The Public Ledger Building
Philadelphia, PA
The Subcommittee on Populations of the National Committee on Vital and Health Statistics was convened on November 8, 2002 at The Public Ledger Building in Philadelphia, Pennsylvania. The meeting was open to the public. Present:
Present:
Committee members
Staff
Others (including presenters)
The Subcommittee on Populations held a hearing in Philadelphia to talk with speakers about the issues for states and regional surveys in collecting data on racial and ethnic groups, to enhance their use in understanding and eliminating health disparities.
The meeting began with presentations by the following representatives of state departments of health:
The speakers addressed the following questions, which had been sent to them in advance:
Despite the variety among their states, the state presenters delivered a consistent message. It was summed up as follows by Dr. Cohen: States need detailed ethnicity data, not broad race data. They have unique populations they need to track. They need intracensal small area race denominators for race calculation and consistent guidance from all federal agencies regarding collection, use, conversion, bridging, and trend analysis. Instead, federal agency practices regarding race and ethnicity data are inconsistent.
One point that arose in the discussion of surveys was the emerging limitations of telephone surveys and the need for creative thinking on reaching hard-to-reach populations. In the discussion, as well, Ms. Harshberger encouraged the Subcommittee to support the Census Bureau's proposed American Community Survey, which would provide intercensal data.
Delton Atkinson, Project Director, NCHS Office of Vital Statistics
The re-engineering project is a joint effort of NCHS, the Social Security Administration, and the National Association of Public Health Statistics and Information Systems (NAPHSIS). The aim is to continue to have a decentralized system, but one that conforms to national standards. Mr. Atkinson's presentation focused on how the collection of race and ethnicity data in vital statistics will be implemented long-term, and what is needed to get to that point. Asked whether the re-engineered certificate will be flexible enough to support state needs, Mr. Atkinson encouraged the Subcommittee to make recommendations to the Department on that point.
Ms. Brett's project uses state-level data on women's health and is supported by the DHHS Office on Women's Health. The target audience is state and regional women's health coordinators. Data are available on CD ROM in an interactive format, enabling the user to compile tables and graphs.
Mr. Meltzer's project features county-level data and is a joint project of the Office on Women's Health and the Office of Minority Health, creating a joint focus on minority health issues and women's health issues. The target audience is public health professionals with operational responsibilities, but an effort is being made to make sure the text is meaningful to clinicians as well. Data are available on CD ROM.
Both systems are designed to be user-friendly. Both databases are set up to be consistent with and related to Healthy People 2010 data and objectives. Both projects comply with the OMB directive on race and ethnicity.
In response to a comment about the apparent redundancies between the two databases and the advisability of consolidating them into a single project, the presenters explained that the data contents (state vs. county) are quite different. Dr. Krieger commented on the importance of framing women's health in terms of a population health approach that takes factors such as context and indicators of women's social status into consideration.
Harvard School of Public Health Nancy Krieger, Ph.D., Assoc. Professor, Department of Health and Social Behavior,
Dr. Krieger said her Public Health Disparities Geocoding Project is about improving monitoring of social inequalities in health in the U.S. It was conducted in partnership with the Massachusetts Department of Public Health and the Rhode Island Department of Health (RIDH). A premise of the project is that collaborations between universities and health departments are vital to improving the monitoring of social inequalities in health.
The project was undertaken because most U.S. public health surveillance systems contain no socioeconomic data. Data are typically racialized in an attempt to get at health disparities, and there is no ability to assess socioeconomic gradients and health and their contribution to disparities within diverse race and ethnicity groups.
A solution is to geocode health data and link these and population data, stratified by Census-derived area-based socioeconomic measures (ABSMs). The purpose of Dr. Krieger's investigation, which looked at seven health outcomes, was to determine which ABSMs at which level of geography would be most useful for public health monitoring. The goal was to develop valid, robust, and easy to construct and interpret ABSMs for use by state health departments for any health outcomes and for all demographic groups.
The study, which looked at nearly 1 million records, found overall that persons in the worst-off areas had mortality rates 1.3 to 1.4 times higher than the persons in the best-off areas. Yet despite the evidence of significant socioeconomic gradients, current U.S. public health reports, "routinely ignore" all these gradients. This type of analysis could reveal the contribution of socioeconomic inequalities within racial, ethnic and gender groups. The study found that the most sensitive ABSMs across all outcomes were those measuring economic deprivation, and the best single candidate is the percent below poverty at the Census tract level. Dr. Krieger concluded that for monitoring social inequalities in health, both race/ethnicity and class matter. She recommended further work to assure the consistency of coding of race and ethnicity across public health databases, and routine geocoding in all public health databases using standardized ABSMs in order to routinely assess the contribution of socioeconomic inequality to racial and ethnic disparities in health.
A final document summarizing methods and key findings will be prepared in the spring of 2003 and sent to all U.S. health departments.
In the discussion period, Dr. Krieger cautioned people to check on accuracy before selecting a geocoding firm, and then to regard its work as an assay to be tested.
Dr. Mays observed that it would be useful for the Subcommittee to stay in touch with NAPHSIS as it develops its recommendations and considers how to have an impact.
At her request, Dr. Friedman summarized the day's discussion in terms of three themes: commonalities and differences among the represented states; future directions; and potential roles for NCVHS. He identified the following main points in the presentations and discussions:
Dr. Mays said that once the Subcommittee on Populations has finished gathering the information it needs, it would not only prepare a report and recommendations but also think about how to "make movement occur as a function of the recommendations."
Dr. Mays called the meeting to order, saying its purpose was to talk about the issues for states and regional surveys in the collection of data on racial and ethnic groups for the purpose of understanding and eliminating health disparities. The Subcommittee advises HHS on the collection of data on racial and ethnic groups as it applies to health. It has concerns in the wake of the OMB-15 directive as to how and whether data are being collected and, if so, how useful they are. This is part of a series of hearings and meetings on those topics. She read aloud the questions that had been sent to all speakers. (See Executive Summary)
Dr. Mays then introduced Dr. Dalton Paxman, Regional Health Administrator for HHS Region III. Dr. Paxman welcomed the group on behalf of the Assistant Secretary for Health. He noted the importance of this meeting in helping HHS achieve its goal of reducing and eliminating health disparities. He cited the wide racial disparities in childhood asthma as an example of the need for data on racial and ethnic groups for a complete picture of population health and the appropriate interventions.
Dr. Mays thanked Dr. Paxman and his staff for their hospitality, and introduced the first presenter.
Note: The transcript containing the details of these presentations is on the NCVHS Web site http://www.ncvhs.hhs.gov/021108tr.htm
Bruce Cohen, Ph.D., Director, Division of Research and Epidemiology
Dr. Cohen began with comments on the differences in state and federal needs for race and ethnicity data, differences that are key to understanding the state perspective on the issue. He defines ethnicity as national origin, ancestry and heritage, and prefers the term "ethnicity group." For states, "ethnicity may be more relevant than the concept of race." At the state level, the five categories in the new OMB standard are not sufficient to capture the heterogeneity of the population. He cited the case of Cape Verdians, who are classified merely as blacks by the Census but only 25 percent of whom identify themselves as black. Also, having the Hispanic ethnicity question precede the race question did not work. In addition, the check-box categories in the revised standard will have unintended negative consequences for state and local public health because they force departments to collect less relevant data.
States need detailed ethnicity data, not broad race data. They have unique populations they need to track. They need intracensal small area race denominators for race calculation, and consistent guidance from all federal agencies regarding collection, use, conversion, bridging, and trend analysis. Instead, federal agency practices regarding race and ethnicity data are inconsistent.
State needs are different from federal ones because they use fewer face-to-face situations with trained interviewers to collect data; they are dealing with smaller areas; and they have different approaches to bridging.
Dr. Cohen recommended that states receive uniform guidance from federal agencies and that states develop partnerships with federal agencies to address technical issues around bridging and small area estimation. Finally, states should emphasize ethnicity rather than race, and they should focus on self-report.
He then described the Massachusetts Department of Public Health experience. He stressed that the fundamental data collection questions that all states need to answer are how they will use the OMB and Census categories and how they will adopt the NCHS recommendation for their birth and death certificates. He described the approaches his state is considering. He added that staff is also interested in knowing language preference, country of origin, and something on migration patterns.
Although states have not received adequate federal guidance on bridging, Massachusetts has implemented its own strategy, which he described. On tabulating the data once collected, he stressed the need for a unified federal statement supporting the OMB recommendation.
He then discussed Massachusetts' use of race and ethnicity data in its publications and reports, calling particular attention to its online information service, Masschip. The Health Department will continue to focus on a detailed array of ethnicity groups, and it will standardize its approach to be consistent with OMB guidelines and develop its approach for implementing vital statistics certificates.
In conclusion, Dr. Cohen stressed the need to consider how race and ethnicity data are used, as well as the need to link the data to SES, linguistic, and other demographic indicators. He pointed out that OMB standards were developed to enforce civil rights laws, and there may be better approaches for collecting race and ethnicity data for health surveillance activities and for targeting interventions to end health disparities.
Peter Abbott, M.D., Acting Director, Health Information and Strategic Planning
Dr. Abbott began by underscoring the points made by Dr. Cohen. California, too, is very heterogeneous and no group is in the majority. He focused his presentation on the California Health Interview Survey (CHIS), first done in 2001 after three years of planning by multiple organizations. It received federal, state, and foundation funding.
The survey was designed to collect county-level data and data on minority subpopulations. It oversampled Alaska Native and American Indians and other subpopulations. It included interviews with adults, adolescents, and the most knowledgeable adult on behalf of children in over 55,000 households. The findings reflect the state's diversity. An attempt was made to be consistent with the U.S. Census and OMB guidelines. He noted that most Latinos identified their race as Other. He described California's decision rules for constructing a single race and ethnicity variable, based on the race with which the respondent most identifies. Information was also asked about tribal identification. The CHIS was administered in six languages.
There has been a major effort to disseminate CHIS data, through public use and confidential files and various reports. Policy briefs and focused reports are planned on food insecurity, asthma, and health insurance coverage, among other topics. Detailed information is available on the Web at AskCHIS.edu. The data access center is at the UCLA Center for Health Policy Research. Another cycle of the CHIS, of the same magnitude, is planned for 2003. The State Department of Health Services' Center for Health Statistics is holding workshops for researchers and other users, as well as invitational conferences for various racial and ethnic groups. Many states are using the database and California hopes to do small area estimates. In conclusion, Dr. Abbott stressed the heterogeneity within the major OMB classifications in his state (Latinos/Hispanics and Asians/Pacific Islanders).
Dr. Mays said race and ethnicity data collection and report methodology are special interests of Ms. McKendry. Ms. McKendry (participating by telephone) began her presentation with comments on California's data needs stemming from its demographics and the fact that, as noted, it does not have a racial or ethnic majority. She discussed the differences within large ethnic groupings. Hispanics in California are primarily of Mexican origin, but there are significant differences within this group. For example, there are big differences in health outcomes and culture between Mexicans born in Mexico and those who were not. California does not have denominator data on this variable and public health people are unable to compare rates for these two categories. Asian subgroups are also very diverse, again with an important variable being how long people have been in the U.S. in addition to country of origin. Currently, 2.7 percent of Californians identify as multiracial, and that is expected to increase markedly, as is the diversity of the state in general.
Ms. McKendry did a survey that revealed considerable variety in the ways different programs collect health data in California. In general, they report it in collapsed categories. Some programs collect place of birth or origin and principal language. She heard complaints about the different standards and practices in different branches of the federal government. Some state programs are more concerned about HIPAA than conforming to OMB standards. Another problem for the state is that they are waiting for the release of population data from the Department of Finance, the gold standard for the state, but counties need the data sooner than the state can provide it. For example, Sacramento does not know the size of the large Russian immigrant community that needs to be served there. She summarized, "Detail, detail, detail is what California is all about, both racially and geographically."
The state also needs standards and guidance. A workgroup was formed through UCSF's Family Health Outcomes Project (FHOP) to address the need for standards, with various county and state participants. They are revising state guidelines developed in 1998 to standardize state race and ethnicity data. They used the U.S. Census standards and did not address problems of multiple federal standards, and they insisted that state data could be collapsed into minimum OMB standards. The state's Center for Health Statistics was instrumental in this process, which also tried to acknowledge programs' practical constraints. They were the first part of the Health Department to reflect FHOP standards, and implemented "a semi-Census model of data collection," which she described. When possible, they use self-reporting for race and ethnicity identification, and they permit reporting of up to three races. They use text boxes on birth and death certificates, partly to keep the former to a single page. Mexicans are allocated to White ¾ the biggest difference from the Census. This format will be retained on the revised certificates. The death certificate revision is completed and implemented in January 1, 2003. The electronic death registration system is expected to be implemented in January 2005, as well. The FHOP guidelines will have considerable detail on the birth system's method of collecting and coding race, which she described as "quite elegant" and flexible.
Ms. McKendry then showed slides using California's current race and ethnicity data in various contexts. She noted that "a real big problem" is the current small sample size of the two-plus races category in mortality rates, due to what appears to be under-reporting. They are working with funeral directors to improve this.
Summarizing her remarks, she said tabulation issues include balancing the detail programs need with sample size quality, denominator quality, and simplicity. There have been several denominator problems, and they have to decide whether to use first-listed race as a bridging method. She emphasized that more race and ethnic detail than the OMB minimum standards is crucial for California in order to see the differences among subgroups, and that they need consistent and appropriate standards from the federal government.
Alvin Onaka, Ph.D., Chief, Office of Health Status Monitoring
Dr. Onaka began with an overview of race and ethnicity data collection in Hawaii, 58 percent of whose 1.2 million residents is Asian. The geographic distribution of racial and ethnic groups among the seven inhabited islands varies a great deal. Dr. Onaka agreed with previous speakers "the concept of ethnicity, which includes culture, language and immigrant status, is a better indication of what we are trying to measure [than race]." Migration status is an important element, along with language (including dialects). A non-English language is primary for 27 percent of Hawaiians.
Hawaii has been collecting multi-race data for more than a century. It had to aggregate its data to meet NCHS requirements. It has a long-standing program to assess the quality of its data, partly using an annual health and demographic survey. This survey shows growth in the proportion of the population with multiple races. 21.4 percent of state residents were of two or more races in the 2000 census, and more than half of the children in 2001 birth data indicate two or more races of the parents. Hawaii's health and demographic survey is similar to California's. One aspect is looking at whether races are being misclassified. The 2000 Census confirmed its count of American Indians and its estimates of multiple races. Race and ethnic data are collected on all state government surveys as well as in vital statistics, as well as in Medicaid managed care and hospital discharge data. A problem with the latter two is that race and ethnicity are often ascribed rather than self-identified. The Hawaii Health Survey is modeled after the National Health Interview Survey. They started using a telephone survey in 1996, calling 5,000 households. Dr. Onaka discussed the variety within various OMB categories and the need for more detail with reference to Hawaiian sub-populations. The state disseminates most of its information electronically.
A major methodological challenge concerns the comparability of the race numerator to race denominators. They hope to get denominators through their annual health interview survey. To continue to collect diversity, they will use a modified Census/OMB collection instrument, but with a text box where people can self-identify. Dr. Onaka said the state also needs small area information and information for intercensal periods.
Discussion
Dr. Newacheck asked the presenters how they adjust for presumed under-representation of racial and ethnic minorities in their telephone surveys, and whether they know of any survey mode effects on reporting of race and ethnicity. Dr. Abbott said CHIS data are adjusted for under-representation, adding that having the language of the respondent available has proven to be critical. Dr. Onaka said Hawaii has determined how many households do not have telephones and adjusts accordingly. For language, they do a callback. They have not found a difference between in-person and telephone interviews. Dr. Mays wondered who, in particular, is most likely to be missed by telephone surveys. Dr. Cohen said Massachusetts's response rates on telephone surveys have declined markedly. He said this discussion points to the emerging limitations of telephone surveys and the need for creative thinking on reaching hard-to-reach populations. He speculated that people who are geographically and linguistically isolated, the homeless and migrants, are being missed. Dr. Abbott added that with screening devices and cell phones, the higher end of the economic spectrum and younger people are also being missed; and the elderly tend not to respond. He said the CHIS sends letters ahead of the survey in six languages. Dr. Onaka observed that given the diffidence of many Asians, having relative anonymity on a telephone call may increase their willingness to provide information. All presenters agreed that alternative ways of collecting information must be found, and all noted disappointing response rates. On underreporting in vital statistics, Ms. McKendry said Native Americans are underreported in death data, as are Asians and Hispanics, the latter because they "go home to die."
Ms. Breen asked the states' experience collecting information on tribal detail, SES, sexual orientation, geography, disability, access to health services, and health insurance, receiving varied responses. Dr. Cohen said Massachusetts got the best results on tribal detail working with tribal groups to collect the information.
From the audience, Stephen Thomas, Director of the Center for Minority Health at the University of Pittsburgh, spoke on behalf of the Society for Public Health Education. He pointed out that the concept of majority and minority with respect to race and ethnicity is a proxy for who has the power, and is not just about numerical proportions. He asked what the Subcommittee and panelists saw as the purpose of the costly effort to improve the precision of measurement.
Dr. Abbott noted the distinction between producing the data and making sure they are used to address health issues, including disparities. He said the CHIS are trying to get at the issue of discrimination in health services, to help guide health policy and programs. Dr. Mays said that the Subcommittee is trying to assess whether the Department has enough data of the right kind to meet its targets in reducing and eliminating health disparities. It hopes to do so before Healthy People 2010 reaches its midpoint review. The information can then be used by a large body of user groups to talk about health disparities. She added that states are asking the same questions as Dr. Thomas. Dr. Friedman pointed out that most state health departments actually do public health practice, and ethnic-group-specific data are used to target and evaluate programs as well as to increase understanding of the variations and health disparities among groups. The thrust of the current investigations is that much greater detail is needed to accomplish that goal.
Ms. McNeil of the HHS Office for Civil Rights asked about the impact on accuracy of the fact that race and ethnicity are typically not self-identified in encounter data. Dr. Cohen agreed, and stressed that the federal government needs to provide more resources for training those who collect the data.
Delton Atkinson, Project Director, NCHS Office of Vital Statistics
The partnership to reengineer the nation's vital statistics includes NCHS, the Social Security Administration, and the National Association of Public Health Statistics and Information Systems (NAPHSIS). The aim is to continue to have a decentralized system, but one that conforms to national standards. Mr. Atkinson noted that "part one" of this presentation, concerning race and ethnicity data in vital statistics, had been delivered by Dr. Jim Weed of NCHS at the Subcommittee's September 25 meeting. Today's presentation focuses on how the collection of race and ethnicity data in vital statistics will be implemented long-term, and what is needed to get to that point.
A survey of states found that changing information systems just enough to collect race and ethnicity information would cost $10 million nationwide. Most states have old information systems supporting vital statistics, and vendors would charge a great deal to change the systems, making even minor changes to these outdated systems arduous and costly. After much internal debate, the vital statistics community concluded that re-engineering business practices, applying new technology, and working collaboratively at state and national level is the answer. Among other things, the partnership will assist states in evaluating vendors' base-model products against a prototype model, and it will develop better ways to integrate the vital statistics system with other data systems. The aim is to minimize person-intensive processes in vital statistics. Also, NCHS is working with NAPHSIS to establish a national standard for the electronic exchange of vital certificate information. The partnership will work with states to find money to re-engineer their systems, to ensure that all 57 states and territories can do so.
These changes will require public health in every state to make re-engineering a priority, and federal agencies other than NCHS and SSA to become active partners. The users of vital records and vital statistics will also have to become involved. The central theme of this national policy is coordination, collaboration, and a comprehensive vision.
Discussion
Dr. Friedman noted the lack of congruence, highlighted in the foregoing discussion, between the model certificate on race and ethnicity and state-specific needs. He asked whether the re-engineered certificate will be flexible enough to support state needs. Mr. Atkinson encouraged the Subcommittee to make recommendations to the Department on that point. He noted that both policy and technology issues are involved.
To another question, he said the re-engineering project is assessing state readiness to implement the new systems and what resources are needed. It is estimated that over $100 million would be needed to institute electronic death and birth systems.
Kate Brett, Ph.D., Office of Analysis, Epidemiology and Health Promotion, NCHS
Dr. Mays said the next sessions would focus on research uses of vital statistics data. Dr. Brett said her project uses state-level data on women's health and is supported by the DHHS Office on Women's Health. The target audience is state and regional women's health coordinators. It has data on men, as well. The project allows the viewing of previously published data by specific population groups within states, with information by sex, race and Hispanic origin, and age. The data are released as electronic interactive data tables, available on CD ROM. Users can format the tables. The focus is measuring Healthy People 2010 objectives. The focus is not trends but current information. Data sources include BRFSS data, mortality data; the youth risk behavior surveillance system, and AIDS incidence. The data are presented as three-year moving averages, to increase the reliability for estimates with infrequent events or small sample sizes and to decrease confidentiality issues. Suppression criteria suppress estimates with low precision.
The three-year collection means that three years of data are needed to present the new race and ethnicity categories. They will use the OMB standards once the data are available for three years and the data system collects the data appropriately. They are following what the Division of Vital Statistics does with their data until mortality data are in the new format. On the known race and ethnicity classification problems in mortality data, Ms. Brett said it is not in her project's purview to change the data; however, they note problems in their technical notes. She described technical aspects of the project involving its data dissemination software, Beyond 2020, and then showed examples of the Healthy Women system's capabilities. She noted that it is very user-friendly, which is "the whole point." Users can create tables, graphs, and maps. It is not possible to sort by risk factor. The data can be found on the NCHS home page in a section called "Tabulating State Data." Having now determined that the system works, NCHS is now figuring out how to promote it.
Alfred Meltzer, Quality Resource Systems, Inc.
This project started recently on a national scale. It is the outgrowth of a regional project of the Office of Women's Health in Denver (Region 8) to provide six states with reports on useful data, particularly county-level or below. The Region 8 project generated interest in the database, which was spun off as a separate entity. In response to interest from other states, the central Office of Women's Health decided to make it a national program. The Office of Minority Health is a partner in the project, creating a joint focus on minority health issues and women's health issues. The ultimate target audience is public health professionals with operational responsibilities, not necessarily clinicians or researchers, but an effort is being made to make sure the text is meaningful to clinicians as well. The database includes user-friendly software that enables access to "a large data base" with little training or experience with database systems. Initially, the database will be loaded on a CD-Rom with self-installing capabilities. The CD-ROM will contain hard links via the Internet to principal sources. The project is being phased in region by region over the next 12 to 14 months, and the resulting database will eventually be placed on the Internet. The project is now in its fourth region. To supplement national sources, regional people are asked to help identify state and local data resources. On-site training lasting 7 hours is provided.
Turning briefly to the details, Mr. Meltzer said the data base includes extensive Census data and data on SES, health resources, infectious and chronic diseases, mortality, natality, violence and abuse, mental health, and prevention. Data issues include the existence of many cells with very small values. State agencies' suppression issues override the presentation of data. Also, there are consistency issues among sources, not only across states but also within states.
Discussion
In response to a question, Mr. Meltzer said his project is keeping track of data gaps as they are identified and will provide the list to OWH.
Both Mr. Meltzer and Ms. Brett said their databases are set up to be consistent with and related to Healthy People 2010 data and objectives. Regarding race and ethnicity, both presenters said their projects would comply with the OMB directive. Mr. Meltzer said his database has a fixed format and would reformulate more detailed state ethnicity data into the OMB categories.
Ms. Breen remarked on the apparent overlaps between the two databases and said someone should decide why there is not a single project that could be built on as needed. Both presenters explained that the data are quite different: Mr. Meltzer's data are county-level, and will contain the range of topics discussed above, while Ms. Brett's system uses state-level data.
Asked about variations related to differing county needs, Mr. Meltzer said regions are allowed to vary the basic model to accommodate local needs. The database will also necessarily be limited by any limitations in state and county practices, e.g., in their data on race and ethnicity. The data on infectious and chronic diseases, fetal deaths, and abortions are being collected from state compilations of local and county reporting. Ms. McNeil suggested that the state centers for health statistics might be helpful in accessing data.
Dr. Krieger commented on the importance of framing women's health in terms of a population health approach that takes factors such as context and indicators of women's social status into consideration. She noted that while clinicians had reportedly helped ensure that these databases had clinical meaning, it appeared that social scientists were not consulted to give meaning to the social nature of the variables. In particular, she expressed surprise that work experience and environmental exposures are not included. Dr. Brett responded that her database's framework is Healthy People 2010, and it is limited to available state data. Mr. Meltzer said one region's women's health coordinators and minority health coordinators drove his project's database content with input from DHHS' central office.
Nancy Krieger, Ph.D., Assoc. Professor, Department of Health and Social Behavior, Harvard School of Public Health
Dr. Mays said the Subcommittee was interested in both the content and the process of Dr. Krieger's work, the latter because of her experience working with state departments of health. She invited comments on opportunities, through collaborations between academic institutions and state offices, to produce data sets that are larger than would otherwise be possible.
Dr. Krieger said her Public Health Disparities Geocoding Project is about improving monitoring of social inequalities in health in the U.S. It was conducted in partnership with the Massachusetts Department of Public Health and the Rhode Island Department of Health (RIDH). She acknowledges the contributions and support of NCVHS member Dr. Dan Friedman, of the Massachusetts Department of Public Health. A premise of the project is that collaborations between universities and health departments are vital to improving the monitoring of social inequalities in health. Healths departments have the data and know the research is needed, but lack the time and resources to do the work; university-based researchers are well placed to get grants and have the expertise to conduct the research, but need access to the data.
Dr. Krieger's team obtained an NIH RO1 grant to conduct the current study with Massachusetts and Rhode Island data. A final document summarizing methods and key findings will be prepared in the spring of 2003 and sent to all U.S. health departments.
The project was undertaken because most U.S. public health surveillance systems contain no socioeconomic data. Data are typically racialized in an attempt to get at health disparities, and there is no ability to assess socioeconomic gradients and health and their contribution to disparities within diverse race and ethnicity groups. Survey data are not sufficient for monitoring local and national health for all populations. A solution is to geocode health data and link these and population data, stratified by Census-derived area-based socioeconomic measures (ABSMs). A problem is the lack of consensus on what level of geography and which ABSMs should be used.
The purpose of Dr. Krieger's investigation was therefore to determine which ABSMs at which level of geography would be most useful for public health monitoring. The goal is to develop valid, robust, and easy to construct and interpret ABSMs for use by state health departments for any health outcomes and for all demographic groups. Ecosocial theory suggests that different ABSMs might function differently for diverse outcomes. She listed the seven outcomes of interest in this study: low birth weight, childhood lead poisoning, sexually transmitted infections, tuberculosis, non-fatal weapons-related injury, cancer incidence, and mortality. She then briefly described their methodology for establishing the study base, creating ABSMs, geocoding health data, linking the records to ABSMs, and generating rates stratified by each ABSM at each level of geography. They worked with roughly one million records. Socioeconomic variables related to occupational class, income, poverty, wealth, education, crowding, and combinations thereof.
The investigators assessed the ABSM and their cut points by visiting randomly selected addresses in Boston and comparing their impressions to the ABSMs' characterization of the area. Before choosing a geocoding firm, they studied the accuracy of several candidates (which ranged from 44 to 96 percent) and chose the one with the highest accuracy. Dr. Krieger noted that the technical issues they encountered included problems with coding race and ethnicity because of inconsistencies in health databases.
The study found, overall, that persons in the worst-off areas had mortality rates 1.3 to 1.4 times higher than the persons in the best-off areas. She showed tables presenting various dimensions of this finding. To consolidate their findings, they devised a scaled RII (relative index of inequality) plot for each ABSM, across outcomes. For each of the outcomes, most of the ABSMs suggest that the impact of socioeconomic position on health outcomes is quite robust. She noted differences among the various ABSMs with respect to different outcomes and said the poverty measure worked well in detecting socioeconomic gradients.
She then showed examples of what U.S. public health data would look like if they were routinely stratified by "an apt ABSM," looking at all-cause mortality and then at several specific conditions such as blood lead levels and low birth weight. Despite the evidence of significant socioeconomic gradients, current U.S. public health reports, "routinely ignore" all these gradients. This type of analysis could reveal the contribution of socioeconomic inequalities within racial, ethnic and gender groups. Dr. Krieger showed data illustrating this point.
Summing up, she said that their data suggest 1) that it is feasible and informative to monitor U.S. socioeconomic inequalities in health using ABSMs, and 2) that the choice of ABSM and the level of geography both matter. She stressed that the error and bias inherent in the study would result in an underestimate of the socioeconomic gradients, because it misses the worst-off part of the population. There are no comparable studies to permit a comparison to the literature. The study found that census tract and block group-level data behave similarly, and zip code-level data were more problematic (and will become moot for this purpose in the future because of changes in the system). The most sensitive ABSMs across all outcomes were those measuring economic deprivation, and the best single candidate is the "percent of persons below poverty" at the Census tract level. Dr. Krieger concluded that for monitoring social inequalities in health, both race/ethnicity and class matter. She recommended further work to assure the consistency of coding of race and ethnicity across public health databases, and routine geocoding in all public health databases using standardized ABSMs in order to routinely assess the contribution of socioeconomic inequality to racial and ethnic disparities in health.
Discussion
Dr. Friedman said the gradient charts shown by Dr. Krieger would be very useful at the state level. He asked her opinion about encouraging states to add measures of individual economic position to surveillance data sets. Dr. Krieger said data collection depends on clinicians asking routine questions in medical histories about people's social and economic experiences. She noted the problems with quality of educational data as a proxy. She added that she was not in a position to know what would be required to get states to collect such data.
To another question, she said her team's next goal is to determine the generalizability of the Massachusetts and Rhode Island data. They have applied for a grant to look at mortality data for multiple outcomes in one state in each federal region, including rural areas.
Mr. Hitchcock reported that the HHS Data Council is about to be briefed on geospatial coding, and there is a White House Executive Order and an OMB directive encouraging all departments to make better use of geocoding. Dr. Krieger said she's observed that epidemiologists "are a little naïve about geocoding," and she cautioned people to check the accuracy before selecting a firm, with the same attention to rigorous methodology as when sending out an assay to be tested. There is a learning curve to doing it correctly, and there must be attention to the rigor of the methodology. She also recommended using geography plus ABSMs, thus making it possible to release data without violating confidentiality.
Dr. Subramanian, a member of the Geocoding Project, explained the conceptual differences between individual socioeconomic position and the ABSM, and said that they have found that the context in which people live, itself, has an impact on their health.
To another question, Dr. Krieger said that it is an empirical question whether it is possible to validly use one measure across multiple racial and ethnic groups. Early evidence is that economic deprivation variables hold the most meaning across race and ethnicity groups. She expects to be able to give a more detailed answer in six months. She stressed that analysis within groups, not just between groups, is critical.
Dorothy Harshberger, State Registrar
Ms. Harshberger referred people to www.adph.org, the department's Web site, which shows the Census data. Alabama has 4.5 million people, of whom 70 percent are white and 26 percent are black. One percent of the state marked two or more races in the Census, so the bridging issues are minimal. The Hispanic population, while only 1.7 percent of the total, is growing. The state has data on national origins for Hispanics.
She showed the data compiled for a chart book on maternal and child health, which includes race and ethnicity data. Data on disparities have been presented for years, and Alabama is presenting more data on subpopulation groups than before. Ms. Harshberger distributed the book, Alabama Book of Racial Disparities in Mortality. She noted that it is difficult to get down to the county level because some counties have such small populations. However, they did produce an atlas with county data, combining three years' -worth.
One serious data issue is the lack of denominator data in non-census years. As a result, for example, while they know the Hispanic population is growing, they do not know its growth rate. The Census Bureau's proposed American Community Survey would provide data for the intervening years, and she asked the Subcommittee to support that survey. Another data issue is that historically, state data have combined the Black and Other categories. They are trying to change that. Survey samples are another issue, especially in representing migrants.
The differences in reporting requirements across programs and data sets is a problem¾for example, the definition of regions in different program contexts. Urban/rural definitions vary as well. Alabama has the same difficulties with race and ethnicity categories as described by other presenters. Ms. Harshberger noted, though, that while standardized data collection would be nice, it is a problem for states if they are expected to collect data on ethnic groups for which there are only very small numbers. In general, "it would be nice if there would be some way to get the data that the states need."
Alabama is "one of the poor states," she said. They use a DOS system for collecting birth data and are facing large costs to change the system. They want to do so, but the money is not available. She stressed that everybody wants good data, but this is expensive.
Richard Urbano, Ph.D., Assistant Commissioner, Bureau of Health Information
Dr. Urbano said the short version of his presentation is "Add a million people to Alabama, and you have Tennessee." His state population is 83 percent white and 16.3 percent black, with large differences in distribution among regions (as in Alabama). It also uses a DOS-based system for birth data but is moving toward Windows-based reporting. He listed some of the state data sets that report race and ethnicity, and said they will expand the recording of race and ethnicity data although the numbers in the latter category are very small. Tennessee posts its reports on its Web site rather than printing them.
Tennessee is a waiver state and Tenncare covers about half its children. It has a large database of encounter data for Tenncare. Case managers in five agencies use the data in a database called Children's Information, Tennessee for public health and health services analysis. Access to the information is controlled by parent or guardian permission. The program is being expanded to include private practice pediatricians. There are now 1.8 million children in the system. Data collection is consistent across all state programs. Unique identifiers are used to link data across programs.
Discussion
Dr. Mays asked about the states' abilities to identify the national origins of black state residents. Ms. Harshberger said Alabama can identify national origin, although in general the state has a very low in-migration rate (also the case for Tennessee). She said it would be very useful to have SES data, particularly income.
Dr. Abbott called attention to a Racial Privacy Initiative on which Californians will vote in June 2003. If successful, it would prohibit the collection of race and ethnicity data by most state governmental entities, not including vital records, medical research and programs with federal requirements.
In response to a question from Dr. Mays, Ms. Harshberger described the barriers in Alabama to adding new variables to data collection instruments to improve understanding of diversity and disparities. The cost of changing an automated system is the greatest problem, and there are political issues at every level.
Dr. Krieger pointed out that the acceptability of asking "certain questions" is culturally determined and can be changed with adequate leadership. She noted that the United Kingdom has a long history of collecting social class data, but until recently they thought race and ethnicity was irrelevant and impossible to collect. But when people demanded these data, they found a way to ask the questions. The leadership has to be national, with a feedback process.
Dr. Abbott agreed that national leadership and standards are important, and he added federal funding to the list. Dr. Lengerich noted states' desire to set their own direction and the need for an intermediary group to speak collectively for the states. Dr. Onaka, the current president of NAPHSIS, described how it plays that role. Dr. Mays observed that it would be useful for the Subcommittee to stay in touch with NAPHSIS as it develops its recommendations and considers how to have an impact.
She then introduced Dr. Friedman, whom she had asked to comment on the day's discussions "because he lives and breathes state issues."
Daniel Friedman, Ph.D.
Dr. Friedman commented on three themes: commonalities and differences among the represented states; future directions; and potential roles for NCVHS.
A consistent theme through the presentations was the need to think "well beyond race" because it is not, by itself, a useful concept to the states. Ethnicity is the important concept, with detail, especially regarding its impact on health status and health outcomes. In this regard, many presenters had discussed the difficulties with the Census construct of race and ethnicity, particularly for Hispanics, who think of themselves as a separate race group. There are substantial differences among the states in the ethnic populations they need to focus on, as well as in the significance of the multi-race identification. Finally, there is a need for additional relevant data with respect to ethnicity, including language and country of birth as well as on small areas.
In terms of future directions, Dr. Friedman noted the crucial need for consistency across HHS agencies in race and ethnicity data collection, bridging tabulation and reporting. There is also a need for clear operational guidance from the federal government around race and ethnicity data collection, bridging, and so on. In addition, there is a need for federal support for state flexibility in meeting federal standards in the software, formats, code sets, etc, they use to collect data on race and ethnicity. And states need to be able to focus the uses of the data on communities themselves, including local health care providers.
The presenters also noted the need for intercensal denominator data, and for ensuring these data are consistent with numerator data. They also stressed the need for training around race and ethnicity data collection. This can and must be done at the federal level. Finally, it is necessary to address this kind of flexibility in the re-engineered vital statistics system.
Concerning the role of NCVHS, Dr. Friedman suggested that it highlight the need for training and bring "friendly pressure" to bear on the need for consistency across and within federal agencies, as well as the need to support geocoding of data sets and the use of area based socioeconomic measures.
Dr. Mays said this meeting had been useful in broadening the Subcommittee's perspective, from a focus on federal data collection to one that includes state needs. The Subcommittee will now identify what information it still needs and then pull together a report and recommendations. It will be important to think about how the Subcommittee can "make movement occur as a function of the recommendations." She thanked the presenters for their contributions.
I hereby certify that, to the best of my knowledge, the foregoing summary of minutes is accurate and complete.
Vickie Mays, Ph.D. /S/ 12/30/04
Chair Date