for access to people who control disease incidence data and the two other topics that we can't cover with national data, and those are ITOPS(?), abortions, and fetal deaths.
So, disease incidence data that we can't get at the national level, fetal deaths and ITOPS we can't get at the national level, that is the extent of data that we are going to ask each state to provide to us. Everything else in the data base will come from national sources.
This is the structure of the report, which will also give you a sense of what the data base will be. So, we have complete demographics, mortality, infectious chronic diseases, mental health, and these other topics.
Each one of these sort of is a component of the overall data base. Again, when we synthesize the individual region-specific findings, they will end up in a report with this sort of structure.
In addition, the reports will cover the data sources, the data frailties, the methods used, and all other techniques that go into it, plus there are some more detailed tables and data presentations.
The report doesn't convey the entire data base, which is just not practical or possible. The data base is a CD that, at the outset, will be handed to each of the regions.
At the end of this process, when each of the regions has a CD and a report, we are going to be providing training in each of the states, in each of the 50 states.
There is not time today for me to show, but I have a copy of the CD that we developed for region 8 on the system, and if anybody is interested, I am happy to stay around after the presentation or after the proceedings today, and show you how it works.
Each of the regions will be provided their own CD at the outset and, after all 10 regions are phased in, the system will be put up on the internet. In the intervening period of time, each region will have access to their data on a CD.
What we do, we go on site and provide training. This is just some of the details. Each of the training sessions will last about seven hours, and will have an extensive amount of hands on time.
Roughly five of the seven hours allow people to sort of work through examples and get comfortable with the system.
We have had, as I said, a whole cross section of people that we have trained already in Wyoming, Colorado, Montana, people from all sorts of settings, with relatively little experience, and they have all come away as users.
So, an extensive amount of time spent on training. Also, what we have found in the past is that, there is only room for 20 or 30 people at each training session. What we found in the past is that, as long as some one person in an area knows how to give these things, we are going to have lots of CDs to distribute. As long as there is someone locally, we will provide TA remotely, but what we have found in the past is that, as long as there is someone in an office who is comfortable using something, it is more likely to be used.
Getting into the details, we have an extensive amount of census data, as well as socioeconomic data included in the system for characterization of geographic areas, mortality and so on, infectious diseases.
The infectious diseases that we are carrying are shown here, and STDs are broken up into chlamydia, gonorrhea, syphilis, and in terms of hepatitis, it is A, B and C.
Mental health is a problem. There was a question earlier in the day about mental health. These are the other data.
In addition to the mortality and natality, we have violence and abuse prevention data. In terms of the major data sources, we have FBI data at county level. Prevention data clearly is from BRFSS, and that is not at the county level, that is at the state level. All the other categories are at that level.
In terms of the data, as I mentioned, this is the data that we go to the states for, and we have had a lot of good success.
So far, we are into our fourth region now, and we will be getting to Alabama shortly, and the other states. We started out in the Chicago base region. So, we are sort of moving our way across the country.
What we are doing is, we are making contacts in each of the states, and we are gaining access to people that have access to these data.
What we have been doing is, we have been sending them sort of spread sheets to show them exactly what we are looking for.
Data issues, lots and lots of cells with very, very small values. When you think of Native Hawaiians in Colorado that have TB, very, very small numbers.
As a result of that suppression, each agency that provides data, whether it is a national source or a state source, their suppression issues override the presentation of data.
Consistency, these things have been talked about, three-year averages, transition between ICD-9 and ICD-10, we are dealing with those issues.
Now only are there source differences -- I think I may have a chart here. This is within individual states, but I think Jane mentioned that on the phone. This is a particular state, and the color codes mean the way they report racial data.
So, not only is it across state problems, it is within state problems, and you are all aware of that, and we are addressing that.
Anyway, these are the project contacts. These are the people from the Office of Women's Health that are directing the project. I am sorry I ran over time. I am afraid that is all there is.
As I said, I am perfectly willing to stick around after the presentation to answer questions.
DR. MAYS: We are actually going to have a discussion now. So, let us open up the floor for questions for either.
MR. HITCHCOCK: I have a question. This is maybe for both of you, I guess. I think both of these are really great efforts, and it is a giant step forward.
I was looking at your list of data issues and limitations. I was thinking, you must be learning about data gaps, too, and what are the people telling you that they would like to see that they are not seeing, and if anyone is keeping track of this.
MR. MELTZER: We are. As we get to the end of each of the chapters in the document, which represent components of the data base, we are recording gaps, and we are going to be providing that back to OWH as part of our final activities.
In fact, we are acting almost like a data agent across jurisdictions. What we are finding -- for example, in North and South Dakota, there are Indian reservations across state boundaries.
In the past, the people in North Dakota had data and the people in South Dakota had data, and they really couldn't pass across data. Even if they could pass across data, they were inconsistent.
This data base has allowed people to sort of do that. When we get nationally, we will be able to cross not only state borders within a region, but across regional boundaries. So, the Dakotas and Minnesota, for example.
We are identifying data gaps that way, as well as just internally.
MR. HITCHCOCK: Maybe I can ask just another one. You and I were mentioning the area resource file after lunch. That is a different set of variables. As I remember, a lot of it is health resources, hospitals and that sort of thing.
MR. MELTZER: That is correct.
MR. HITCHCOCK: Are the two files going to be linked in some fashion?
MR. MELTZER: What we have done is, we have extracted from the area resource file those variables that are of interest in this data set and just sort of moved it over.
One of the things we do on the CD and in the internet version, there are hard links to the principal sources.
For example, we can't possibly contain all the information in CDC Wonder, but there is a hard link to the Wonder system. The same thing with Census.
So, as people want more detail than is available in this data set, they can hard link directly over to it.
In terms of the area resource file, we have sort of looked at the variables in that file that might be of interest to this, and we have just sort of incorporated them.
DR. MAYS: Let me just ask a question that would come for both of you, and that is, I am trying to get a sense of what the bigger organizing factor is here, in terms of the data.
Like for example, when you commented about the areas of the data that you have, it doesn't link up necessarily with Healthy People 2010.
Then I look at yours, and yours is different than hers, and there is a long list. So, if you both could comment on what it is that you think these projects -- I know you are being asked to do them, but I want to get a sense of what you are trying to achieve with the projects, or what you think they will do.
Two, when you have questions, like for example, when you put up the data challenges, who resolves these? Is it that the state resolves them for you? Is it that you go to a group of data users? Is it that you go to statistical experts?
Here, I am particularly interested in how those questions are resolved around race and ethnicity. Now, NCHS, I am going to imagine that they are resolved by the NCHS policy.
Since you are not NCHS, if you could tell me how yours are resolved? So, this question is for you both.
DR. BRETT: As I said in the presentation, while we didn't really have a focus, and Suzanne Hanes has many focuses and is all over the place, we have resolved that.
We then came around and decided that the way to organize ours is Healthy People 2010. So, there may be topics -- and there are very few -- there may be five or six topics that are not Healthy People objectives currently, and they may or may not be phased out, but that is our focus.
That also helps us resolve issues in terms of availability, because Healthy People 2010 data people have already decided, here is the question, here is how it is going to be addressed at the state level.
We are just doing it that way. We are not trying to reinvent any boats.
MR. MELTZER: My answer in regard to that is, our existing data base is set up to be consistent with Healthy People 2010.
Each variable that we pop up has the Healthy People 2010 objective identified. So, when you see the county's data, or if you pick out more than one county, if you pick out a state's data or a cluster of counties, you can see the value of the indicator for that county and, on the same screen, is the Healthy People 2010 value, which, of course, is a national value.
So, you can see how that county or cluster of counties, how the values relate on inspection.
The question about how to resolve data anomalies or data issues --
DR. MAYS: No, not anomalies. I don't want to call them that. They are not anomalies. I am asking when the issues are put on the table, race ethnicity are not anomaly issues.
MR. MELTZER: Right. We have just been asked that very question, are we going to be in compliance with the OMB directive.
Well, we don't create the data. The data comes from the states or the data comes from census or the data comes from birth records or death records. So, we don't actually create the data.
To the extent that the data are consistent with the OMB directives, our data base will be.
DR. MAYS: Let me just go ahead and ask more specifically, if Massachusetts gave you their data, with their full array of their view on ethnicity, would you display their data in a way that Massachusetts would like to have its data displayed or is it then going to be reformulated into the OMB display?
MR. MELTZER: The latter. In other words, we have sort of a data base, and it is sort of a fixed format data base. So, to the extent that we expand the data base, it has got to be for everybody.
To the extent that Hawaii or California may have 15 breakdowns for Asian, if we decided that that was going to be the standard, then everybody would have it and everybody would have zero cells.
We have been asked to be minimally compliant with OMB and, to the extent that we can, that will be our model.
DR. MAYS: By OWH, is that who has asked you to be compliant? I am just trying to understand.
MR. MELTZER: Yes.
DR. MAYS: If it is state data, what we have heard from the states today, and you have not finished yet --
MR. MELTZER: It is not really started yet, to be quite candid. We are starting to collect data, and we have structured a data base, but the data base has not been sort of hard finalized yet.
It is our intent to be as compliant as is practical. We could leave a lot of empty cells for breakdowns within the Filipino population but, to the extent that we can't collect it any place but one state, the data base becomes outsized, and it is just not practical to deal with it on that basis.
DR. MAYS: Let me see if there is anybody on the committee, and then I am going to turn to others.
MS. BREEN: Thank you. I was wondering about the two data sets. One is called the women's data base, and the other one is called the women and minorities data base.
I guess I am wondering why there is not one project that is building on the first, instead of a second whole new project.
Then I am wondering, are we going to have a third, which will be race and minority and SES. I am being a little bit provocative here.
MR. MELTZER: You are talking, at least from my -- you are talking to the wrong person. I am a contractor.
MS. BREEN: I am just wondering. Maybe you are not the person to answer the question, then, but I think someone should.
DR. BRETT: I think that is a good question. In fact, last year I think there was some effort at the Office on Women's Health to decide on one or the other, and their constituency was pretty evenly split on liking one or the other.
So, they decided to maintain it for another year. I think that is the quick answer to that question, because I thing I was going to be the project that was going to be eliminated.
On the other hand, there has been talk internally about taking -- the project that was presented second uses software that is slightly more -- I mean, it takes seven hours to teach rather than an hour and a half.
There has been talk about using the data and then melding it, so that both data bases are accessible in both formats. That has been some internal talk, so that there is actually some linkage between the two of them.
MS. BREEN: So, the software is different and the data is a little bit different.
MR. MELTZER: The data is extensively different. We are talking about county level data versus state level data, humongously different.
DR. MAYS: You have county?
MR. MELTZER: Yes, and you can go to MMWR and get state level disease incidence data, but try to find disease level incidence data nationally at the county level.
DR. LENGERICH: Have you run into any problems where the states or counties have wanted to define their indicators differently, particularly in light of, maybe they have set up their own Healthy People 2010 objectives and defined them a bit differently.
MR. MELTZER: Yes.
DR. LENGERICH: How does that get resolved, then?
MR. MELTZER: We incorporate their state and regional variation from the basic model. We have a template that has literally thousands of variables for each county.
What we are doing, at the very outset of our conversations with each region is, we ask the region, what is the issue.
For example, one region is interested in incarcerated women. So, we have actually gone to the Bureau of Justice Statistics and gotten data on women who are incarcerated.
Another region is interested in Caribbean immigrants. So, we are going to put that in in region four. So, on a region by region basis, we are allowing them to vary the basic model.
For things that are able to be gotten nationally, like incarcerated women, we are going to put that in for everybody.
People may not want it in all regions, but -- then, when we deal with the individual states, we are allowing them some variation also.
DR. LENGERICH: That example you gave, I would see that as a population group that they are particularly interested, but I could also see variation in states, the way they might define a particular outcome, a particular health outcome or health status or health risk factor, for example.
They may just interpret it differently and set objectives differently. So, have you seen that sort of thing, not specific to populations, but to the outcome or indicator that they are talking about?
MR. MELTZER: We haven't, yet, because we are providing them pretty much the raw data and indicators.
So, they have a shot at the list of variables at the very beginning. If they want more data, they are asked to define it up front, and we are getting small numbers of variables that are getting added to the basic mix.
Then, as long as they have the components of the indicator in the data base, then they can ultimately choose a new indicator.
DR. KRIEGER: My question was, it was interesting to see the list of variables that were included. I was just as a consultation for what will be Canada's first health surveillance report.
It was interesting to note that there were similar omissions in both, that got discussed extensively at the Canadian meeting.
For example, nothing on occupational health, nothing on environmental health. In framing how women's health is defined, there are very conventional approaches and there are approaches that are more premised on a population health approach.
I notice that you said you had a panel of clinicians that were helping with the clinical meaning. Was there consideration given to, one, having a panel of social scientists to give meaning to the social nature of the variables that you are using?
Secondly, for both, these data don't exist simply as numbers. They have a context. I am wondering what the obligations are with regard to interpretation of some of the kinds of variables that you put forward.
There are many different ways you can interpret these data. Part of it is making sure the data themselves are accurate, but another part is what they actually mean.
I am just wondering what is being done there, because the list of variables connoted one approach to defining women's health, and there are others out there that are bring in more notions of gender and context.
I am just wondering about that rationale and, again, particularly striking leaving out the work experience, leaving out environmental exposures, other things like that.
There are also indicators along the lines you mentioned that are being brought in, through UN reports and others, for example, the proportion of the population in the legislature that is female compared to the proportion of the population, for example, measures that start to get at different aspects of women's social status. The same has been done around racial ethnic groups as well.
DR. BRETT: I will answer that first. We are using Healthy People 2010. That is our framework. So, if it is not in there, we aren't going to try to get at it. If there is not a data set available giving data at the state level for indicators, we are not going to be able to measure it.
That is part of the problem that Healthy People 2010 is currently having to work with. We are just following in their footsteps.
Once they identify a data set, they have analyzed how to analyze the question in a way that is meaningful, then we will follow suit.
DR. KRIEGER: There are some environmental goals in Healthy People 2010.
DR. BRETT: There are some, and I am not sure, though, that there is state data. I know that there is EPA data, I don't remember -- I haven't looked to see the state data.
MR. MELTZER: My answer to that is, our data base was driven by the state women's health coordinators and minority health coordinators in one region, and then it was passed up to the central office of OWH.
It was the people who generated the requirement who specified the data base. So, it wasn't something that came from nowhere. It came from the Office of Women's Health.
MS. VENTURA: I just have a quick clarification, which is that the fetal deaths are actually in the National Center for Health Statistics data base. That is part of the data that we collect from states. That is available nationally. The ITOP data are not, but fetal deaths are.
MR. MELTZER: Thank you, and that is available at the county level?
MS. VENTURA: They may not be. There is a lot of data suppression because of the small numbers.
DR. COHEN: The question I had was about county level. In the New England region, county is not a viable political entity. What are you planning to do?
MR. MELTZER: We have wrestled with this on the area resource file. We used to deal with -- I can't remember, there is a term that slips my mind for the units in Massachusetts. That term slips my mind. Maybe you mentioned it earlier. We have dealt with that, and we plan to deal with it.
DR. COHEN: The metropolitan areas? CHENAS, Community Health Network Areas, but I don't think would be used anywhere else.
MR. MELTZER: No.
DR. COHEN: So, there is another geopolitical entity that you are going to use?
MR. MELTZER: Probably so. We just completed a study in Connecticut. We ended up using townships. That is their unit. We can't do that for the national data base.
We are going to work with the regional office and perhaps, if you don't mind, we could contact you when we get ready to deal with the New England region.
DR. COHEN: We prefer neighborhoods.
MR. MELTZER: That will help the data base size.
MS. HUNTER: Mildred Hunter, regional coordinator, the Office of Minority Health, Office of Public Health and Science in Region five, which includes Illinois, Indiana, Michigan, Ohio and Wisconsin.
My question is, for those states, local and county health departments, where, in their reports, that they don't reflect all racial and ethnic minority populations, is your report going to make the adjustment and how are they going to address that, because not all states collect, or county or local level health departments collect data for all the racial ethnic populations that are identified by the Office of Management and Budget.
MR. MELTZER: The way we have dealt with it in the past, I am trying to remember. What we have had to do is, in order to get a consistent cross region statement or make comparisons, getting back to the question that was asked before that we didn't answer, do you just present data or do you present some kind of findings or interpretation, we have had to collapse it.
In certain cases for certain variables, it has been white, non-white. To the extent that there is white, African American and other -- we have tried to use the least common denominator to provide the data, so that we can provide consistent data.
That is in the report. The data base will contain all the data that is available. So, if one state reports everything according to OMB, that will be, and if a contiguous state just reports white and non-white, that is all that would be in that state's data base.
When we present a synthesis, and we try and interpret the data, in order to have some sort of commonality of understanding, we have had to collapse them.
DR. BRETT: Because our data is all data that is so far collected at the national level, we will not be getting state data directly from the states. We won't have that same issue.
MS. HUNTER: Are you also relying on -- I might have missed some of this -- on county and local data as well, local data sources?
MR. MELTZER: We have been able to find all the data we need at state sources, like cancer registries or public health offices, in the state government.
MS. HUNTER: Because in certain areas it may be collected at the local level and at the county level, but may not be included in the data report at the state level. That is how you can get the inclusion of some of the populations, if you go to other levels of data collection sources.
MR. MELTZER: We haven't been pointed. We have gotten to the person in each state government that has knowledge on the particular -- hepatitis, for example. We have never been directed below that level, saying, we don't have that data but you can go to each of the county units or whatever substate units.
We haven't had that sort of a reference. We try and find the data wherever we can. So, we have asked for alternative sources.
MS. MC NEIL: Do you work directly in each state with the state center for health statistics staff?
MR. MELTZER: On a state by state basis, we get directed by the state women's and minority health coordinators to the proper place.
The other thing that we are doing, we are doing an exhaustive search of the internet sites, and we are finding individuals or units within the state, even if we are not directed to them by the referrals.
We will sometimes find a unit -- again, each state calls things slightly differently --
MS. MC NEIL: Each state, as I understand it, has a designated state center for health statistics. They usually, but not always, are responsible for the vital statistics, the behavioral risk factor surveillance system, the cancer registry and other major data systems.
They also might be useful to you if you would train staff then. Then, the state people would have a statistical support within the state, that would be appropriate for ongoing statistical support.
MR. MELTZER: That is good except that, for example, I just found out that in the State of Kentucky, the state cancer registry is at the University of Kentucky.
It may be in the same office. We may be going different places in the same office, but our first attempt is, rather than trying to search a state directory, is to try to go to people in the state who should be knowledgeable and get pointed to the people.
MS. MC NEIL: I am suggesting that the state center for health statistics, if they are not responsible for each data system, should be able to help you get there.
MS. MELTZER: Very well.
DR. BRETT: Is it called the state center for health statistics or is it like Medicaid, it might have its own name in every state.
I would like to add that, for our project, we are doing trainings at the regional level that are inviting a state health statistics center personnel and then the women's health and minority health people and having them sit down. Not only do they all get trained, they all get to know each other, and I think that is really important.
DR. MAYS: Great. Any other questions? Okay, we are going to switch, because one of the individuals who needs to be in the next segment, hasn't arrived. So, rather than doing part of the next two states, we will wait, and I am going to ask Dr. Krieger if we can switch to the next presentation with Dr. Krieger.
I think it really fits. I think you are going to get an example of some of the things and some of the reasons that some of the questions are being asked at this point in time.
Let me tell you a little bit about Dr. Krieger. Dr. Krieger is an associate professor at the Harvard School of Public Health in the Department of Health Behavior and Health Education.
One of the reasons that we had asked her to join us is because of her experience in working with state departments of health.
She has had some interaction with California, with Rhode Island, with Massachusetts, and also has had some projects where she has been involved in geocoding data, that I think will be illustrative of the issues that we have raised today.
We have asked her to do double duty. So, in addition to providing us with some of the statistical information around these issues, I have also asked her to talk about like how these collaborations come about with the departments of health that she has worked with, so we have just some sense of, if this is a model to be considered, what are some of the difficulties and what are some of the resources needed to do it.
That was part of also the CHIS presentations, for people to see that there are partnerships and collaborations that end up resulting in data sets that are larger than what the Department of Health, by itself, will do.
We have given her a little bit of extra time so that she can do this double duty kind of thing. So, welcome, Nancy. Thank you for being here, and putting time into doing this two-fold presentation.
Agenda Item: Geocoding State Data and Establishing Collaborations.
DR. KRIEGER: I am delighted to be here, and to present on behalf of my co-authors, our public health disparities geocoding project, which is about improving monitoring of social inequalities in health in the United States.
I would like to begin by acknowledging our partners, listed on this slide, from both the Massachusetts Department of Public Health and the Rhode Island Department of Health.
I would also like to acknowledge, in particular, Dan Friedman, who has been extremely supportive of our project. I have also brought along one of our co-investigators, Dr. Subremanian, who is an expert in multi-level methods and modeling. I would also like to thank Bruce Cohen, who has been involved in our project, and supportive in the early days.
One premise of our project is that collaborations between universities and health departments is vital to improving the monitoring of social inequalities in health.
Health departments are aware that this research is needed, yet typically lack the time and resources to do the work.
University-based researchers, in turn, are well placed to get the grants and are well placed to get the grants and have the expertise to conduct the needed research, much of which is methodological in nature.
Of course, we also need to collaborate with the health departments for access to the data.
The origins of this particular project drew on my experience as PI in collaborating previously with the SEER cancer registries to geocode and analyze their data, in order to use area-based socioeconomic measures to look at the impact of socioeconomic position on cancer incidence and survival.
Shortly after I arrived in Massachusetts, I contacted the Massachusetts Department of Public Health to conduct the first state-level analysis of the incidence of AIDS in relation to socioeconomic position, and also race, ethnicity and gender.
Building on these experiences, the team that I assembled worked with relevant staff at MDPH, and also to bring in another state for comparison purposes, the Rhode Island Department of Health.
To make arrangements to get access to their data, while paying careful attention to all the confidentiality stipulations, we then wrote and obtained, on the first round, an NIH RO1 grant to conduct our study, which is currently in its last year.
Among our project's products are, first, scientific publications, secondly, geocoded public health surveillance data that we have given back to the state health departments, and which otherwise would never have been geocoded, training also to the health department staff in our methodology.
Once we finish all the work, we are planning to prepare, this coming spring, a final document that we will send to all U.S. health departments, summarizing our key findings and methods.
We undertook this project because of an important problem. Apart from data on education, birth and death certificates and also some of the occupational data in the death certificates, most U.S. public health surveillance systems contain no socioeconomic data.
The net result is that 85 percent of the tables and health status determinants in the annual federal report, Health of the United States, 70 percent of Healthy People 2010 objectives fail to include any socioeconomic measures.
Instead, the data are typically racialized. We have no ability to assess either socioeconomic gradients and health within diverse racial ethnic groups, let alone their contribution to racial and ethnic disparities.
As is obvious, we need to have routine monitoring of health at the local and national level for all populations stratified by socioeconomic position, not reliant on survey data alone, in terms of getting the whole population itself covered.
One possible solution to make an intractable problem more tractable, as it were, involved tracts. By this, I mean that we can geocode our health data and link both these data and our population data, too, as stratified by census-derived area-based socioeconomic measures, of ABSMs.
One problem, however, is that there is absolutely no consensus on which area-based socioeconomic measures, at which level of geography -- for example, census tract, block group, or zip code -- should be used.
The literature, instead, is incredibly eclectic, with myriad studies using different measures at different geographic levels, thereby precluding meaningful comparison across studies or over time.
The purpose of our empirical investigation was, thus, to determine which area-based socioeconomic measures -- again, ABSMs -- at which level of geography would be most apt for public health monitoring.
Our goal was to come up with valid, robust, easy to construct and easy to interpret ABSMs that could be readily used by any U.S. state health department for any health outcomes, from birth to death, for women, for men, young and old alike, among any racial ethnic group.
Guided by ethnosocial theory, we anticipated that different ABSMs might function differently for diverse outcomes, given likely different pathways contingent on the cumulative interplay of exposure, susceptibility and resistance over the life course.
Thus, our specific outcomes included mortality, all cause and cause specific, cancer incidence, all sites and site specific, low birth weight, childhood lead poisoning, sexually transmitted infections, tuberculosis, and non-fatal weapons-related injuries.
Our a priori criteria for evaluating the ABSMs listed on this slide pertain, first, to their external validity, second, their robustness, third, their completeness, and fourth, very important as well, their user friendliness.
This slide lists the key steps we undertook from establishing our study base to creating our ABSMs, geocoding the health data, linking these records to the ABSMs, and generating rates stratified by each ABSM at each level of geography.
We conducted these analyses, first, for the total population in each state, and are right now in the process of additionally stratifying the results by race, ethnicity and gender.
We have also been addressing various methodologic issues, which I will discuss, pertaining to multilevel data and analysis.
This slide presents our study population in terms of people. In 1990, the population of Massachusetts was approximately six million persons, and that of Rhode Island about one million.
The number of records we obtained from each surveillance system varied by outcome, with the total equalling approximately one million.
In both states, approximately 90 percent of the population was white, four to five percent were Hispanic or black.
In terms of areas, as expected, block groups and tracts averaged approximately 1,000 and 4,500 people respectively, and in these two states, the zip codes average about 13,000 to 14,000 people.
The population size was most variable at the zip code level and least at the block group level.
Listed on this slide are the census-derived ABSMs we generated. Eleven were single variable measures. Eight were composite, and they were intended to capture diverse domains of economic position.
Our stance was that it was an empirical question, that we wanted to see how these different kinds of measures worked. There were some hypotheses going into it about economic deprivation, but we really were conducting a real, empirical investigation.
So, our study, our domains included variables pertaining to occupational class, income and income inequality, poverty, wealth, education, crowding, and also combinations of these variables.
For the latter, we generated either pre-established indices such as the United Kingdom Townsend and Carstairs indices of economic deprivation, also the CDC index of local economic resources.
We also generated several study-specific ABSMs, by factor analysis, and also by creating a priori categorical combinations of such variables as poverty, wealth, and occupational class.
Because we knew we wanted to display results stratified by the different ABSMs, we created two types of cut points, those based on a priori categories, and also those based on percentiles and generally quintiles.
As shown on this slide, which gives variables about percent working class, median household income and percent below poverty, only the a priori categorical cut points highlighted in yellow are comparable across the different geographic levels within and across states.
Cut points for the quintiles, not surprisingly, highlighted in green, varied across these different areas, and that has implications for comparisons.
It is one thing to see output. However, it is another to have a sense of its relevance to the real world. So, to see whether our measures made any sense, we randomly selected several addresses in Boston and took a look at them, writing down our impressions of the neighborhood, and then comparing these impressions to how the area would be characterized by our ABSMs.
This first slide, of an economically depressed area in Boston's Chinatown, turned out to be characterized as highly working class, poor, low income area with high unemployment and very few expensive homes.
By contrast, this one house -- and it is a house, not an apartment building -- in Beacon Hill looked like it was, and turned out to be in a fairly affluent area, over 75 percent professional, low poverty, high income, low unemployment and lots of expensive homes.
Just to give you a sense of the full terrain of our project, this slide shows the poverty rate by census tract in both states.
The highest rates, in dark green, are clustered in key cities. The poverty is also high in several outlying areas.
Next, before geocoding our health data, we did a small study the accuracy of several candidate geocoding firms. As we reported last year in the American Journal of Public Health, we found considerable variation in accuracy and cost, and chose the firm that achieved 96 percent accuracy as compared to, say, 44 percent accuracy.
Overall, we were able to geocode about 92 percent of our nearly one million records to the block group level, 98 percent to the tract level, 98.2 percent to the zip code level.
Note, however, that about 6.1 percent of the approximately 840,000 records geocoded to the zip code level, could nevertheless not be linked to 1990 census data, because their zip codes were either for non-residential sites, or were created or changed after the 1990 census.
As we will show in a moment, as we likewise reported recently in the American Journal of Public Health, this produced some serious discrepancies between zip code versus tract and block group level results.
Also of note, relevant to these hearings, were the problems we encountered coding race ethnicity, in relation to inconsistencies within health data bases maintained within a given health department, and also in relation to the 1990 census categories, which delimited our denominators.
Although most of the data bases we worked with contained separate fields for race and ethnicity -- referring to Hispanic origin or not -- several, indicated by the Xs in pink, included Hispanic in the race field.
This problem was particularly important for the Massachusetts sexually transmitted infectious data, the Rhode Island lead data, and the Massachusetts weapons-related injury surveillance system data.
So, having done all of this, what did we find? I will start by showing you one example, using all cause mortality for Massachusetts, presenting a table that I know has way too many numbers, and I will be showing other ways of presenting the data shortly. This table does get us started.
For each of the 11 ABSMs we focused on, this table presents, in the first three columns, the age-standardized rates for the areas with the least resources, for people living in those areas, for each level of geography, first the block group, then the tract, followed by the zip code, followed by a second set of three columns with rates for the areas with the most resources -- again, block group, tract and zip code.
The next set of columns shows the incidence rate ratio, comparing people living in the worst to best off areas.
Of note, the findings for the different ABSMs for all cause mortality, within and across levels of geography, were actually quite similar, with results well summarized by the highlighted data in the very last row at the bottom.
Overall, persons in the worst areas had mortality rates 1.3 to 1.4 times higher than the persons in the best off areas.
For a different picture, however, consider this slide for colon cancer incidence. If you look at the highlighted data now in the last row at the very bottom, for the median value for the incident rate ratios, you will see that, whereas both the block group and tract measures indicated that people in the worst off levels were at somewhat less risk for colon cancer than persons in the more affluent areas, the zip code results suggested a socioeconomic gradient in the opposite direction, and opposite to what has been reported, in fact, in the literature.
One problem, however, was comparing incidence rate ratios, as the classifications producing smaller groups at the margins might conceivably lead to larger effects because a finer discrimination of extremes is achieved.
To address this problem, we used an alternative parameter estimate, the relative index of inequality, or the RII, which was first used with UK social class data by L.C. Pannick(?).
Its values at the REI provides a slope estimate of risk across the full range of the distribution of the determinant, taking into account the population size of each strata, thereby permitting meaningful comparison of gradients across different socioeconomic measures.
To consolidate our key findings, we also devised what we called a scaled RII plot, in which we displayed the RII for a given ABSM, divided by the median value for all the ABSMs being compared.
This lets us determine which ABSMs are likely to detect RIIs similar to, higher than, or lower than those of the median RII.
Here, we show a scaled RII plot for ABSMs at the census tract level, looking across all outcomes for both states.
Thus, each dot represents an outcome. I realize it is a little hard to read for the descriptors, but I will help you through the slides. Each line of connected dots, as you can see, is for different variables.
So, for example, poverty is the yellow triangles, the Townsend's index is the blue line, and we will work through what it means.
So, the first thing to note across all the different outcomes, which start in the far left with all cause mortality, then gives different forms of mortality, then goes through the cancer incidence data, then goes to the low birthweight data, the lead data, the syphilis data, the tuberculosis data, and the intentional injury data.
For each of the outcomes, most of the ABSMs, in fact, hovered close to the median, suggesting the impact of socioeconomic position on a given health outcome is, in fact, quite robust.
That said, measures of economic deprivation, such as the percent below poverty, the yellow triangle line that I mentioned, the Townsend index, which is the blue line, routinely picked up socioeconomic gradients either at or above the median.
Moreover, for several outcomes -- most notably HIV mortality, homicide, tuberculosis and the sexually transmitted infections -- these measures picked up gradients far larger than those detected by the other ABSMs, and also, importantly, consonant with what has been reported in the literature.
By contrast, measures of wealth and income inequality generally detected associations falling below the median, or those detected with measures of education hovered around the median.
What I will now show you are examples of what U.S. public health data could look like if they were routinely stratified by an apt ABSM.
In these slides, we use the tract level variable for percent below poverty, chosen because 98 percent of the records were geocoded to the tract level and, as noted already, the poverty measure worked well in detecting socioeconomic gradients, and is readily interpretable and, in fact, worked as well as any of the other, more complex, composite measures.
In these slides, each bar represents the population living in a specified socioeconomic stratum, ranging from people in areas with less than five percent below poverty on up to areas where 20 percent or more lived below poverty, which is a federally defined poverty area.
The height of each bar depicts the rate for the health outcome. The width of the bar is proportional to the amount of population living in the specific socioeconomic stratum, and the upper X axis gives the incident rate ratio, using as reference group, the rate among persons in the least impoverished areas.
Note that none of the outcomes that I am about to show you, except the death and birth outcomes, could be included in the 1998 socioeconomic status of health chart book, because these health data bases lacked socioeconomic data.
So, starting with all cause mortality, this figure clearly shows a poverty gradient, with persons living in the census defined poverty levels, with more than 20 percent below poverty, experiencing the highest death rates.
This next slide shows the expected reverse socioeconomic gradient for breast cancer incidence, with rates highest among women in the least impoverished areas.
Next are the data on percent low birth weight. Of note, the two-fold excess risk among women in the most compared to least impoverished areas is equivalent to the two-fold excess we observed comparing women with less than a high school education to women who had completed four or more years of college education, using educational data from birth certificates.
Here are data on the percent of children with elevated blood lead levels, showing over a nine-fold excess among those living in the most, compared to least, impoverished areas.
Ditto the results for tuberculosis. Here, the excess risk is eight-fold. For syphilis, the excess risk jumps to 18-fold. For non-fatal weapons-related injury, it is 11-fold.
In other words, for none of these outcomes do we have trivial socioeconomic gradients. Yes, in current U.S. public health reports, all these gradients are ignored routinely ignored and unreported.
Particularly germane to this hearing, these next two slides hint at what these types of analyses could reveal about socioeconomic gradients within racial ethnic gender groups, and also the contribution of socioeconomic inequalities to racial ethnic disparities in health, understanding that both economic inequality and discrimination are important determinants.
Using the example of premature mortality before the age of 65, this first slide shows two key findings for three census-defined groups, white -- which includes Hispanics and non-Hispanics and, in Massachusetts, that is 98 percent of the whites identified as non-Hispanics -- blacks, both non-Hispanic and Hispanic -- and in Massachusetts about 91 percent of the blacks identified as non-Hispanics -- and Hispanics for all races, combined -- and in Massachusetts about 45 percent identified as white, 47 percent identified as other.
The first, shown by the data in red, a finding that is rather of note, is that whereas nearly half the white women and men lived in census tracts with less than five percent of the population below poverty, half or more of the African American and Hispanic population lived in tracts with 20 percent or more below poverty.
Second, as shown by the data in blue, within each racial ethnic group, there were marked socioeconomic gradients and premature mortality, ranging from the low of a two-fold excess among the white women, comparing those in census tracts with more than 20 percent versus less than five percent below poverty -- up to a four-fold excess for black and Hispanic men.
In other words, the socioeconomic gradient was steeper among the black and Hispanic population compared to the white population.
Next, this slide shows that, first, for men and women, rates of premature mortality were higher among blacks compared to whites, the data in bright red, chiefly in the more impoverished census tracts.
Secondly, adjusting for census tract poverty, your data in dark red, reduced the overall age adjusted excess risk of premature mortality -- which is the data in pink -- from 1.8 to 1.3 for men, and from 1.7 to 1.3 for women, a non-trivial reduction.
So, to start summing up, our data suggests that, first, it is feasible and informative to monitor U.S. socioeconomic inequalities in health using ABSMs and, second, the choice of ABSM and the level of geography both matter.
Before you accept our results, it is important to consider several possible sources of error and bias. First, if anything, our results underestimate rather than overestimate likely socioeconomic gradients in health.
Because the poorer persons were less likely to be geocoded or to be included in these health surveillance systems, we would be missing the worst off part of the population.
Second, suggesting that there was little bias in the ascertainment of the determinant, the ABSM data were typically missing for under one percent of the geocoded areas.
Third, from the temporal standpoint, the simultaneity of measurement of the ABSMs and the outcomes, is appropriate, because the point of monitoring is to determine where the burden of disease falls.
Studies with a more etiologic focus would have forced me to take into account etiologic period.
From a spatial standpoint, the results we have presented have not taken into account spatial correlation. We are actually doing analyses right now to examine its impact on our findings.
Of particular importance, however, I want to stress the issue of ecologic fallacies not germane to the present study design, since individuals constituted the unit of observation for both the dependent variables, the health outcomes and the independent variables, living in an area with a certain demographic characteristic.
Instead, the validity of using ABSMs depends on the extent to which areas constitute meaningful geographic units, which is a very different question from whether they are proxies for individual level socioeconomic data.
Lastly, a lack of comparable studies means that we can't say how our results are similar to, or different from, those in the literature.
That said, among the handful of studies comparing gradients detected within individual, versus area-based socioeconomic measures, they typically have found more consistency in results with block group and tract compared to zip code level measures.
So, to offer some interpretations, starting with level of geography, our finding that the tract and block group level data behaves similarly, where zip code level data were more problematic, was consistent with our expectations.
In addition to noting problems introduced by the spatial temporal mismatch between zip codes and U.S. census-defined areas, we remind you that zip codes have been replaced by zip code tabulation areas, or ziptas(?), in the year 2000 census.
This effectively renders moot the possibility of simply using people's mailing address zip code to link to U.S. census data, because zip codes and ziptas sharing the same code may, in fact, encompass different geographic regions.
Second, with regard to ABSMs, what stands out is the robust impact of socioeconomic position on health, in that, by and large, whatever measure you use, with some exceptions, you can document powerful socioeconomic gradients in health.
That said, the most sensitive ABSMs across all outcomes were those measuring economic deprivation. Here, it is striking to note that the single variable measure, percent below poverty, performed as well or better than virtually all the more complex composite measures, which are far harder to construct, and definitely harder to explain.
Third, we think it is important to flag some unanswered questions we are right now in the midst of addressing.
One pertains to whether our results, shown for the total population, hold for different racial ethnic gender groups, and our preliminary results suggest they do. We are in the midst of cranking out those numbers right now.
The second pertains to the multilevel nature of our data. We are just now completing analyses investigating whether the nesting of blocks within census tracts matters for estimate effects, and our tentative answer is sometimes, only in cases where both the ABSM and the outcome exhibit strong spatial clustering.
We likewise are exploring the contributions of different levels to the spatial distribution of outcomes, and finding that geographic variation at the track level, at times, increases when we take into account individual level data, contrary to what typically is expecting. We will be reporting on these findings in the scientific literature in the next year.
So, based on the evidence you have seen today, and our additional analyses underway, our tentative conclusion, drawing on both our a priori criteria, and also several desirable attributes of indicators, as summarized in this slide, is that efforts to monitor U.S. socioeconomic inequalities on health using ABSMs will be best served by those tract or block group measures that are, one, more attuned to capturing economic deprivation, second, are meaningful across regions and over time and, hence, use a priori categorical cut points, third, have little missing data and, fourth, are easily understood.
In our view, the best candidate ABSM meeting all of these criteria is the percent below poverty at the census tract level.
In conclusion, then, the monitoring of social inequalities in health in the United States requires that health departments collect data on both race ethnicity and class. Both matter.
The realities of socioeconomic inequality and the impact of past and present racial discrimination, both economic and non-economic, mean that we need to have data on both of these dimensions of social life, if we are to generate informative data on the distributions and determinants of population health.
Moreover, we need identical on race ethnicity and socioeconomic position for the numerators and denominators, hence, the critical importance of working with the U.S. census categories and counts, understanding at the same time what their limitations are.
Our overall recommendations are that, first, more work needs to be done to assure the consistency of the coding of race ethnicity across public health data bases and the census, and also within state health departments across their very different data bases.
Secondly, as important, all U.S. public health data bases should be routinely geocoded and employ both standard area based socioeconomic measures that could be compared across states and over time, so that we can routinely monitor socioeconomic disparities in health within racial ethnic groups, and routinely assess the contribution of socioeconomic inequality to racial ethnic disparities in health.
Lastly, as a resource to the methodology we proposed, we refer you to the following papers from our public health disparities geocoding project, some of which are already published, some in press, with a promise of more underway.
I also have two additional slides, and I gave this information to Audrey before, for handouts, on other references, for studies I have published using socioeconomic measures -- Gracie is holding them up in the back -- and also articles pertaining to the conceptualization, measurement and analyses of social inequalities in health, involving race ethnicity, class and gender. Thank you very much.
DR. MAYS: Okay, we will open the floor. Dan?
DR. FRIEDMAN: One comment and one question. The comment is that the gradient charts that you presented, I think, are going to be extraordinarily useful at the state level, and they pack a real visual impact, and they present the data in a way which, as far as I know, I have not seen them presented before. So, I really think it is going to be a terrific addition.
Secondly, several years ago, you did a study which I believe was published in Public Health Reports, looking at state surveillance data bases and the collection of, amongst other things, economic position measures.
I realize that there is, you know, conceptually, as well as methodologically, a difference between area-based measures of economic position and individually based measures.
I have a two-fold question, one of which is, what is your intuitive sense about how realistic it would be to encourage states to add measures of economic position to some of our surveillance data sets -- in other words, individual economic position.
Then, secondly, what is your sense of what the value added, if any, would be by doing that.
DR. KRIEGER: Great. First, I am glad that visual presentation was helpful, of our results. That is actually what we are aiming to put out in the chart book that we develop.
We are just right now actually beginning to start to figure out what this report is going to look like, that we are going to give out to all the states, and we are going to be definitely talking to you and also Jay Bruckner, and we are looking at different NCHS reports and trying to figure out, visually, what makes sense.
So, we are literally just sort of starting that, as the end of this project comes into sight.
On the collection, part of the issue, obviously, is when you have things like cancer registries, they are relying on what is in the medical charts.
Until and unless people start routinely asking questions in medical histories about people's social and economic experiences, there are going to be real limits as to what you can actually get from those data.
I am sure you have maybe seen the recent issue of Public Health Reports. There are real questions about the quality of educational data on death certificates.
In fact, we found in our own data you get that funny upside down U-shaped curve, if you go by the individual educational data.
People with lower education, there is embarrassments, there are misconceptions, there are other problems with reporting. They end up getting put into the higher level of education.
So, they end up, their rates get artificially deflated of mortality, in the lowest end, so that you actually end up, I think -- it is a point we have been talking about on our team -- you end up dealing with error in a different way when you use the census data.
I mean, they both have problems with error, but I think that when you are dealing with death certificates and education levels, that is very different from asking someone, at that point in time, what their education level is.
Education is a tricky variable. People want it to mean many different things. I think one thing it is very useful for is probably childhood life circumstances. It says something about the kind of family and resources that they had, what you grew up with.
As we move to more and more of a life course perspective, education, I think, is valuable, but many people say education is great, because it is stable. The whole point, actually, with a lot of problems in life is that your economic position fluctuates.
Greg Duncan has shown, for example, that it is fluctuations in income and poverty that have a major toll on people's health. It is not just being poor at one point. It is the fluctuation of poverty and going in and out. Education is a constant measure and doesn't capture that.
I am not sure the area based measures capture that either, but I think that sometimes, in the language I have seen for why education is the variable to use on birth and death certificates, it ignores a lot of the complications and problems with educational data.
So, I think that there are questions that could be asked with some of these systems, and it would be helpful to get that. In fact -- and maybe, Sabu, you might want to speak some to this, in terms of multilevel frameworks -- you actually can get a much richer analysis that may be more pertinent for etiologic studies than for state monitoring, I am not sure, if you were able to have both individual and area-based data.
The point is not to say it should be one or the other. Both contribute information, and the area-based measures are not a poor cousin, if you will, to the individual.
To the extent that I, as an individual academic, don't know what the resources are to ask for these kinds of data. Obviously, you always want to say the more the better, but I realize that it has a cost.
The good thing about the geocoding reports is that, once those data are collected, the address is there, it is possible to geocode accurately, and we have actually been investigating some new sources of geocoding that would be less expensive and more fun than cleaning addressing like we were doing.
You have it right there, and you also -- the good thing is that you have that data and it is consistent wherever the census has been done. So, you can actually do regional, state, et cetera, comparisons.
DR. LENGERICH:: Thank you. It was a very nice presentation and display of information. I am just wondering of there are any issues around generalizability outside Massachusetts and Rhode Island, that you might see.
DR. KRIEGER: That is a great question. We have put in for competitive renewal a study that takes one state from each federal region to look at mortality data for all different kinds of outcomes exactly to address this kind of question.
We have chosen the mortality outcomes that showed the kind of marked variability for sensitivity for different ABSMs. So, that is absolutely our next goal.
DR. LENGERICH: I guess, as a follow up, then, does that include largely rural areas? Will that be a criteria for selection, then, as well?
DR. KRIEGER: Yes, that is part of it. The other point that I want to add that we are going to be addressing in this study, and it speaks to concerns that Dan raised as well, is the temporal question, which is how good is a census estimate good for.
In other words, you are collecting data and here it is 2001 and you are still stuck using the 1990 census data. Is that a problem or not.
So, we are actually going to be doing part of this study to see what happens -- if we get funded, keeping fingers crossed -- to look at what happens for outcomes that are smack in the middle, 1995-ish. If you use 1990 or 2000 census data, do you end up with markedly different estimates.
I think one thing that is pretty clear for these kinds of gradients that we are seeing, I am sure there is going to be some variation, but when you have got an 11-fold difference, it is pretty big.
You may end up with a little bit up or a little bit down for some of these approaches, but that is pretty hard to explain away by various forms of error and bias.
MR. HITCHCOCK: I have one comment, but Gene has prompted me to think of something else to just mention, and that is my colleagues at NCHS have told me that the Boston area is probably among the hardest in the United States to conduct a survey in. Response rates tend to be really, really low.
That is not a problem among lower socioeconomic class folks as it is people in middle income and higher. They just don't want to participate.
What I was going to say is that, it is going to be encouraging, I think, that it just so happens that the HHS Data Council, which is our internal group that is parallel to this, basically, has a presentation next Wednesday on geospatial coding.
There is a White House Executive Order encouraging all departments to make better use of geocoding, and there is also an OMB directive along those same lines.
We are going to be taking a close look at that starting on Wednesday. I would like to say, we have got a lot of data collected already, and most of our surveys go back to the raw data sets, where you would see census tract information, identifying information. They haven't been available at all to outside resources.
DR. KRIEGER: Two comments I would like to say. One is, yes, and there are Healthy People 2010 objectives as well about getting 90-odd percent of the data bases geocoded.
I think there are things that I am concerned about. What I am starting to see, as I am reviewing manuscripts for different journals is that a lot of people, I think, in public health, or at least some of the epidemiologists -- I won't generalize beyond that -- are a little naive about geocoding, and they just sort of send it off somewhere and don't check on the accuracy.
As I said, we found 44 percent accuracy in one firm that also charged twice as much money as the others.
The thing that I think is important to note is that this should be seen as an assay. You know, you deal with sensitivity, specificity. You have to test it out. You have to actually pay attention.
I think sometimes the social variables get less attention or are seen as less rigorous than some of the biomedical variables.
So, the basis of geocoding, what the base maps are, needs to be specified. I think that is something that, if health departments are going to start doing this, they should say what their methodology is.
There is a difference between sending your data elsewhere and how often those base maps are updated. For people who are using their own software, like ARCview or whatever, there is an enormous learning curve to actually get it right.
Just like you have with the search on deaths, you can set different parameters for what the match might be.
I think that, if this is going to be used more, attention to the rigor of that methodology is going to be very important.
Second, if there were a standard chosen, geography and also area-based socioeconomic measures, it would be possible to release data that would not do anything to people's confidentiality.
You could say, this person lives in this census track, is this percent below poverty, and not say anything about where it is. There is no release of the person's confidential information, and it would be subject to the same kinds of suppression rules as possibly other data.
I think that it is possible, with this kind of using publicly available census data, it would be possible to actually come up with a methodology that really could be used without violating confidentiality concerns, if that kind of work were done in house.
For this project, we have had some of the data, like the mortality data, have been given to use to work with and analyze at school, at Harvard. The other data, we go in with our laptop and work at the health department, because that is what the confidentiality stipulation is.
What we come back with is only the aggregate level data, and that is what we crunch through our files.
DR. NEWACHECK: I just wanted to follow up with you, Dale, about this new directive or this new set of directives. Does that mean that there might be some loosening of the confidentiality constraints like the NCHS operates under to permit identification?
MR. HITCHCOCK: I wouldn't say loosening.
DR. NEWACHECK: How does that translate?
MR. HITCHCOCK: I don't think we can really say at this time.
DR. NEWACHECK: It seems like one of the big obstacles for like the health research community is having to go to specialized data centers to be able to use that geocoded data.
MR. HITCHCOCK: What Nancy said made some sense, how they went about doing it, and there might be something like that on the federal level.
I think you could probably get very close to that now with the NCHS data center, if you were to go to Hyattsville.
DR. NEWACHECK: Exactly. That is the problem.
DR. MAYS: Let me take some questions from the audience. Do you want to go to the mike and identify yourself?
DR. SUBREMANIAN: Subu, from the School of Public Health and the Geocoding Project. A couple of comments. One, it is extremely gratifying, Dan, from the department of public health, that you are appreciating the kind of science we are doing. I think this is precisely what our motivation has been in this project.
Two comments. One on this notion of individual socioeconomic position, or a variable, versus the area-based socioeconomic variable.
I would just reiterate what Nancy said. There are key differences between how we conceptualize these. I think if we have the individual data that is appropriately measured, I think that, with the given level of methodological sophistication that we have, we could obviously enrich the analyses.
For instance, what does economic position buy you in terms of health. It could well vary across places. Now, that is a point that we can validly study if we have the individual economic position.
So, clearly, there is a case for that, obviously, and one could talk about how we could do that.
I would really like to stress -- this is what Nancy and I were discussing on the train today, some of the recent models I have run, multilevel models -- so, having controlled for age, gender and race ethnicity, and given its limitations, we actually found that the between block group variation and between census group variation in Massachusetts actually increased, when we took into account these three variables.
That is suggesting that the place variation actually increases. I was telling Nancy I think this is a very strong substantiation to look much more upstream, into the context in which people live.
Just one more thing, which we are again trying to think of constantly, is this notion of geography and the multilevel aspect of this geography.
We have these maps at block group levels, census tract. These are clearly units. At one level, they seem like everybody knows about it, but I think key mind sets are based on these levels.
This just drives us. So, it is not just independent. It is not in a vacuum. The linkages between these two levels, I think that is what we are really exploring in terms of the scientific objectives and how connected are these, and are there different approaches one could take.
I just wanted to add to what Nancy said and comment on what Dan said. Thank you.
DR. MAYS: Nancy, let me ask a question in terms of racial ethnic minority groups. That is, sociology has been doing some work in which they are demonstrating that things like, for example, education, income, that they don't have the same meaning, that they don't have the same value, if you think of Oliver and Shapiro's work, and we can take that, the difference between blacks and whites.
How, then, do you overcome that when you are doing this kind of work?
DR. KRIEGER: Vickie, that is a great question. What we are doing now, the studies we are doing, we say, let's take each of these different racial ethnic gender groups and let's just see, flat out, do the area based socioeconomic measures line up in the same kind of way in terms of being sensitive or not to different outcomes. Could you validly use one measure across lots of different racial ethnic groups or not. That is an empirical question.
Again, we have just really managed to look at the mortality data thus far, and getting to the cancer incidence data.
It is looking like it is the same story, that the economic deprivation variables pack the most punch, even in the different racial ethnic groups, stratifying race, ethnicity and gender.
Now, if there is an effect modification, what does that mean? In other words, we have there bigger effects, more powerful gradients detected by the poverty measure.
Well, that becomes an interesting question to explore. So, it is important, within the group, to understand what is going on and then, in terms of what the validity is for the between group comparisons, it provides more information, but it has to be done with that caveat of what are the variables actually capturing.
When you think about area-based measures, though, it is a little bit different, and we can bring in some more information about the areas, where people are, that it is not just what an individual person's odds are at giving a job at that level of education that has that kind of economic return, which is where some of the discrimination factors fit in.
So, this is what we are exactly looking at. Like I said, we have only done this for mortality. We see that we do get different sizes of effect.
One question that we have, for example, right now, that we are just about to run the tables to see is, when we see this stronger effect, let's say we have people that are in an area that is 20 percent below poverty.
Well, are the African Americans poorer than the whites in that area? Is that part of what the reason is? If that is the case, there are ways that you could, again, bring that into the analysis, if you are trying to do a comparison that is etiologic in nature.
DR. MAYS: The question, though, would be, for example, in trying to find that out, what are the variables that you are going to use?
In finding that out, for example, if we just take income, then we already know we have a problem. If we use, again, Oliver and Shapiro's work, they would say, people may make the same amount of money, but the taxes aren't, in terms of their ability to purchase and things like that. Give me a sense of how you get the good data?
DR. KRIEGER: Again, I think that is why it is also important to make that jump from individual level data to area level data.
There are lots of different problems that are going on, once you are in an area that is poor. It is missing supermarkets. It is missing all kinds of different things.
I think that these are capturing area-based phenomena of what happens if you live in that kind of area. Whatever is going on with you personally, you are in that neighborhood and something is happening.
So, I think it has to be another level of research, which I actually haven't seen done yet, with those area-based measures to see what those things mean, except that people are starting to contextualize areas.
That was Subu's point, about keeping the same geographic unit and understanding that there is a lot that is creating what that geographic unit is. It is part of the conceptual work that has to run with these models.
Again, I can say right now, empirically, we are going to sort through empirically and see whether we get consistent or different patterns of effect modifications for these different variables, and it is uncharted work for us. We haven't seen other work like this. So, it is a big empirical question, and in six months I will be happy to give you a more detailed answer.
I think that where it would matter the most importantly is remembering, first, that the analyses within group are of critical importance. This isn't just about between group explanations.
So, within groups, whatever it means, that is what it means within that group. Then, for the between group, when you are doing the adjustment, well, there are a lot of caveats that go into doing adjustments for anything, and one would have to be very explicit about that conceptually as well as methodologically.
DR. MAYS: I think it will be interesting to see, in terms of Eugene's point, whether or not it will hold in other areas.
I know, for example, you could take Illinois and take very specifically Chicago, and people could live, people who are poor, end up having to move a lot. They will end up in neighborhoods but they are poor, and the quality of difference -- I mean, to live on the south side of Chicago and be poor and live on the west side of Chicago and be poor is really kind of different. It will be interesting to see, in terms of people's mobility and other factors.
DR. KRIEGER: We realize, looking at the wonderful work that NCHS supported on cognitive issues about mapping, those were done, however, primarily with professionals in mind. I mean, they are the ones they based the cognitive tests on.
So, even trying to figure out what maps mean to different people and what the issues are and how you use the different colors and how you can present spatially the data in ways that are meaningful to different kinds of audiences, is another thing that we want to look at, and would start to get to some of your questions about -- and we are also, with that, bringing in other non-census data.
This project was explicitly designed to use readily accessible, easy to get, uniform data that is available throughout the United States. I think that meets one objective.
I think the questions you are asking bring us to another level of analysis, which is very important, but we have to do this first, to get that broad monitoring on the table.
DR. MAYS: I want to thank you for an excellent presentation. Again, I think that what you have put on the table for us to consider will help us in terms of shaping the kinds of recommendations that we can make about the types of things we think should be done. Let's take a five-minute break. Then we are going to turn to our other two states.
[Brief recess.]
DR. MAYS: We are going to get started. For those of you on the internet, we are back after a very short break.
We have been hearing from states throughout the morning. We have two other states that we are going to hear from. We are going to start with Alabama, and from Alabama we are going to have Dorothy Harshberger talk to us, who is the state registrar for the Alabama Department of Public Health.
Dorothy is a known entity to us, and she has been useful over the years in this area. So, thank you, and we are appreciative of your time with us.
Agenda Item: Alabama Department of Public Health.
MS. HARSHBERGER: Thank you for having me. Yes, I have been around for a long time.
I wanted to show this. This is the home page for the Alabama Department of Public Health web page. I have also handed out a brochure that tells how to get into the data for the center for health statistics.
The data that I am going to show, and the charts that I am going to show, are mostly on the web page.
DR. MAYS: Can I just make a comment, then, for those people who are on the internet? If you go to www.adph.org, they can then see the pretty home page.
MS. HARSHBERGER: Okay. One of the things that you asked us to look at is population for the state. We have about 4.5 million people. As you can see, about 70 percent are white, 26 percent white, and approximately three percent, 2.9 percent, other.
The other is broken down into some very, very small groups, .7 percent Asian, .5 percent Native American -- these are census as they identified themselves -- .7 percent other races, and only one percent in the state of Alabama marked two or more races in the census.
This leads us to some problems, obviously. We have small population groups in some of these which, if you are trying to get data for them, it gets very difficult.
It may also eliminate some problems when we are trying to work with data going back and doing some bridging, because we don't have some of the problems other states do on the bridging issues, because we have so few that marked more than one race.
I think it will become more of an issue as we go forward in time, and people start identifying themselves that way, but it is not quite that issue today.
This is a map that shows the distribution of the black or African American. You can see there is a large belt of that through the central part of the state.
The northern part of the state is much more white. It is quite a different population there. So, we are dealing with a lot of different groups across the state.
Our Hispanic population is becoming an interesting issue for the health department. We have about 76,000 Hispanic or Latinos. It is only 1.7 percent of the population, so it is still a very small group.
You will note, I put up there, that the increase from the census in 1990. So, we increased tremendously. When we look at the breakdown, we have a lot of them that aren't giving anything. They weren't classifying themselves into a particular group.
About 60 percent are Mexican. Then, the other group is broken down into Central American, Cuban, South American, which is only 2.7 percent, and Puerto Rican, which is about 8.3 percent.
One thing that I would like to mention about our Central American, about half of those are Guatemalans. As we are working with the health department, we are finding they are Guatemalans with relatively little education and they are a particularly hard population to deal with.
This is the distribution of the Hispanic. Much of it is in the northern part of the state. Their percentages range from virtually zero in a county to about eight percent in the county. In the northern counties, we have a lot more.
The Hispanic population in the northern part of the state are the ones that come in from Mexico or Guatemala to work in the chicken processing plants.
In the southern part, where you have some around Baldwin and some of the southern counties, they are migrant farm workers.
I want to show this one. This shows the birth rate by race. These are some of our traditional ways of doing things. We have shown these data for years.
I want to point out that we are having a problem with population data, in the sense that we have to break our population primarily into white and black and other. There is a reason for that.
That is because, in the years between censuses, we don't really get data that is broken out in more detail. That is a big problem for us.
The other thing, you can just see approximately there, the black other rate is about 16.5 per thousand and the white rate is about 13.3. I wanted to just point that out because I am going to show you show you something a little bit later.
This is the Hispanic population. These are numbers. This is the births to Hispanics in Alabama. You can see the tremendous increase that we have had in the past 10 years. We are up to about not quite 2,000 births.
One of the things, on the Hispanic, with 1,900 births and 76,000 population, that gives a rate of about 25 per thousand. It sounds a little high, a little out of whack with some of the others. It means that either we have a very, very young population, which is true in some cases, but it also means that the census is not counting everybody.
Because of the movement of these people, these are the ones that are really hard to count. So, we are not picking up all the people that are coming into our state, or that were in the state when they did the census population.
Moving on, when we look at the data, this is all data that we have, and it is out on the web. We did a chart book for maternal and child health people. So, I have pulled a lot of data, charts and things we did for that, just to illustrate some things.
This is the method of payment. We ask, on our birth certificate, primary source of payment for this delivery, so we can get our Medicaid information.
As you can see, there are quite a few differences in the population groups, for those paid by Medicaid. The blacks have nearly 70 percent paid by Medicaid.
When you look at the Native Americans and the Asian and Pacific Islanders, you can see they are different. One of the things I would like to point out is numbers here.
White births, about 40,000, black births, about 20,000, when we are talking about Native American, 171. So, the number is extremely small. We can do this, but you are dealing with very, very small numbers. Asian and Pacific Islanders, 525. So, the numbers are very, very small that we are working with here.
On Hispanic, the method of payment is different. You can see that nearly one third of them are self pay. Basically, these are the people who don't have insurance. They have to take it on themselves. So, it is a different group we are dealing with. It is one that there are a lot of problems within the health department.
One of the things that I think you also ought to be aware of, in the south, at least in Alabama and many other southern states, the county health departments are all part of the state health department, and they provide care. They provide direct services to a lot of the population groups.
So, we have people coming in for prenatal care, et cetera, in the clinics, and they come into the health department clinics for a lot of their care, as far as their health is concerned.
This is the percent of the mothers who had adequate prenatal care. You can see that it does vary. Again, we are dealing with small groups, but we can get a little bit of data about them, as long as we are staying within the context of the particular program, and looking at percents.
You have a lot of problems when you are going out and looking at it, trying to relate it back to your overall populations and using population data as denominators.
This is our Hispanic population and their care is much lower. Only 48 percent had adequate prenatal care.
I threw this one up here. This is our standard, again, our infant mortality, to answer a couple of questions that you asked about presenting data on disparities.
Data on disparities has been presented for years. I mean, this is the traditional way of doing this. We have always looked at the breakdown of white and black and other infant mortalities as one of the big indicators. So, it is not new to present data.
I think we are presenting a lot more data on a lot more groups now than we ever have before, and we are trying to do a lot more with the different groups.
The other thing that I wanted to mention with this one is, when you are talking about the misclassification of data, people keep telling us that the death data are wrong, and we have heard that.
We are looking at different things. The question is asked about different ways. When we are talking about birth data, which is a denominator here, we are talking about the mother's race, and generally what the mother identifies herself as, when we are doing the birth certificate.
Death data are taken from the death certificate, identified by the family. So, where you have a mixture in the family, you will have differences in the data, and are they wrong? It depends what you are really asking and what the context is. You are comparing different things, I agree, but is it wrong? I can't answer that.
Other things, I just want to show quickly. We do a lot of data for programs. This is smoking and showing low weight for smoking status, for helping in the tobacco prevention programs.
We try to tailor a lot of the things that we do specifically for some of the things that they want. The whole book, incidentally, was done for maternal and child health, and some of the things they needed for maternal and child health.
This shows infant mortality rates for teenagers and adults. We have a lot of programs that deal with teenagers, teenage pregnancy, et cetera, that want data from us all the time.
Again, they want more data, they want it by county, they want it in a great deal of detail. It is really had to do that, to get down to the county level. We have small populations in some of the counties.
I wanted to show you this one for two reasons. This shows the current death rate, and it is a three-year rate versus an age-adjusted rate.
You can see that the whites actually have a higher death rate in Alabama than black and other. However, when you age adjust, it reverses quite a bit, so that the black and other rate is higher than the white rate.
Of course, we are finding that the black and other die at earlier ages for almost all causes of death.
That leads me to, I have passed out the book, the Alabama Book of Racial Disparities in Mortality that we just did.
We worked closely with our different programs, as I said. This was done for the Office of Minority Health. They specifically wanted something that they could have to look at to use to talk to people out in the communities.
That is why we did it as an atlas. They wanted county data. They wanted something by counties. We had a hard time doing this to show, something that they could go out and use, and this is what we came up with for them to use.
Even in using three years worth of data, we had some problems in doing rates, when you get down to detailed causes of death. You just don't have the numbers of death in the population available to do them on. So, we even, on this, had some places where we just couldn't present the data.
Some of the data issues. As I have already mentioned, the lack of denominator data in non-census years. We just don't have the data to use for looking at these different population groups in between census years, and it is a big issue.
We don't even have data on black, really, it is combined black and other which is done by our state data center, which are the data that we have used. So, that has been a big problem all along.
We are doing much more to try to attempt to break them separate, and I think we are going to get better data in the future.
Of course, Hispanic, we have nothing for the 10 years that we had that huge increase. When we try to look at what is going on in that population group, we really have had no data in between.
There has been no real way to really go out and even tell where they are, other than what we are seeing from other areas, where they are coming into the clinics and where we are getting the births and deaths. We can tell we have a growing population there, but we don't have the denominator data to really look at it.
The one thing that I would like to mention that nobody else has said anything about today is the American Community Survey.
This is something that the Census Bureau is proposing to do, and they are actually testing it out now. The American Community Survey would be the thing that would give us the data in between years.
The idea is to do a survey like the long form survey on a continuous basis, so that data would be available for small geographic areas.
It would be a five year group data for some of the very smallest areas, but once you get through that five years, you would have the data available to use.
There are a lot of issues, a lot of technical issues and everything, but I would like to ask those of you who can support this, or have factors to put in to support this, I think this is a very important thing. It would certainly help us to have a lot of detail in between census years.
Other data issues, you have heard these before, the data limitations with small numbers. It is a big, big problem when we are trying to look at these different groups, to try to go out there and get something to look at.
Survey samples, when they were talking about going out and doing a survey, you were talking about telephone surveys, even mail surveys.
When we are talking about some of the population groups that we are dealing with here, particularly migrant workers or whatever, they are having births in the state, but they may or may not be there.
The people that come in to work in the chicken processing plants live in trailer parks, oftentimes many to a household. It is not easy to locate these, even if we are doing a mail survey. So, it is an issue when you are trying to get data on these groups.
Just targeting programs, we tried to provide data for the diabetes program, breast and cervical cancer, all kinds of screening programs, hypertension program, all these different things.
Everybody, since we are working with the local county health departments, wants data for their county health department, and it is difficult to get that and get the detail, particularly by race, when you are dealing with the smaller groups.
It is not so bad in some of the larger areas, of course, but it is when you are dealing with some counties.
The other thing is the differences in reporting requirements across programs, across data sets. Geography is one.
We are being asked to provide data in all kinds of geographic groupings. You have the issue of, we have perinatal regions, we have health regions, we have all sorts of regions there that we are trying group data for.
We have the problem with urban rural definitions. There are different definitions for them. How do you define these things?
Some of the things that people want to collect are for federal programs, for grants, that have different definitions in geography.
The other one is the race ethnic categories that you have already heard about, and the differences in program. I really won't say a whole lot about that.
One thing that I did want to mention, though, while it is nice to have a way to collect data in a standard format for everything and to be used for everything, it gets very difficult when you are working in a state office.
I think Bruce mentioned it a little bit this morning. Asking the state to collect data in all sorts of very very small categories, to try to get the data that may be wanted at the national level for very small populations, where we have maybe one or two in a population group or something like that, it gets very difficult to collect that data or have it meaningful.
We have to carry a lot of data that we can't use in the state because it is meaningless to us. It would be nice if there would be some way that we could work out a way to get the data that the states needed. In some cases, they need more detail than at the national level, but in a lot of cases, like in states in the south, where we don't have a lot of the different populations groups, it is a big problem to try to collect all those data when you are carrying lots and lots of breakdowns, and you have to keep them on your computer files and keep them in files and put out a form for people to complete, and it doesn't mean anything to them. So, that is a problem within the states.
Other than that, as I said, we don't have as many problems as the other states do in looking at tabulating across the bridging, for example, because we don't have as many groups to deal with. We don't have the multiples.
We don't have all the luxury of having the surveys and the household surveys. We don't have a discharge data set.
We are one of the poor states, by the way. We still use a dos system for collecting the birth data, due to the fact that it costs so much to have the vendor change it.
We want very much to change it, and we are working to do that, but the vendor wants a great deal of money to do that.
I know you all said that, get your health officers aware. I think they are aware. At least some of them. Some of them may not be aware of the vital records data and the usefulness of it.
It is really hard in a state, when budgets are being cut and everything is really tight, and it is particularly bad right now.
Money is not available anywhere. When it comes down to particular in a state in the south where they are providing care, are you going to cut money to not give poor children care to do vital records and collect better data, or are you going to do the care. I think you know the answer to that as well as I do.
So, there are a lot of issues there and money, of course, is the big one. I don't think there is anybody that really doesn't want good data. It is just that it is an expensive process to get it sometimes.
I will basically end there. Here is my contact information and, just to end, this is the civil rights memorial in Montgomery.
DR. MAYS: Thank you.
Agenda Item: Tennessee Department of Health.
DR. URBANO: I can make my presentation very short. Add a million people to Alabama, and you have Tennessee.
I took the 11 questions that I got as my outline for what I want to talk about. Basically, Tennessee is a white and black state, 83 percent white, 16.3 percent black, and that pretty well covers it.
If you look at the distribution of race across Tennessee, the southwestern counties around Memphis are predominantly black. The rest of the state is white. We have some counties where I don't think there are any non-white individuals in those counties. There may only be 10 people in the county, but they are all white.
In addition to Memphis, Nashville, Davidson County and Chattanooga have about 20 percent white. a couple percent, 1.4, 1.5 percent Hispanic. Yes, we do, on most of our data sets, have the question, Hispanic, yes or no, plus black, white, other or black white and, on the birth certificates, the categories.
I guess you used Genesis also for yours? We are in the same boat with our electronic birth data system, dos based.
Fortunately, we raised the fees for birth, death, marriage, divorce and cremation permits in the state last month and we were able to fund, with those additional things, the revision to go to windows based reporting.
About a third of our hospitals right now actually use the internet to send their birth files to the state for inclusion, although they are running dos-based programs.
We surveyed the hospitals asked them, of those 96 birthing centers and hospitals in Tennessee that are using our electronic birth reporting, the Genesis system, how many of you will be switching to Windows XP within the next two years.
Genesis told us that they would not run, that the software would not run under XP. Half of our hospitals said, if they weren't already there, they were moving there. So, we had a great motivation to move in that direction. In fact, next week, Genesis will be on site to work with us on the revisions that we want to make to the birth certificates. Our goal is to move to the NCHS standards for those, and go to internet transfer of the data.
The question about ethnic country codes, in our fetal death, induced termination of pregnancy, our emergency medical service, we do have a hospital discharge data system, and the traumatic brain injury system, in addition to our vital records systems, do report ethnicity and race.
Will the same race ethnicity standards be used for data collection and reporting? No. Yes, we will go to the expanded recording of ethnic data, and expanded race data, but we have such small numbers that most of the state level reports that we do, it is black, white other, and if you go back a few years, it was white and other, for many of those older reports.
On the state portal site, if you go to the HIT component, Health Information Tennessee, we have almost every report we produce.
You can go in there dynamically and look up data by county, by race ethnicity, and a bunch of various other variables.
On our health statistics and research side, we have PDF files of every state report we generate. The reason we did that, we had a little budget problem last year and one of the things they said is, you cannot print any reports.
So, all of our annual reports could not be printed. So, that was a great motivation to create PDFs of all of them. We had some tax increases and the budget is a little bit more flexible right now, so our major reports we can start printing again.
I think it is a great thing that we were able to do with the budget problems, to make those reports much more widely available through the web.
Do you have significant misclassification in categories? Well, yes and no. We are sure we have misclassifications, and our field reps, who do the training on reporting of vital records say, you go out there and look and basically the people reporting know black, white and other. Unless you have the primary reporter giving the data, the rest of it, they don't have much confidence in.
Funeral homes collect the data on most of the deaths. The county certifies those and transmits those. We are even less certain of the quality of the data that we are getting in that regard.
NCHS standard birth certificates, our goal is January 1, 2004, to go statewide with the modified forms.
Do we think there is a problem with bridging? No, we have so few, small numbers that bridging will not be a major issue for us.
Do we report racial and ethnic data routinely in reports? Yes, I would say every report that we do does have those breakouts.
We are doing another thing in Tennessee that all of this impacts on. We are using our vital records systems plus the other administrative systems -- hospital discharge, our TennCare encounter.
Tennessee is a waiver state. Tenncare covers about half the children in Tennessee, or I should say that it does as of last week and, with some changes, we don't know how many will be covered.
About half the births in Tennessee are covered by Tenncare. As part of our health data system, we get the Tenncare encounter data, and we are in the process of creating, I don't want to call it a data warehouse, but a consolidated set of data bases where all these sets are linked.
Part of this is to create a child health profile, where longitudinally and across data sets, we link the information that we have on children, so we can see what is happening across time.
In Tennessee we have an initiative called Childrens Information Tennessee. Case managers in five agencies in Tennessee use these data. They have on line access, web access, to the information.
Health, human services, children services, Tenncare provides data, and education, primarily our early intervention program have access to the data.
Depending on how frequently those data sets are updated, we get the data into CIT and do those linkages, make them available for our epidemiological programs and for research and for the case management.
Access to the information is controlled by parent permission or guardian permission. So, a case manager can go in and look up Johnny B. Good. To see what services and what programs Johnny is in, electronic recordings of the permissions to release that information have to be there.
Gathering those permissions varies across programs. So, if you go in and you don't see anything on Johnny, what you can do is send a blind e mail to the system, since the system knows who all the case managers are and say, I have permission to do case management services for Johnny. If you have like permission, contact me, and let's make sure we are coordinating services.
We are expanding that program to include private practice pediatricians who are doing our EPSDT exams in the state.
Vital records are used two ways. We wanted a background demographic file so that, if you pulled up a name, you wouldn't know whether the child you pulled up in the data set was in a special program or not. All you would know is two things in our set. They were either born in Tennessee or they got an immunization in Tennessee. They also could be served by a program in Tennessee if they weren't born there.
So, we have about 1.8 million children in the system, demographic records on kids in the system right now and, depending on the program, between a couple hundred thousand and, for small programs, about 1,500 service records on those children. Any questions?
DR. MAYS: We are actually going to open it up to both of you, if you are finished. If you will join us at the table? Okay, let's open it up to questions.
DR. LENGERICH:: I guess starting with Tennessee, so are the local public health people employees of the state health department? Is that how that works, it is the same as in Alabama?
DR. URBANO: We have two kinds of relationships, the metropolitan areas -- Memphis, Jackson, Nashville, Chattanooga, Knoxville and Johnson City -- are metropolitan areas.
The health department, although it is under the state, actually operates under local government with contract.
The rest of the state is in seven regions. They are rural regions. The rural regions are directly under the Health Services Administration within the state, yes.
DR. LENGERICH:: That leads to the second question, then, do those metropolitan areas collect different information or request additional information, more detail than you necessarily can gather at the state?
DR. URBANO: No, they don't. In Tennessee, we have a system that is used in all the health departments for service. It is called the patient tracking, billing, management information system. That is a regionalized system, but it is consistent across all programs across the state.
Unlike Alabama, I think we only have two rural regions that provide direct service. When Tenncare came in, the state said, we have got providers under Tenncare who are providing these services. The health department doesn't need to do it.
Except in a couple of rural areas, where we have had issues of provider availability, we got out of the business. It looks like, in the next year, we are going to get back in the business, as people are disenrolled in Tenncare.
DR. LENGERICH:: I was just curious how you were able to link data across programs. Does Johnny B. Good have a unique identifier assigned to him at some point?
DR. URBANO: Eventually, yes. First, we have a memorandum of understanding with all the participating agencies and, as a new agency comes on, all the people has to sign on, that we will share some specific limited information on all the kids so that we can do the linkage.
Almost every service program in Tennessee, including Tenncare, uses social security number as their internal. Whether it is their only ID, which is almost not the case, but they get that recorded. So, a specific identifier to link those data sets exists.
Basically, what we do is, we take the agency's unique identifier and look to see if that is in the system. If that is in the system, then we know we have got a link, and we verify that at least three other things agree on that.
If we don't have that, we have got a social security number, and from there, it is a very small percent, under 10 percent of the records, we then go and look for multiple variable matches.
Our strategy has been, if we don't find a match there, we go ahead and add the records and let our case managers beat us up because we have got duplicates in the system, and fix it there.
DR. MAYS: One of the groups that we haven't talked about the diversity in was a category that is often black or African American.
Again, I don't know Tennessee or Alabama well enough to know, for example, who comes there. Could you discuss a little bit, in your category of your black population, what that might be, whether it is individuals who have come there and they are southern born African Americans, or whether they might be Africans and they have checked black, or they might be just any number of subgroups that sometimes get categorized as black, Jamaicans, Nigerians.
MS. HARSHBERGER: I would say we have a few coming in from other areas. We have a few Jamaicans, a few Nigerians, a few from a variety of places. Primarily, in Alabama, they are going to be the people who have been born there and been there for many years, and have stayed in that area.
Alabama, I think at one time a number of blacks from Alabama went north, and some of those are coming back now. So, they went north to Chicago and Detroit, and some of those are coming back into the state now. Those people would be different from the people that have been there.
In Alabama, the blacks who have been there for years are obviously the ones that are in that middle part of the state, poor areas, rural areas. You are lacking industries, you are lacking all sorts of things.
It is a group that has lacked services and lacked a lot. I think that, while traditionally, things have not been provided there, I think the attempt has been made within the last number of years to try to provide better services, better health care, and an attempt is really being made to address some of the issues and some of the disparity issues that have existed.
So, that is one of the reasons that I think everybody is interested in the data, and really having data for a lot of these groups in detail.
DR. MAYS: Do you have ways to show the differences in those groups? For example, in terms of when you report the data, it is black. Would you argue that it is probably even more salient for you to use either the other techniques that have been discussed here today or something other than just race to talk about those differences?
MS. HARSHBERGER: Yes.
DR. MAYS: Suggestions?
MS. HARSHBERGER: Years ago, I went to a meeting, it is probably 10 or 12 -- it is probably more than that, because I hate to say how old I am, but it has been a long time.
I went to a meeting and we came out with a recommendation for SES. Many, many years ago we said that was a really important thing to be developed and should be measured in some way that we can measure something else.
Yes, I would argue that we need something else to look at, to really identify the differences in these groups, because we are not doing that now. We are not doing that now.
The other place where we see some of the really poor and some of the ones with major problems with infant mortality are in the city, in the city of Birmingham and Mobile.
You have a really diverse group, I think, of black or African American even within the cities. You have some that are doing very, very well and upper class, and then you have some that are very, very poor still.
We are really not, when we put this data out, distinguishing between those groups in any way, and we really should. We really need to.
DR. MAYS: Can you just say what you would want, what variables? You say SES and I am a master at pushing people to say what is it.
Often people will say, well, it is income. We all know, depending on the data set, either income isn't there or people don't want to answer income. So, can you tell me what kinds of variables you have been thinking about?
MS. HARSHBERGER: I don't know. Income is the one that is most often talked about. I think when we were revising the birth and death certificates, particularly the birth certificate, actually that was brought up, the possibility of putting it on there, and it was one that got very little discussion because everybody said, no, you are not really going to be able to get that data.
That would be something, if a state wanted to try it, would be a good thing to do. We do have Medicaid, and half of our births are on Medicaid in Alabama. So, we do have that on the birth certificate as a source of payment.
That is sort of a proxy that we can use, and we do a lot of work with Medicaid. We have some programs within the health department that are specifically targeted at the Medicaid populations, and we try to do a lot to identify those and where those are and we have used the data in that way.
That has proved to be a very good question and, as you noticed, I put a couple of slides up there with that on there.
That is a possibility, that if there is some program that people are in, that you can use as a proxy, that might help. That might not get you everything, but it would get you a little bit.
DR. URBANO: I would say Tennessee is very similar. We have a very low in-migration rate. I have no measure of the diversity within the African American population that I can think of right off, that we can go to.
Again, Memphis and southwest Tennessee is predominantly historically a black area. I don't have any anecdotal evidence that says we have much diversity within the black population.
In Nashville, surprisingly, we do have a large number of Kurds. One of the local churches worked in bringing people in. So, there may be small areas where that is the case.
On the SES or proxy measures for that, we are going to have discussions as we do our rework on how we may include something to do that.
I am sure that I am going to get the same argument that collecting the data consistently, reliably, will be a problem.
Since we work very closely with Tenncare, being able to look at availability of insurance, or being on Tenncare or not, being on Tenncare for either areas or counties or census tracts or whatever, we can get participation in Tenncare.
In many of the reports that we do, we look at Tenncare and non-Tenncare participations, and we can do that on a group level and on an individual level within the data sets.
DR. MAYS: Let me open it up in terms of the broader group, if there are any other questions.
MR. MELTZER: Richard, you said you were collecting Tenncare data. Have you included mental health data in that?
DR. URBANO: The question was, have we included mental health data in the current data set. Again, it is a yes no answer.
In the broad picture of CIT and our five participating agencies, as you might imagine, getting five state agencies to agree on sharing individual level data involved a lot of good, political discourse.
One of the parameters we set on doing this was that the programs that participated would be doing case management, so that the data that we collected would be available to case managers.
In Tennessee, our department of mental health and developmental disabilities basically contracts for much of the services they provide. So, we do not have state case managers, and they opted out of participated.
Having said that, we have got a great push now for ETSD, and making sure that all kids in custody and state services get complete ETSD. We have got a judge who is on our case about it big time.
So, we have got a remedial plan and, in that remedial plan, mental health is a component. We are trying to get the private pediatricians who are providing that additional service, plus the state agencies, to start reporting ETSD, and the referral services.
The largest unmet service for this population is mental health services. So, we have six centers of excellence and we are working on a way that we will have some of those data. Right now, the answer is no.
MS. HARSHBERGER: Stephanie just reminded me that we also have a question on the birth certificate of the place of birth of the parent. We could use that to get some information for people who are coming into the country who were not born here.
DR. ABBOTT: I didn't mention this this morning, but under the theory that a lot of things happen in California before they happen elsewhere in the United States, Californians, in June of next year, are going to have an opportunity to vote on something called the racial privacy initiative.
This is an initiative being put forward that basically would prohibit the collection of race and ethnicity data by most governmental entities within California.
There are some exceptions. Arguably, vital records will be an exception. Medical research will be an exception and requirements imposed by the federal government are an exception.
We may well have much less information on the diversity of our population if this, indeed, is going to be passed next year. I wondered if the other states perhaps are facing that.
DR. MAYS: Only California at this point. Can I ask a question? What would be the most helpful intervention for you in terms of being able to ask, or being able to get this data?
What would it be for you to be able, for example, to get additional data on, say, for example like -- I hate using the word SES any more, but these additional variables that can help us to understand diversity within racial ethnic groups.
MS. HARSHBERGER: Well, identifying what those variables are.
DR. MAYS: I think you can be helped on that one, but it is like whether we can have it consistent across states is another story, but if there is an issue of being able to get these things, I am trying to understand what the barriers are.
MS. HARSHBERGER: You have to figure out how to ask the questions and so on. So, if we know what to ask and how to ask the questions, we can, at least as far as vital records, we can put some of our own questions on there. States can have their own questions, and we do, and I think many states do. It is a good way to try things out, to see what works and what doesn't work.
The biggest problem that we have is just changing the whole system, which is what we are talking about doing when we go to a new certificate. We want to go completely from one thing to another thing.
One of the things that automation does is, you can collect the data and tabulate it, et cetera, but changing an automated system is not a cheap thing to do. That is the problem.
It is hard to go back, once you have the system set up, to go back and make those changes, and that is what we are finding. It is not as easy as just throw in one question into an electronic birth certificate, for example, because you have to change the thing in every hospital and you have to go to a vendor to get the software changed, and they charge you a lot of money to do that and all that. So, there is a big problem in doing it that way.
It does make it difficult. The other thing that, particularly with vital records -- and this varies from state to state -- are the issues of what you have to go through to put something on a certificate, to collect the information.
In our state, the variables have to be approved by the board of health. That involves public hearings, et cetera, to add or delete an item from a record. It will vary from state to state, depending on how their laws are set up, what they have to do.
So, that also, sometimes, makes it not easy to do some of the things in the ways you want to do them. The Census is somewhat political, too, and the questions that they ask are not necessarily the ones they think are good to ask. They are the ones that Congress says, these are what you will ask and this is the way you will ask it.
So, sometimes, in going before the board of health, you may end up with something you don't necessarily want and not ask the question in the way you want it.
It does help to have other people doing things. The one reason for this standard is good. It is support, when you go before the board of health, to say, this is the standard, this is what is recommended, this is what we need to do. In most instances, they go along with that. Sometimes they want to throw some other things in, but that is pretty much what we do.
DR. KRIEGER: This is just a comment. I apologize for not being here earlier today. I don't know everything that was discussed. I am just responding in part, Vickie, to something you raised.
When it said you can't ask certain questions, part of that depends on the leadership that is applied to ask those questions.
I would just like to make the instructive parallel to data collection in the United Kingdom, for example. There was a long history of collecting social class data which now, there are additional systems of socioeconomic classification, because the original one to five registrar general social class measure has been shown to be wanting, as occupations have diversified and as economic conditions change.
They did not collect in their census, until recently, racial ethnic data. They thought it was irrelevant and they thought it was impossible to collect.
Their first go around of collecting racial ethnic data caused some problems. You couldn't simultaneously, for example, be black and British. Obviously, that created some serious problems.
So, that has been addressed in the next census. It was seen as iterative, but it was seen that this was not something that you could ask about and not see it then as populations organized internally within the United Kingdom, and also dealt with the different immigration questions within the United Kingdom.
They started demanding those data and, guess what, you could ask the questions. Of course, there are problems and it goes through iterations.
They look to our data in race ethnicity as examples of, you can ask these things and it is actually important and you pick up trends that you don't pick up elsewhere.
They also will have opinions about ways in which the United States approaches to handling race ethnicity reflects our very complicated history with race ethnicity.
We, simultaneously, in the United States can look at some of the United Kingdom data. They have collected data on occupation, but they actually have not collected on income, partly for many of the reasons that are discussed.
Whether occupation and class structure here, whether and how that is the right way to go, what they do in the United Kingdom, is an interesting question.
The leadership for collecting the racial ethnic data came because of local dissent translated up politically to a national mandate. That played out through the census.
I think there are instructive examples that we can have cross nationally. Similarly, in Brazil right now, and it will be interesting to see, with the new elections in Brazil, what happens, there was not collection of data by color or race ethnicity. It was said that the problems were socioeconomic.
The groups that were affected organized to ask to have those data start to get into the census and also into the health data. PAHO is now beginning to try to look at initiatives in different Latin American countries to look at race ethnicity in association with economic data.
I think that one of the things that has stood out to me, in talking to colleagues and looking at these different data systems in different places, it is local groups organizing, feeling they are left out, that raises the issues.
Leadership has to be national, in order to actually then begin to be able to get data collection locally. So, there is a feedback process between the two and it is not one or the other, but without national leadership, the local initiatives stay local and are hard to compare to other things because they are local issues.
I think some recognition of that interplay between the local and national needs needs to be brought into account.
DR. ABBOTT: I agree very much with what Nancy said. I think national leadership, standards, are very important. We mentioned this morning consistency in those standards.
I think that, even though it is probably not a good time, federal funding is important. Delton mentioned, and I think Dorothy did as well, that many states are experiencing significant financial problems, including in California.
The only way that we can really address these issues and move forward, in terms of improving the scope, the quality and the reliability of the data that we collect, particularly on racial and ethnic status, is to have some funding accompany that leadership and standards.
I think it is also imperative that this be done together with the states. I applaud your involving state representatives in your hearing today. I think it is very valuable that we are here.
I think, though, as we move forward, there still needs to be a continuing dialogue between the states and the federal government, as these standards are developed.
DR. LENGERICH: Just as a follow up to that, the question, I have not seen too many states that are willing to give away their opportunity to set their own direction, but yet I also hear you talking about federal guidance on this.
There has to be some sort of intermediary group or organization. Who is that? Who sort of speaks collectively for the states, and is there some group that needs to come forward to do that, or we need to encourage or empower to make that happen?
DR. MAYS: Maybe you can say something about NAPSIS, what it does.
DR. ONAKA: I am changing hats from the representative of Hawaii to the current president of the National Association of Public Health Statistics and Information Systems. That is the acronym NAPSIS.
We represent 57 jurisdictions. They include the 50 states, the territories, the District of Columbia and New York City.
We have been in existence since 1933. So, this coming year we will be celebrating our 70th birthday.
We have, through the years -- I think our mission statement in 1933, our mission statement in 2002, have been very consistent in terms of being the representative.
I think it has been mentioned several times that the vital statistics system is a state-based system, and NAPSIS has been the glue, if you may, to speak and represent the 57 different jurisdictions. I think we have played that important role.
DR. MAYS: Can you advocate for things?
DR. ONAKA: Yes.
DR. MAYS: Is advocacy something that your organization can do?
DR. ONAKA: Yes.
DR. LENGERICH:: Is there a subcommittee or a group, then, that would particularly take up the race ethnicity question in the data system, or would that be the entire executive committee or whomever?
DR. ONAKA: As a matter of fact, NAPSIS did have a race ethnicity summit that we brought to, especially in regard to the bridging.
We had representatives -- Dorothy chaired the subcommittee for the revision of the birth certificate. I chaired the subcommittee for the revision of the death certificate.
So, we were looking at national standards that all of our jurisdictions could work with. We are a 501(c)(3).
MR. HITCHCOCK: I think you need to look at the history of how OMB promulgated its most recent change to the categories that are used in the collection of data, too, how that was accomplished.
That was basically OMB driven, and I think it originated with a group of concerned citizens or an alliance of citizen groups that bombarded OMB with letters and testimony to get the race category into the guidelines.
DR. ONAKA: I think I owe a little to my predecessors. I am the current president of NAPSIS, but Dorothy is a past president, Chris is a past president of NAPSIS. Again, you have about 20 years of leadership of the 70 years.
DR. MAYS: It might be useful for us to talk to your group at some point in time, for you to really change hats and for us, as we get closer to thinking through what we want to do, to talk with you.
It is important to us, not just that we generate some piece of paper and that we say, here is what happened in the hearing, but that we also understand where we should be sending these things as a way to try and make something move.
Just sitting here having hearings to have fun is not any of our idea -- we have lots of other things to do. I really do think that, in a serious way, we have probably underestimated and, and I thank Gene for bringing it up the way he has.
We probably should have, at some point in time, a meeting with you to talk about these issues. Bear with us, if we invite you again.
DR. ONAKA: We appreciate the opportunity.
DR. MAYS: Thank you. One of the things that is of value in terms of how the committee is put together is that different people come from different backgrounds. You know, we have people in academia, people in the states, people in the federal government.
Today, what I have done is kind of put on the spot one of our members, because he lives and breathes state issues and has been very helpful to us. He is probably why we are having a state hearing, because I have learned a lot about the difference of the issues for the state versus the federal level, and I really appreciate his pushing us to hear that difference.
What I have asked Dan to do is to actually try and do a commentary on what has gone on today. That is a tall order, but he is a very tall man in terms of dealing with those kinds of state issues. I have asked him to please share some thoughts with us about what we have heard today. I know you also want to catch a train.
DR. FRIEDMAN: It is actually going to be a short order, since I am going to excuse myself shortly.
I wanted to comment on basically three different themes that we heard today. First of all, commonalities and differences among the represented states, second, future directions that the states need to consider and move toward and, third, potential roles for the national committee.
First of all, in terms of commonalities and differences, I think one issue that came up very consistently, especially this morning, is what do we mean, in this country and in the states, when we talk about race.
I think what we heard, particularly from Alvin and Peter and Bruce is that race, in and of itself, is not an especially useful concept to the states, and we need to think well beyond race.
My own personal view is that the revised OMB-15 construction of races really doesn't necessarily represent how people in this country think of their own identities, and those concepts of their own identities in terms of race clearly changes over time. Those may change through the life course, and certainly changes from decade to decade.
What we heard, you know, especially in the morning, was that ethnicity and not race is an important concept to the states, and detailed ethnicity, particularly as it may impact upon health status and health outcomes.
We also heard about the difficulties that we have had in the past, and currently have, with the census construct of race and ethnicity, particularly for Hispanics.
We knew in the past that the construction didn't work. Clearly, census believed that having the whatever, are you Hispanic ethnicity question first was going to solve the problem, and clearly, it has not in any way solved the problem.
Hispanics think of themselves not only as being separate ethnic groups but also as being a separate race group.
State needs, one of the things we heard is that we have substantial differences in many of the states in our, if not unique ethnic populations, in the ethnic populations that we, in each of the states, need to focus upon.
While in one state Cape Verdians may be important and Chamoorans may be statistically irrelevant, in other states we have very different situations.
Similarly, we have substantial differences in the extent to which multi-race identification is germane in individual states.
In some states, obviously in Hawaii, it is extremely important. In other states, it is not important at all.
We have also heard about the need for additional relevant data, including certainly language, country of birth in addition to ethnicity, as well as the need to be able to drill down, not only to individual ethnic groups, but also to small area geographies, be those counties or neighborhoods or cities and towns and so forth.
Secondly, future directions. One area that we have heard about is the need for consistencies across the Department of Health and Human Service agencies and, within those agencies, in race ethnic data collection, bridging tabulation and reporting. This is clearly an absolutely crucial area.
Second, and related to that, is the need for clear and consistent operational federal guidance around race and ethnicity data collection, bridging and so forth, for public health. I am not going to editorialize on that right now.
Third, we heard about the need for federal support for state flexibility in race and ethnicity data collection, while still enabling states to meet the federal standards, flexibility around software, flexibility around item formats, flexibility presumably around code sets, especially for race and ethnicity, but then also, for economic position, language and country of origin.
Something else that we have heard from several of the presenters has to do with the need to focus on the uses of race and ethnicity data to communities themselves, as well as how those data can be usefully disseminated to local health care providers and to the communities.
Another issue revolves around the need for intercensal denominator data, as well as ensuring that our denominator data are consistent with our numerator data and, obviously, with race and ethnicity, in the next 10 years, that will continue to be a major issue and a major problem.
Another issue that was brought up is the need for training around race and ethnicity on data collection, particularly for those instruments in the data collection system where the data are not self reported.
There have been -- my personal favorite example for this is Australia, which has had a very good public information campaign around the collection of data on Aboriginal people and Torres Strait Islanders, with posters, instruction booklets, posters for hospitals and physicians, posters for data respondents themselves.
This is something that can be done. This is something that can be done and led at the federal level. Certainly, Census tried to do something similar, stressing the importance of answering race and ethnicity questions that were on posters, that were also in different languages for groups, and it is something, I think, that we can turn our attention to.
A final issue that I wanted to mention that was brought up was the need to address this kind of flexibility in a re-engineered vital system.
In terms of the role -- I think there were many issues that were brought up today that the committee, the national committee, could turn its attention to.
Certainly, one of those areas is training needs. Certainly, a second is to bring light to bear and, as it were, friendly pressure to bear on the need for consistency, across federal agencies and within federal agencies, and also the need to support geocoding of our data sets, and to support systematization, as Nancy Krieger said, of means to test that geocoding, increasing the sophistication of how we do it, and our self awareness about it, as well as our use of area based socioeconomic measures. With that, I am going to turn off my IPAC and go.
DR. MAYS: Thank you very much. I am sure we will continue the conversation at the next meeting, which is a couple of weeks away. I appreciate it, Dan.
Agenda Item: Summary and Next Steps.
DR. MAYS: Part of what I think has emerged today is that we focused a lot of time and energy previously at the federal data collection level, but that the state has a unique set of both desires of what it wants, as well as issues, and that the committee, I think, has been well served today, for the testimony, and we thank you for giving it to us to receive, so that we can work with it.
Our process at this point in time will be to, as a committee, sit down and determine what we think are the best directions for us to go in, what additional information we might need, and then try to pull together a report with a set of recommendations.
As I said, I don't think it is for us to just sit and think about recommending, but to think about how we can actually make movement occur as a function of that recommendation.
All of that is to say that this won't happen tomorrow. It won't even happen next week. Probably the joke on this committee is that I think these things should happen a lot faster than they seem to always happen.
I will be one to say that it is going to happen fast, but we will see. I know many of you have to catch trains and get out of here, so I just want to take the opportunity to thank you very much for spending the day with us.
I know that some of you put a lot of time and energy into your presentations. So, I appreciate the time you took away from your other day job and your family, et cetera, to do this, because we will receive it and take the mandate to try to do some good with it. So, thank you very much, everyone. I think we are going to adjourn at this point.
[Whereupon, at 4:28 p.m., the meeting was adjourned.]