Midcourse Review > Table of Contents > Appendix C: Technical Appendix
Appendix C provides additional information on a number of issues related to monitoring progress in Healthy People 2010:
Measuring Progress Toward Target Attainment
Progress toward the target for each objective in Healthy People 2010 is measured using the progress quotient, or percent of targeted change achieved. The progress quotient expresses any change from the baseline relative to the initial difference between the baseline and the target. The progress quotient also was used to monitor progress in Healthy People 2000.1, 2
Baseline data values were published at the beginning of the decade for Healthy People 2010 objectives and subobjectives for which data were available.3 Baseline data for additional objectives and subobjectives have become available since the publication of Healthy People 2010.4 Progress Quotient Chart PQ = (most recent value – baseline value) / (year 2010 target – baseline value) × 100
For example, school-based objective 7-2c calls for an increase in the proportion of middle, junior high, and senior high schools that provide education to prevent violence from a baseline of 58 percent in 1994 to a target of 80 percent in 2010. In 2000, 73 percent of schools provided education to prevent violence. With the use of the formula above, 68 percent of the difference between the baseline and the 2010 target had been achieved in 2000. PQ = (73 – 58) / (80 – 58) × 100 = 68 percent For the population-based objectives, the PQ also can be used to measure progress toward the target for each population group with data beyond the baseline. For example, the PQ for objective 16-1c to reduce infant death to 4.5 infant deaths per 1,000 live births in 2010 can be computed for the total population between the 1998 baseline (7.2) and 2002 (7.0). When the formula above is applied, 7 percent of the difference between the baseline and the 2010 target had been achieved for the total population in 2002. PQ = (7.0 – 7.2) / (4.5 – 7.2) × 100 = 7 percent In contrast, among infants of Asian women, the infant death rate declined from 5.0 at the baseline in 1998 to 4.1 in 2002. With the formula above, 180 percent of the difference between the baseline and the year 2010 target had been achieved in 2002. PQ = (4.1 – 5.0) / (4.5 – 5.0) × 100 = 180 percent In this example, the PQ indicates that the target was exceeded by 80 percent of the difference between the baseline and the target. The PQ is positive when the rate moved toward the target and negative when the rate moved away from the target. The PQ can be used to compare progress for one objective, relative to its baseline, with progress for other objectives, relative to their baselines. There are some limitations to the interpretation of the PQ statistic. First, the PQ measures the observed difference between the baseline year and the most recent year only. Fluctuations in the measure during the intervening years are not considered. This variability can cause substantial fluctuations in the size of the PQ from year to year. Second, the number of years between the baseline year and the final data year for Healthy People 2010 might be different both between objectives and within objectives. Between objectives, differences in the number of years available to meet targets are a function of data sources and choices made regarding the most appropriate baseline year for a particular objective. To assist the reader in the interpretation of these comparisons, the baseline data year and the most recent data year for each objective are shown in parentheses in the PQ charts for each focus area. Within objectives, differences in the number of years available to meet targets for specific groups within the population template can be affected by changes in the classification of race during the tracking period. In these cases, the period used to compute the PQ is the same as that used to measure disparities. (See the section on Measuring Health Disparities for more details.) Third, the absolute change required to attain the target might differ among select populations or across objectives and/or subobjectives with identical PQs. Therefore, equal PQs do not reflect equal absolute progress from the baseline. Fourth, estimates of the variability of the PQs are not included in the midcourse review because of lack of standard errors of the estimates for some objectives. Estimates of variability for the PQs are expected to be available for the final review at the end of the decade. In addition to the above limitations, there are a small number of cases in which the PQ could not be calculated or did not accurately reflect change in an objective. Four example cases are as follows:
In the figures for each focus area, objectives like examples 1 and 3 are shown with arrows in the positive direction, with the value "100+%" indicating that the target was exceeded. Objectives like examples 2 and 4 are shown with arrows in the negative direction. In all cases, a footnote indicates that the precise amount could not be calculated. Finally, when the targeted amount of change was small relative to the actual amount of observed change, the PQ can produce relatively large values that are difficult to interpret. Measuring Quality and Years of Healthy Life
Goal 1 of Healthy People 2010 is to increase the quality and years of healthy life. This goal is tracked with three summary measures of health that belong to the family of measures called "healthy life expectancy." The measures are (1) expected years of life in good or better health, (2) expected years of life free of activity limitation, and (3) expected years of life free of selected chronic diseases. The measures are given in life-years, which indicate the average number of healthy years a person can expect to live if age-specific death rates and age-specific illness rates remain the same throughout his or her lifetime. Thus, healthy life expectancy is a "snapshot" of current death and illness patterns and can illustrate the long-range implications of the prevailing age-specific death and illness rates. The methods used to create the healthy life expectancy measures are described here.
Methods To produce healthy life expectancies, age-specific death rates are combined with age-specific health prevalence rates to produce an estimate of overall healthy life expectancy. (See Molla et al.8 for details on calculating healthy life expectancy.) The following terms are used as column headings in a life table:
Life tables used to calculate healthy life expectancy contain all of the columns described above, as well as the following columns of terms that refer to illness:
The use of healthy life expectancies allows for easy comparisons across populations, as well as over long periods of time. The use of the Sullivan method for estimating healthy life expectancies is most appropriate for the cross-sectional data used to track Healthy People 2010.9 Data Systems These data systems are used for the study of healthy life expectancy because they contain detailed information on health and death. However, the institutionalized population is excluded from the NHIS sample. Because the institutionalized population is more likely to report poor health, the Healthy People 2010 healthy life expectancy measures might underestimate the effect of poor health on healthy life expectancies. Survey Questions Activity limitation is measured using questions about personal care needs, limitations of activities, and use of special equipment. Adults are asked whether they need assistance with personal care needs, such as eating, bathing, dressing, or getting around inside the home; whether they need assistance with routine care needs, such as household chores; and whether they have a mental or physical problem that keeps them from working at a job or limits their activity in any way. They also are asked whether they have health problems that require the use of special equipment, such as a cane, wheelchair, or special telephone. If a respondent answers "yes" to any of these questions, he or she is classified as having activity limitations. Children are considered limited in activity if the proxy adult respondent responds "yes" to any of the limitation, special services, or special equipment questions that are specific to children. Selected chronic disease prevalence is measured by several questions that ask respondents whether a doctor has ever diagnosed them with a given disease. The list of selected chronic diseases represents those chronic diseases included in Healthy People 2010 and NHIS: arthritis, asthma, cancer, diabetes, heart disease, high blood pressure, kidney disease, and stroke. Thus, the current measure might underestimate the contribution of chronic disease to healthy life expectancy because all chronic diseases are not included. Additional chronic disease measures might be added in the future if they are supported by a nationally representative data source. If a respondent answers "yes" to any of the selected diagnoses, he or she is classified as having a chronic disease. For children, not all of the selected chronic diseases have the same relationship to risk of death, and ideally such a healthy life expectancy would adjust for the severity of the disease. However, NHIS does not collect data on the severity of the disease. Healthy People 2000 Limitations In addition, the Sullivan method can be biased when evaluating trends over a short period of time. Biases in trends of healthy life expectancy can occur if there are fluctuations in health over a short time period. The Sullivan method is less likely to give misleading estimates of trends in healthy life expectancy when changes in death rates and health status rates are smooth and relatively even. Future Plans Goal 1 of Healthy People 2010 challenges the Nation to assess and measure the complex interactions of health, disease, disability, and premature death in order to increase quality of life and years of healthy life. Refining summary measures of population health, expanding data collection, and implementing effective disease prevention and health promotion interventions will likely assist in making effective progress on this goal by the end of the decade. Measuring Health Disparities
The second overarching goal of Healthy People 2010 calls for eliminating health disparities among segments of the population, including differences that occur by race or ethnicity, gender, education or income, geographic location, disability, or sexual orientation.3 These characteristics are applicable to objectives that measure aspects of the health of the population and do not apply to objectives that are based on schools, worksites, States, or other nonpopulation units of measures. Information about disparities for the population-based objectives is provided in the second figure in each focus area section. These disparities tables summarize information about the size of disparities at the most recent data point and changes in disparities over time for each population-based objective and each relevant characteristic.
The methods used to create the disparities tables are described here. The information in the tables is summarized in the Executive Summary under Goal 2: Eliminate Health Disparities. The rationale for methods used in measuring disparity in Healthy People 2010 is provided in a previous report.11 Measuring Objectives and Defining Groups
Federal data systems have been revising their collection and tabulation procedures to comply with the new standards on racial and ethnic identification. Some data systems began reporting data for calendar year 1999 using the new standards. Other data systems are still in the process of adopting the revised standards. Consequently, the availability of comparable data for racial and ethnic groups varies by data source and across objectives. DATA2010, the interactive online database containing the baseline and tracking data for the Healthy People 2010 objectives, is adding data for the new categories as soon as they become available. Seven racial and ethnic groups are shown in the disparity tables in this report: American Indian or Alaska Native, Asian, Native Hawaiian or other Pacific Islander, two or more races, Hispanic, white non-Hispanic, and black non-Hispanic. The first four groups might include small numbers of persons of Hispanic origin. The data systems used to track the population-based objectives in Healthy People 2010 might not provide data for all of these groups. Departures from this classification scheme are indicated by footnotes in the disparity table for each focus area. To maintain comparability of data by race and ethnicity over time for some objectives, a more recent data year might be used as the baseline because of the revised standards.9 For example, NHIS began reporting data according to the new racial and ethnic categories in 1999. Although the baseline year for objectives tracked with NHIS might be 1997 or 1998, data for 1999 are used as the baseline for measuring disparities for race and ethnicity data only. These departures are indicated by footnotes in the disparities table for each focus area. Availability of Data Content of the Disparities Table Measuring Disparities From the Best Group Rate Disparities are measured as the percent difference from the best group rate for each characteristic. The group with the best, or most favorable, rate (that also meets an additional criterion specified below) is identified for each characteristic in the table by a "B." The percent difference for each of the other groups associated with a characteristic is computed as follows: Percent difference = (Ri − RB ) / RB × 100
where RB is the best group rate for a particular characteristic and Ri is the rate for any of the other groups of interest for the same characteristic. For example, racial and ethnic disparities are measured as the percent difference between the best racial and ethnic group rate and each of the other racial and ethnic group rates. Gender disparities are measured as the percent difference between the better group rate (male or female) and the rate for the other gender group, and so on. In rare instances when two groups for a characteristic have identical best rates, both groups are identified by a "B." To ensure that disparity is measured from a reasonably stable data point, the most favorable group rate must have a relative standard error of less than 10 percent. When the relative standard error for the most favorable group rate is greater than or equal to 10 percent, a small letter "b" is included in the cell and the next most favorable group rate with a relative standard error of less than 10 percent is identified as the reference point. The percent difference is not calculated for cells identified by a small letter "b." When only one group has a relative standard error of less than 10 percent, a best group is not identified for purposes of measuring disparity, and the cells for all groups with data are blank—indicating that disparity cannot be assessed. When standard errors are not available, the best group is determined by the most favorable rate. The first section of the legend for the disparities table (reproduced below) addresses the identification of the best group rate for each characteristic.
Figure 1. Legend for the second figure in the focus area chapters
To ensure comparability across objectives in the measurement of disparity, dichotomous measures are expressed in terms of adverse events when the percent difference is calculated.9, 15, 16 The objective is not restated or changed; it is expressed in terms of the adverse event only when the percent difference is calculated for the analyses presented in the disparities table. For example, objective 1-1, to increase the proportion of persons with health insurance, is expressed in terms of the percentage of persons without health insurance when the percent difference from the best group rate and the change in disparity over time are calculated. Representing the Size of a Disparity by the Color Gradient The statistical significance of the difference between groups can be assessed using the following Z statistic: where Ri is the rate for a group of interest, Rb is the rate for the best group, SEi is the standard error of the rate for a group of interest, and SEB is the standard error of the best group rate. This formula assumes that the groups are independent. Because the comparison is made to the best rate, measured in terms of adverse events, the other group rate can only be larger; therefore, a one-tailed test is used. If |Z| ≥ 1.645, the difference between the groups is significant at an alpha level of 0.05. If the difference between the rate for a group of interest and the best group rate is significant, the percent difference is considered significant. Changes in Disparities Over Time If the group with the best rate changes over time, the percent difference between the rate for a group of interest and the best group rate at the baseline is subtracted from the percent difference between the same two groups at the most recent data point—except that the group of interest is now the best group rate and the best group rate is now the group of interest. If the difference is positive, the increase in disparity applies to the group that had the best rate at baseline. If the difference is negative, the decrease in disparity applies to the group with the best rate at the most recent data point. When standard errors are available for a data system, only statistically significant changes between the baseline and the most recent data point are indicated with arrows (see the legend above). Several steps are required to evaluate the statistical significance of a change in the percent difference over time. The percent difference (PD) at each point in time is based on the ratio of the simple difference (SD) between the rate for the group of interest and the best group rate to the best group rate. The relative standard error (RSE) of a ratio is computed based on the RSE of the numerator and the denominator. The RSE for the SD is calculated as follows: where SEi is the standard error of the rate for a group of interest (i), SEB is the standard error for the best rate (B), Ri is the rate for a group of interest, and Rb is the best group rate. The relative standard error of the rate for the best group is computed as follows: The relative standard error for the percent difference RSEPD is computed based on the relative standard errors of the numerator (RSESD) and the denominator (RSEr) as follows: The standard error of the percent difference is obtained from the relative standard error as follows: The statistical significance of a change in the percent difference from the best group rate over time at the 0.05 level is assessed using the following Z statistic: where PD1 is the percent difference at the most recent time (1), PD0 is the percent difference at the baseline (0) SEPD1 is the standard error of the percent difference at the most recent time (1), and SEPD0 is the standard error of the percent difference at the baseline (0). When standard errors were available at one point in time but not at the other, statistical tests were performed applying the known standard error to the estimate at the point in time for which standard errors were missing. When standard errors were not available, changes are indicated by arrows based on the size of the change alone. Summary Measures Summary index = where PDi is the percent difference from the best group rate for each of the groups of interest (i), and (n – 1) is the number of groups minus one. Because the percent differences are calculated with the best group rate as the reference point, the number of comparisons is equal to the number of groups minus one.These comparisons are made only when data are available for the same groups defined in the same way at the baseline and the most recent data point. The statistical significance of a change in the index over time is assessed when standard errors are available for the rates on which the index is based. The magnitude and direction of changes are indicated by arrow symbols as described above. When standard errors are not available for the rates on which the index is based, changes are classified by size and direction without regard to statistical significance. To obtain a standard error for the index, a type of resampling or "bootstrap" procedure is employed.9, 18 This procedure uses the rate and standard error for each group to reestimate each group rate 25,000 times, assuming a random normal distribution. Based on these group rates, 25,000 estimates of the index of disparity are generated, and the distribution of these estimates is used to estimate the standard error of the index. The bootstrap procedure is used to estimate standard errors for the index at the most recent time (1) and at the baseline (0) to determine whether a change in the index over time is statistically significant. A Z test for the change in the index of disparity can be computed as follows: where ID1 is the index of disparity at the most recent time (1), ID0 is the index of disparity at the baseline (0), SE1 is the standard error for the index of disparity at the most recent time (1), and SE0 is the standard error for the index of disparity at the baseline (0). Because the value of the index could increase or decrease, a two-tailed test is used with a critical value of |Z| ≥ 1.96 at the 0.05 level. Estimates of Variability When standard errors are not available, the variability of the best group rate was not assessed, and tests of statistical significance could not be performed. For objectives based on these data sources, there is no quantifiable assurance that observed disparities and changes in disparity are not due to random variability. As a consequence, more changes in disparity are evident for objectives based on data sources without estimates of variability. In the disparities tables, objectives based on data for which estimates of variability are available and those for which estimates of variability are not available are designated by footnotes following the baseline and most recent data years in parentheses:
DATA2010 DATA2010 is an online, searchable database that contains baseline data, tracking data, and targets for all measurable objectives in Healthy People 2010.4 The database is updated quarterly to provide the most accurate and up-to-date data for tracking Healthy People 2010 objectives.
DATA2010 allows users to search the database for estimates by focus area, objective, data source, and keyword. In addition, users can access current data by downloading standard or statistical data spreadsheets by focus area. Standard spreadsheets contain rounded estimates, whereas statistical spreadsheets contain rounded data as well as raw data and standard errors (both rounded to one decimal place), when available. Users can access the Healthy People 2010 midcourse review data by downloading designated static midcourse review standard and statistical tables accessible at http://wonder.cdc.gov/data2010/ftpselec.htm. These tables do not reflect postmidcourse revisions or updates or data acquired since January 2005. All of the postmidcourse review revisions and updates are included in the current database selections. All of the data used to produce the midcourse PQ and disparities charts are reflected in these static midcourse review tables. PQs were calculated using rounded estimates. Measures of disparity were calculated using raw estimates and their associated standard errors, when available. In addition, DATA2010 contains other technical information related to the Healthy People 2010 objectives at the midcourse, including updated operational definitions for each measure. Target Adjustment
Target Adjustments for Objectives With Revised Baselines
Targets were adjusted for some objectives for which a change was made to the total population baseline data point since the publication of Healthy People 2010.3 Baseline data were changed because of revisions in methodology, survey questions, baseline year, and population denominators. Baseline data for several objectives were revised to accommodate updated public health recommendations. In several cases, baseline data were revised because the published data were based on preliminary analyses. Target revisions were not made in cases in which the baseline data for a select population had changed but data for the total population were unchanged. Baselines for the majority of objectives for specific causes of death from NVSS were revised from data year 1997 to 1999. These changes were made for two reasons. First, NVSS implemented the 10th revision of the International Classification of Diseases (ICD) with data year 1999, creating a discontinuity in cause-specific death data between 1999 and previous years. In addition, as of data year 1999, the standard population used for death age adjustment changed from the 1940 to the 2000 standard million population, creating a second discontinuity with previously published death data.19 The revision of the death baselines to 1999 avoided the discontinuities associated with the ICD and the age-adjustment changes. More information about the impact of the ICD revision and the new age-adjustment standard is available in Tracking Healthy People 2010. Several general methodologic changes for NHIS occurred since the launch of Healthy People 2010. These changes included the imputation of missing values for family income using multiple imputation methodology, revised standard error methodology, and population weights based on the 2000 census. These changes required baseline revisions to objectives using NHIS as a data source. Baseline data for 145 objectives or subobjectives were revised since the publication of Healthy People 2010. New targets were calculated using the following criteria:
Corrections From the Original Healthy People 2010 Publication and Special Cases
Tracking Healthy People 2010
Tracking Healthy People 2010 is a comprehensive guidebook on the statistics used for Healthy People 2010.12 It provides detailed information on how the data are derived and the major issues affecting the interpretation of the statistics. The guidebook ensures greater accuracy and comparability in the data produced for and used by Healthy People 2010 programs at the local, State, and national levels.
During the Healthy People 2010 midcourse review, the three parts of Tracking Healthy People 2010 described below were updated.20 Part A: General Data Issues Part B: Operational Definitions Part C: Major Data Sources The appendices of the updated Tracking Healthy People 2010 provide information on the following:
Appendices on baselines for age-adjusted death objectives using rates age adjusted to 1940 and 2000 standards and on crosswalks between the Healthy People 2000 and Healthy People 2010 objectives (previously included in the original Tracking Healthy People 2010 publication) were not included in the revised Tracking Healthy People 2010 publication. The revised Tracking Healthy People 2010 publication is accessible from the Healthy People 2010 website at www.cdc.gov/nchs/hphome.htm. References1U.S. Department of Health and Human Services (HHS). Healthy People 2000 Midcourse Review and 1995 Revisions. Washington, DC: U.S. Public Health Service (PHS), 1995. 2National Center for Health Statistics (NCHS). Healthy People 2000 Final Review. Hyattsville, MD: PHS, 2001. 3HHS. Healthy People 2010: With Understanding and Improving Health and Objectives for Improving Health. 2nd ed. Washington, DC: U.S. Government Printing Office (GPO), November 2000. 4DATA2010. More information available at http://wonder.cdc.gov/data2010; accessed October 31, 2006. 5Sullivan, D.F. A single index of mortality and morbidity. HSMHS Health Reports 86:347–354, 1971. 6Sullivan, D.F. Disability components for an index of health. NCHS. Vital and Health Statistics 2(42), 1971. 7Anderson, R.N. Method for constructing complete annual U.S. life tables. NCHS. Vital and Health Statistics 2(129), 1999. 8Molla, M.T., et al. Summary Measures of Population Health: Report of Findings on Methodologic and Data Issues. Hyattsville, MD: NCHS, 2003. 9Crimmins, E.M. What can we expect from summary indicators of population health? In C.J.L. Murray et al., eds. Summary Measures of Population Health. Geneva, Switzerland: World Health Organization, 2002. 10Idler, E., and Benyamini, Y. Self-rated health and mortality: A review of 28 studies. Journal of Health and Social Behaviors 38(1):21–37, 1997. 11Erickson, P., et al. Years of healthy life. Healthy People 2000: Statistical Notes. No. 7. Hyattsville, MD: NCHS, 1995. 12Keppel, K.G., et al. Measuring progress in Healthy People 2010. Healthy People 2010: Statistical Notes. No. 25. Hyattsville, MD: NCHS, September 2004. 13HHS. Tracking Healthy People 2010. Washington, DC: GPO, November 2000. 14Office of Management and Budget. Revisions to the standards for the classification of Federal data on race and ethnicity. Federal Register 62/FR:58781–58790, 1997. 15Klein, R.J., et al. Healthy People 2010 criteria for data suppression. Healthy People 2010: Statistical Notes. No. 24. Hyattsville, MD: NCHS, June 2002. 16Keppel, K.G., et al. Methodological issues in measuring health disparities. NCHS. Vital and Health Statistics 2(141), 2005. 17Keppel, K.G., and Pearcy, J.N. Measuring relative disparities in terms of adverse outcomes. Journal of Public Health Management and Practice 11(6), 2005. 18Pearcy, J.N., and Keppel, K.G. A summary measure of health disparity. Public Health Reports 117:273–280, May–June 2002. 19Efron, B. The Jackknife, the Bootstrap, and Other Resampling Plans. Philadelphia, PA: SIAM Publishing Company, 1982. 20Tracking Healthy People 2010 is accessible from the Healthy People 2010 website at www.cdc.gov/nchs/hphome.htm. << Previous—Appendix B: Healthy People 2010 Workgroup Coordinators | Table of Contents | Next—Appendix D: List of Healthy People 2010 Objectives Deleted at the Midcourse Review >> |