J. Michael McWilliams, MD, PhD; Ellen Meara, PhD; Alan M. Zaslavsky, PhD; John Z. Ayanian, MD, MPP
Acknowledgment: The authors thank Katherine Swartz, PhD, for helpful comments on an earlier draft of this manuscript.
Grant Support: By The Commonwealth Fund (grant no. 20060485).
Potential Financial Conflicts of Interest:Consultancies: E. Meara (Employment Policies Institute).
Reproducible Research Statement:Study protocol and statistical code: Available from Dr. McWilliams (e-mail, email@example.com). Data set: Available at http://www.cdc.gov/nchs/nhanes.htm.
Requests for Single Reprints: John Z. Ayanian, MD, MPP, Department of Health Care Policy, Harvard Medical School, 180 Longwood Avenue, Boston, MA 02115; e-mail, firstname.lastname@example.org.
Current Author Addresses: Drs. McWilliams, Meara, Zaslavsky, and Ayanian: Harvard Medical School, Department of Health Care Policy, 180 Longwood Avenue, Boston, MA 02115.
Author Contributions: Conception and design: J.M. McWilliams, E. Meara, A.M. Zaslavsky, J.Z. Ayanian.
Analysis and interpretation of the data: J.M. McWilliams, E. Meara, A.M. Zaslavsky, J.Z. Ayanian.
Drafting of the article: J.M. McWilliams.
Critical revision of the article for important intellectual content: J.M. McWilliams, E. Meara, A.M. Zaslavsky, J.Z. Ayanian.
Final approval of the article: J.M. McWilliams, E. Meara, A.M. Zaslavsky, J.Z. Ayanian.
Statistical expertise: J.M. McWilliams, E. Meara, A.M. Zaslavsky.
Obtaining of funding: J.M. McWilliams, J.Z. Ayanian.
Administrative, technical, or logistic support: J.M. McWilliams.
Collection and assembly of data: J.M. McWilliams.
McWilliams JM, Meara E, Zaslavsky AM, Ayanian JZ. Differences in Control of Cardiovascular Disease and Diabetes by Race, Ethnicity, and Education: U.S. Trends From 1999 to 2006 and Effects of Medicare Coverage. Ann Intern Med. 2009;150:505-515. doi: 10.7326/0003-4819-150-8-200904210-00005
Download citation file:
Published: Ann Intern Med. 2009;150(8):505-515.
Appendix: Classification of Clinical Conditions
Efforts to improve the care of cardiovascular disease and diabetes or expand insurance coverage for adults with these conditions may reduce differences in clinical outcomes.
To assess recent national trends in disease control, trends in sociodemographic differences in control, and changes in sociodemographic differences after age 65 years associated with near-universal Medicare coverage.
Observational and quasi-experimental analyses of repeated cross-sectional data.
National Health and Nutrition Examination Survey, 1999 to 2006.
Adults age 40 to 85 years with relevant clinical conditions.
Blood pressure control (<140/90 mm Hg) and mean systolic blood pressure among adults with hypertension (n = 4521); glycemic control (hemoglobin A1c levels <7.0%) and mean hemoglobin A1c levels among those with diabetes (n = 1733); and total cholesterol level control (<5.2 mmol/L [<200 mg/dL]) and mean total cholesterol levels among those with coronary heart disease, stroke, or diabetes (n = 2928). Temporal trends in these measures were compared by race, ethnicity, and education, and sociodemographic differences were compared above and below eligibility for Medicare at age 65 years.
Disease control improved significantly between 1999 and 2006 for all 6 measures (PÂ < 0.001). These trends did not differ by race or ethnicity or by education (PÂ â‰¥ 0.185 for groupâ€“time interactions), except that whiteâ€“Hispanic differences in glycemic control widened (P = 0.042). Blackâ€“white differences in systolic blood pressure were smaller among adults age 65 to 85 years than among adults age 40 to 64 years (reduction in difference, 4.2 mm Hg; P = 0.009). Blackâ€“white differences in hemoglobin A1c levels were also smaller after age 65 years (reduction in difference, 0.7%; P = 0.005), as were Hispanicâ€“white differences (reduction in difference, 0.7%; P = 0.007) and differences between less and more educated adults (reduction in difference, 0.5%; P = 0.033).
Data were cross-sectional, and estimates may have been biased by coincidental events at age 65 years, such as retirement, that may affect disease control.
Control of blood pressure and glucose and cholesterol levels has improved since 1999 for adults with cardiovascular disease and diabetes, but racial, ethnic, or socioeconomic differences have not narrowed significantly. Medicare coverage after age 65 years is associated with reductions in these differences.
The Commonwealth Fund.
Acquiring health insurance and getting better quality of care could reduce health care disparities. The relative importance of these 2 factors is unknown.
To measure changes in chronic disease control, the authors used blood pressure, hemoglobin A1c, and total cholesterol measurements that were obtained from participants in the 1999 to 2006 National Health and Nutrition Examination Survey. Disease control improved over 8 years, but gaps between white and nonwhite patients did not change. The gaps were smaller after age 65 years, when universal Medicare insurance begins.
Each annual National Health and Nutrition Examination Survey enrolled different persons.
Access to care through universal health insurance reduced disparities in chronic disease control; improved quality of care did not affect sociodemographic differences.
In 3 comprehensive reports since 2001, the Institute of Medicine has advanced recommendations to expand access (1), improve quality (2), and eliminate disparities in health care (3). Although widespread deficits in the quality of care have been reported in the United States (4), some evidence suggests that quality of care has improved in the past decade (5–8). More consistent efforts to provide high-quality care may also reduce racial, ethnic, and socioeconomic differences in health (9). However, quality improvement may not necessarily lead to more equitable care (5, 10, 11), especially if improvements occur among providers who serve fewer disadvantaged patients (12–18) or if new financial incentives to improve quality have unintended, detrimental consequences (18–20). Furthermore, better performance and smaller racial differences in processes of care for cardiovascular disease and diabetes have not been consistently associated with reduced racial differences in clinical outcomes, such as control of cholesterol or glucose levels (8, 21, 22). Although overall disease control in the United States may be improving for some measures (23–25), recent national trends in sociodemographic differences in control have not been comprehensively assessed.
Insurance coverage may be an important mediator of sociodemographic differences in control of cardiovascular disease and diabetes (3, 26). Racial and ethnic minorities and adults of lower socioeconomic status are much more likely to be uninsured (27), and uninsured adults are much less likely to receive basic clinical services for these conditions (28). Near-universal Medicare coverage after age 65 years has been associated with decreased racial and socioeconomic differences in self-reported general health status and receipt of mammography (29, 30). Recent longitudinal studies also suggest that acquiring Medicare coverage increases use of health services and improves self-reported health outcomes for previously uninsured adults with cardiovascular disease or diabetes (31–33). However, previous studies have not assessed the effects of increases in insurance coverage on racial, ethnic, and socioeconomic differences in clinical measures of disease control.
Our primary objectives were to assess national trends from 1999 to 2006 in blood pressure control (for adults with hypertension), glycemic control (for adults with diabetes), and cholesterol level control (for adults with coronary heart disease, stroke, or diabetes); to analyze concomitant changes in differences by race, ethnicity, and education for each of these measures; and to evaluate whether these differences narrow after age 65 years with Medicare coverage.
We analyzed serial cross-sectional data from the National Health and Nutrition Examination Survey (NHANES), a nationally representative study designed to assess population health through interviews, physical examinations, and clinical testing (34). From 1999 to 2006, 41 474 noninstitutionalized adults and children were enrolled (average response rate, 81.4%), including oversamples of black adults, Mexican Americans, and adults age 60 years or older. Data have been released in four 2-year increments. Among participants who completed a standardized interview in English or Spanish, 39 352 (94.9%) had clinical examinations and testing, including 12 079 participants age 40 to 85 years.
We studied adults age 40 to 85 years who had at least 1 of the following conditions: diabetes, hypertension, coronary heart disease, or stroke (Table 1). If we assessed disease control only among those with self-reported diagnoses, improved diagnosis of less severe disease might bias estimates of time trends in disease control in a positive direction. Therefore, on the basis of relevant clinical testing, we identified and included adults with undiagnosed hypertension and diabetes so that disease control was consistently assessed among all prevalent cases (a detailed classification of conditions is in the Appendix).
Because the proportion of immigrant Hispanic adults varies by age and has increased over time (35) and because Hispanic immigrants experience different patterns of chronic disease care and outcomes from those of U.S.-born Hispanic adults (36), we excluded 743 Hispanic adults (53.1%) who were born outside the United States, restricting all analyses to the U.S.-born Hispanic group (hereafter referred to as Hispanic). This restriction improved comparability of Hispanic samples over time and ensured that age-related differences in outcomes were not confounded by health differences between immigrant and U.S.-born Hispanic adults. In contrast, only 5.2% of white participants and 12.7% of black participants were born outside the United States. Results were similar when we excluded white and black immigrants in a sensitivity analysis. Finally, we excluded 202 participants who were not white, black, or Hispanic because samples for other groups were too small for statistical comparisons. The Human Studies Committee of Harvard Medical School approved our study protocol.
We used dichotomous and continuous measures of disease control to compare rates of control and mean levels, respectively, across groups. We assessed blood pressure control (average systolic blood pressure <140 mm Hg and diastolic blood pressure <90 mm Hg) (37) and average systolic blood pressure readings among participants with hypertension; glycemic control (hemoglobin A1c levels <7.0%) (38) and hemoglobin A1c levels among participants with diabetes; and cholesterol level control (total cholesterol level <200 mg/dL [<5.2 mmol/L]) (39) and total cholesterol levels among participants with coronary heart disease, stroke, or diabetes. We defined comparison groups by race or ethnicity (non-Hispanic black and Hispanic each vs. non-Hispanic white) or education (high school graduates vs. non–high school graduates). We also determined age, sex, ratio of family income to poverty threshold, body mass index (BMI), current smoking status, and insurance coverage from NHANES data.
We compared trends in disease control for each measure by using a linear model (E[Yi] = β0 + β1timei + β2groupi + β3timei × groupi), in which Yi is a dichotomous or continuous indicator of disease control for the ith individual; time is a chronologic index of the four 2-year survey periods ranging from 1 (1999 to 2000) to 4 (2005 to 2006); and group is an indicator of membership in a particular racial, ethnic, or educational comparison group. Thus, coefficients for the time-by-group interaction terms represent biennial trends (average change over 2-year periods between survey waves) in racial, ethnic, and educational differences in disease control from 1999 to 2006. We estimated overall trends by using simpler models without group variables. We also calculated rates of control and mean values for each of the four 2-year periods for reporting purposes.
We adjusted all reported estimates of rates and trends for age and sex. To determine whether trends in disease control were related to changes in other population characteristics, we also estimated overall trends that were further adjusted for race, ethnicity, education, income, BMI, smoking status, and insurance coverage. We did not adjust analyses of racial, ethnic, or educational differences in disease control for factors other than age and sex, because we were interested in overall differences that could result from many individual and health care system factors, rather than attributing differences to specific mediators, such as discrimination (40).
To estimate effects of near-universal Medicare coverage on sociodemographic differences, we compared racial, ethnic, and educational differences in systolic blood pressure, hemoglobin A1c levels, and total cholesterol levels before and after age 65 years. For example, to identify changes in racial differences in systolic blood pressure associated with Medicare eligibility, we fitted a linear model predicting mean systolic blood pressure as a function of black race, an indicator of age 65 years or older, and an interaction between these 2 predictors, with white adults serving as the reference group. We fitted similar models for each racial, ethnic, and educational comparison and for each outcome. We used measured values rather than dichotomous outcomes in these analyses to identify clinically important changes with greater sensitivity.
In our study, differences in mortality rates, time-varying characteristics, and use of cross-sectional data posed several challenges to interpreting age-related changes in sociodemographic differences in disease control. Older groups may have differed from younger groups in predictors of disease control other than age or insurance coverage, and these differences between age groups may have differed by sociodemographic characteristics.
Therefore, to estimate effects of Medicare coverage more robustly, we made several prespecified modifications to improve comparability of adults age 40 to 64 years with adults age 65 to 85 years. First, we adjusted for sex because the female share of the population increases with age. Second, we restricted analyses of hemoglobin A1c and total cholesterol levels to 1091 and 1877 participants, respectively, with relevant conditions diagnosed before age 65 years so that adults who developed conditions after age 65 years did not enter comparisons without appropriate comparators before age 65 years. Third, because age at diagnosis was not reported for hypertension, we added 1562 obese participants (BMI ≥30 kg/m2) to the sample of adults with diagnosed or undiagnosed hypertension. This addition captured those who did not have hypertension but were at increased future risk, thereby better approximating an aging cohort from serial cross-sectional samples.
These refinements to the study populations reduced selection bias in comparisons of sociodemographic differences in disease control by age eligibility for Medicare. For example, if hypertension developed at later ages for white adults than for black adults and was less severe or more easily controlled early in its course, limiting comparisons to participants with evident hypertension would disproportionately select elderly white adults with comparatively lower blood pressures into analyses, biasing estimated effects of Medicare coverage on racial differences in blood pressure control toward zero. The addition of obese adults (in effect pooling hypertensive adults with those at high risk for becoming hypertensive) made it more likely that effects of aging on disease control were similar for comparison groups, thereby allowing us to separate differential changes in control due to Medicare coverage from racial differences in age of hypertension onset. Furthermore, these modifications addressed potential bias from underdiagnosis of conditions among uninsured nonelderly adults, which would have caused disadvantaged participants to be disproportionately excluded before age 65 years (41). For example, if comparisons included participants with conditions diagnosed at any age but not undiagnosed cases, and if undiagnosed cases were poorly controlled, we would understate sociodemographic differences before age 65 years relative to differences after age 65 years.
In addition, we conducted 2 sensitivity analyses to determine whether differences in blood pressure, hemoglobin A1c levels, and total cholesterol levels changed in conjunction with Medicare coverage at age 65 years or evolved gradually over earlier or later ages. First, we restricted analyses to age 63 to 67 years to ascertain differences just before and after age 65 years. Second, we varied the cutoff for the age indicator variable from 61 to 69 years by 1-year increments to determine the age at which racial, ethnic, and educational differences in disease control changed most sharply.
We adjusted all analyses for the complex survey design (42) by using SAS, version 9.1 (SAS Institute, Cary, North Carolina), and SUDAAN, version 9.0 (Research Triangle Institute, Research Triangle Park, North Carolina). We used robust variance estimators (43) appropriate to the clustered survey design to test model coefficient estimates, obtain 2-sided P values, and construct 95% CIs.
This study was funded by The Commonwealth Fund. The funding source had no role in the design, conduct, or reporting of this study, nor in the decision to submit the manuscript for publication.
Table 1 summarizes sample sizes and selected characteristics of study samples with specific conditions. The prevalence of hypertension, coronary heart disease, and stroke (data not shown) did not change over the study period (P > 0.23 for tests of trend), but the prevalence of diabetes (diagnosed or undiagnosed) increased from 11.8% (1999 to 2000) to 13.7% (2005 to 2006) (P = 0.043 for trend), which is consistent with other data on diabetes trends (44).
Age- and sex-adjusted trends indicated significant improvements in control of blood pressure, hemoglobin A1c levels, and total cholesterol levels (Figure 1). From 1999 to 2000 to 2005 to 2006, absolute improvements in control rates ranged from 10.3% for blood pressure control among adults with hypertension to 21.0% for glycemic control among adults with diabetes. Estimates of biennial trends in disease control remained essentially unchanged after further adjustment for race, ethnicity, education, income, BMI, smoking status, and insurance coverage.
Rates are plotted by survey period with 95% CIs and fitted trends from linear models. Disease control improved significantly from 1999 to 2006 for all 6 measures (all P < 0.001). A. Age- and sex-adjusted rates of blood pressure control and mean systolic blood pressure among adults with hypertension. B. Age- and sex-adjusted rates of glycemic control and mean hemoglobin A1c levels among adults with diabetes. C. Age- and sex-adjusted rates of total cholesterol level control and mean total cholesterol levels among adults with coronary heart disease, stroke, or diabetes. To convert cholesterol values from mg/dL to mmol/L, multiply by 0.0259.
Pooled over 8 years, rates of blood pressure control were significantly lower and mean systolic blood pressure was significantly higher for black and Hispanic adults than for white adults with hypertension (Table 2). Among adults with diabetes, rates of glycemic control were significantly lower and mean hemoglobin A1c levels were significantly higher for black and Hispanic than for white adults. These racial and ethnic differences in disease control did not change statistically over the study period, except that differences in glycemic control between Hispanic and white adults widened significantly from 1999 to 2006 (Figure 2). Glycemic control was significantly worse for less educated adults than more educated adults with diabetes, and mean systolic blood pressure was significantly higher for less educated than more educated adults with hypertension (Table 2). These educational differences also did not significantly change over the study period (trends not shown). Among adults with coronary heart disease, stroke, or diabetes, total cholesterol level control and mean total cholesterol levels did not differ by race, ethnicity, or education (Table 2).
Rates are plotted by race and ethnicity and survey period among adults with hypertension (A and B) and diabetes (C and D) with 95% CIs and fitted trends from linear models. Differences between black or Hispanic and white adults did not significantly change from 1999 to 2006 for any of the 4 measures of disease control, except differences in rates of glycemic control between Hispanic and white adults widened (P = 0.042). A. Age- and sex-adjusted rates of blood pressure control. B. Age- and sex-adjusted mean systolic blood pressure. C. Age- and sex-adjusted rates of glycemic control. D. Age- and sex-adjusted mean hemoglobin A1c levels.
On the basis of cross-sectional rates, age eligibility for Medicare was associated with increases in insurance coverage that were 6.6 percentage points greater for nonwhite than for white adults (P < 0.001) and 12.6 percentage points greater for less educated adults than for more educated adults (P < 0.001).
Most sociodemographic differences in mean systolic blood pressure, hemoglobin A1c levels, and total cholesterol levels among adults younger than 65 years were substantially smaller than those among adults age 65 years or older (Figure 3). Differences in systolic blood pressure between black and white adults with hypertension or obesity were significantly smaller among adults age 65 to 85 years than among adults age 40 to 64 years (Figure 3, A). Mean systolic blood pressure was slightly but significantly higher among less educated adults than among more educated adults before age 65 years (3.2 mm Hg [95% CI, 0.4 to 5.9 mm Hg]; P = 0.024), but this difference did not significantly change after age 65 years (data not shown). Systolic blood pressure did not significantly differ before age 65 years between Hispanic and white adults with hypertension or obesity.
All differences are displayed by age group; bars represent SEs. Higher values indicate greater disparities. Age groups are collapsed to reduce random variation, but narrower 5-year age intervals are shown adjacent to age 65 years to demonstrate notch effects. Differences among adults age 65 to 85 years were significantly smaller than those among adults age 40 to 64 years for systolic blood pressure (P = 0.009 for black–white comparison) and hemoglobin A1c levels (P < 0.001 for nonwhite–white comparison; P = 0.033 for comparison by education) and tended to be smaller for total cholesterol levels (P = 0.087 for comparison by education). Changes after age 65 years in black–white and Hispanic–white differences in hemoglobin A1c levels were equivalent, were each statistically significant, and were therefore combined into nonwhite–white differences to summarize. A. Differences in mean systolic blood pressure between black and white adults with hypertension or obesity. B. Differences in hemoglobin A1c levels between nonwhite and white adults with diabetes diagnosed before age 65 years. C. Differences in hemoglobin A1c levels between less and more educated adults with diabetes diagnosed before age 65 years. D. Differences in total cholesterol levels between less and more educated adults with coronary heart disease, stroke, or diabetes diagnosed before age 65 years. To convert cholesterol values from mg/dL to mmol/L, multiply by 0.0259.
For adults with diabetes diagnosed before age 65 years, racial, ethnic, and educational differences in hemoglobin A1c levels were significantly smaller among adults age 65 to 85 years than among those age 40 to 64 years (Figure 3, B and C).
Among adults with coronary heart disease, stroke, or diabetes diagnosed before age 65 years, educational differences in total cholesterol levels also tended to be smaller among adults age 65 years or older (Figure 3, D). Total cholesterol levels did not differ before age 65 years by race or ethnicity among adults with these conditions.
Several findings supported the robustness of these results related to the effects of Medicare. When comparisons were restricted to age 63 to 67 years, statistical trends (P < 0.100) remained for all significant reductions in group differences after age 65 years because these findings became even greater in magnitude than in comparisons of broader age groups. In addition, differential improvements in disease control for black, Hispanic, and less educated adults were greatest when group differences were compared before and after age 65, 66, or 67 years rather than other ages between 61 and 69 years.
In time–trend analyses of nationally representative data, blood pressure, glycemic, and cholesterol level control improved significantly from 1999 to 2006 for adults with cardiovascular disease or diabetes. Overall trends were not explained by changes in population characteristics, suggesting that management of cardiovascular disease and diabetes has improved considerably over this period. These improvements probably have prevented adverse outcomes, increased life expectancy, and provided considerable value to society (31, 32, 37–39, 45–52). However, improvements in disease control have not been associated with discernable reductions in racial, ethnic, or socioeconomic differences in blood pressure or glycemic control, and gaps in glycemic control between white and Hispanic adults with diabetes have widened. Therefore, to redress inequities in these health outcomes, quality improvement efforts focusing on disadvantaged groups and their health care providers or broader reforms to address social determinants of poor health are still needed (19).
In particular, because black, Hispanic, and less educated adults are much more likely to be uninsured or underinsured (27), expanding insurance coverage may be especially beneficial for these groups. With near-universal Medicare coverage after age 65 years, differences in systolic blood pressure, hemoglobin A1c levels, or total cholesterol levels reduced substantially. These reductions may substantially decrease racial and socioeconomic differences in mortality as well (53, 54). For example, 1 study estimated that eliminating racial differences in mean systolic blood pressure might reduce deaths from heart disease or stroke by more than 7500 annually among black adults (55).
Our findings are consistent with previous studies that used experimental or quasi-experimental designs to assess effects of insurance coverage on health outcomes for adults with cardiovascular disease or diabetes (32, 56–58). Thus, differences in insurance coverage before age 65 years may contribute substantially to sociodemographic differences in health. Sociodemographic comparisons of self-reported insurance coverage at the time of surveys suggested that diminished sociodemographic differences in disease control after age 65 years were due to narrowed gaps in coverage between groups. However, near-universal Medicare coverage after age 65 years probably improved disease control for greater numbers of elderly black, Hispanic, and less educated adults than these comparisons suggested because insurance status was only assessed at the time of surveys. For example, if the greater reduction in mean hemoglobin A1c levels after age 65 years for nonwhite adults relative to white adults (−0.7%) was due entirely to their 6.6% greater increase in cross-sectional rates of coverage after age 65 years, this would imply an implausibly large reduction in mean hemoglobin A1c levels (−0.7%/0.066 = −10.6%) concentrated in a relatively small group of nonwhite adults. Therefore, black, Hispanic, and less educated adults who are underinsured or intermittently uninsured may also benefit from gaining Medicare coverage (32, 59–61).
Strengths of our study included nationally representative data with standardized measurements of clinical outcomes to compare recent trends in disease control and a rigorous study design to assess effects of Medicare coverage on racial, ethnic, and socioeconomic differences in disease control. We searched the clinical literature on PubMed through October 2008 using combinations of keywords relating to health care disparities, uninsurance, and relevant clinical conditions to identify studies similar to our study. We found no previous quasi-experimental studies of the effects of Medicare coverage on sociodemographic differences in disease control. Furthermore, by comprehensively examining recent national trends in racial, ethnic, and socioeconomic differences through 2006, our study builds on previous research that found persistent or widening differences in some measures of disease control over previous decades (10, 62–64) or assessed more recent overall trends in blood pressure, glycemic, and cholesterol level control (23–25, 65).
Limitations of our study included several potential sources of bias. Because we assessed trends in disease control across serial cross-sectional samples, we could not adjust for mortality rate trends that may have differentially altered the distribution of outcome measurements for certain sociodemographic groups. For example, improved care for black adults with hypertension may have decreased cardiovascular mortality among this group with worse disease control, thereby enabling them to survive longer and possibly masking underlying improvements in disease control for lower-risk black adults. However, from 2000 to 2004, age-adjusted annual rates of death from ischemic heart disease, cerebrovascular disease, or diabetes decreased similarly for white (−48 deaths per 100 000 persons), black (−52 deaths per 100 000 persons), and Hispanic (−47 deaths per 100 000 persons) adults (35). Therefore, persistent racial differences in disease control were unlikely to be explained by differential mortality trends.
In assessing the effects of Medicare coverage on sociodemographic differences in disease control, we used a quasi-experimental design to control for racial, ethnic, or educational differences in observed or unobserved characteristics that remained fixed across age groups. This strategy assumed that predictors of disease control other than insurance coverage did not change more after age 65 years for 1 comparison group than for another. Two potential sources of bias threatened this assumption. First, sociodemographic comparison groups did not comprise the same persons at different ages. Rather, the composition of comparison groups in our study could have changed with age differently for different groups because cross-sectional samples of younger and older participants were selected on the basis of the presence of clinical conditions when surveyed. To address this limitation, we modified inclusion criteria to improve comparability across age groups and distinguish effects of Medicare coverage from differences in disease control related to age of disease onset.
Second, even in longitudinal cohorts, time-varying characteristics, such as employment, may change with age differently for different groups, and death may lead to selective attrition. Our sensitivity analyses suggest that these potential confounding effects would have to coincide directly with Medicare eligibility at age 65 years to bias our results substantially. For example, although differential premature mortality might improve average disease control among survivors in some groups, only a spike in deaths immediately before age 65 years among black, Hispanic, and less educated adults could account fully for the reduced differences in disease control associated with age eligibility for Medicare. Because mortality rates increase steadily with age, including the years surrounding age 65 years (29), such a spike in deaths is unlikely. Furthermore, events that more plausibly might coincide with Medicare eligibility, such as retirement and receipt of social security benefits, occur at similar rates for adults with different sociodemographic characteristics or previous insurance coverage (29, 32).
Our study lacked data about health care providers and organizations to explore potential mediators of improved disease management (66–75). In addition, we used self-reported diagnoses to define study populations. Previous studies have found good overall concordance between self-reports and diagnoses established from medical records or other clinical data, with self-reported measures exhibiting high specificity and moderate sensitivity (76–78). A previous NHANES analysis found self-reported hypertension to have a specificity of 90% to 95%, suggesting few false-positive diagnoses (79). By including undiagnosed cases in our study, we improved the sensitivity of these measures.
In examining trends in disease control, we selected clinical targets currently recommended by national guidelines and used to measure health system performance. However, these indicators do not reflect changes in disease prevalence. Gains in disease control may be offset by increases in disease prevalence so trends in national rates of uncontrolled disease are also important to assess. Because the prevalence of cardiovascular disease was stable over the study period, improvements in blood pressure and cholesterol level control also meant net public health gains. Despite the increase in diabetes prevalence, substantial improvements in glycemic control also lowered the rate of uncontrolled diabetes (disease prevalence × [100 − control rate]) from 7.3% in 1999 to 2000 (11.8% × [100 − 38.1%]) to 5.6% in 2005 to 2006 (13.7% × [100 − 59.1%]), which is a 23.3% relative reduction.
Our study was adequately powered to detect small overall improvements in disease control but not small changes in differences over time. In particular, the significant gap in blood pressure control between Hispanic and white adults may have narrowed somewhat, but sample sizes were not sufficient to assess an effect of this size. Nonetheless, we observed that other significant differences in blood pressure and glycemic control grew larger or remained essentially unchanged, including differences in mean systolic blood pressure between Hispanic and white adults, thereby bolstering our conclusion that national improvements in disease control were not accompanied by consistent reductions in racial, ethnic, or socioeconomic differences. Moreover, the difference in glycemic control between Hispanic and white adults significantly widened among adults with diabetes—the smallest of our 3 study populations.
Our study has important policy implications. Control of blood pressure, hemoglobin A1c levels, and total cholesterol levels has improved substantially for adults with cardiovascular disease and diabetes in the United States since 1999, which suggests that recent efforts to improve quality of care for these conditions have yielded widespread benefits. However, racial, ethnic, and socioeconomic differences in blood pressure and glycemic control have persisted or widened, suggesting that more focused efforts are needed to improve the quality of care for black, Hispanic, and less educated patients. Because age eligibility for Medicare coverage was associated with significant narrowing of differences in disease control, expanding insurance coverage before age 65 years may reduce racial, ethnic, and socioeconomic differences in important health outcomes for adults with cardiovascular disease and diabetes.
Participants were considered to have hypertension if they reported taking antihypertensive medication for high blood pressure or had a systolic blood pressure of 150 mm Hg or greater or a diastolic blood pressure of 95 mm Hg or greater when averaged over 1 to 3 readings taken during a single examination session (80). We used this higher diagnostic threshold to distinguish persons who were likely to have undiagnosed hypertension from those with isolated mild elevations. Participants were classified as having diabetes if they reported a diagnosis of diabetes or had a hemoglobin A1c level greater than 7.0%. We did not use measurements of fasting blood glucose levels to define undiagnosed diabetes because fewer than half of participants were fasting at the time of their examination. Participants reporting borderline diabetes (n = 198) were also included because 21.7% of them were taking insulin or oral medications for diabetes or had elevated hemoglobin A1c levels. Participants reporting diagnoses of coronary heart disease, myocardial infarction, or angina were classified as having coronary heart disease, and reported diagnoses were similarly used to classify participants as having a history of stroke.
The In the Clinic® slide sets are owned and copyrighted by the American College of Physicians (ACP). All text, graphics, trademarks, and other intellectual property incorporated into the slide sets remain the sole and exclusive property of the ACP. The slide sets may be used only by the person who downloads or purchases them and only for the purpose of presenting them during not-for-profit educational activities. Users may incorporate the entire slide set or selected individual slides into their own teaching presentations but may not alter the content of the slides in any way or remove the ACP copyright notice. Users may make print copies for use as hand-outs for the audience the user is personally addressing but may not otherwise reproduce or distribute the slides by any means or media, including but not limited to sending them as e-mail attachments, posting them on Internet or Intranet sites, publishing them in meeting proceedings, or making them available for sale or distribution in any unauthorized form, without the express written permission of the ACP. Unauthorized use of the In the Clinic slide sets will constitute copyright infringement.
James P. Scanlan
James P. Scanlan, Attorney at Law
November 28, 2009
Article on disparities in control of cardiovascular disease and diabetes raises several measurement issues
McWilliams et al.(1) analyze changes in disparities in dichotomous and continuous measures of control of cardiovascular disease and diabetes from 1999 to 2006 and the effects of Medicare coverage on these disparities. Three aspect of their analysis warrant comment.
First, the authors analyze the size of disparities in control rates in terms of absolute differences between rates. The analysis is flawed for failure to consider the way absolute differences between outcome rates tend to change solely because of changes in the overall prevalence of the outcome -- as I have discussed with respect to similar flaws in the references nos. 8, 11, and 20-22 of the article ([2-7). See also Houweling et al. (8), which similarly recognizes the patterns of correlations between absolute differences and the overall prevalence of an outcome and the need to consider overall prevalence in appraising the meaning of standard measures of differences between outcome rates. For research of this nature to be of value, it is necessary to appraise disparities between outcome rates in terms of measures that are unaffected by the overall prevalence of the outcome, such as that described in references 9 and 10, which derives differences between means of underlying risk distributions based on outcome rates (9,10).
But, in any case, it is not useful to analyze disparities in terms of absolute difference or other measures of differences between outcome rates that are affected by the prevalence of the outcome without even addressing the measurement issues. Notably, in observing that they are building on further work that found persistent or widening healthcare disparities in recent years, the authors cite the National Healthcare Disparities Report. But the National Healthcare Disparities Report not only relies on relative differences between rates rather than absolute differences, but relies on whichever relative difference (in the favorable or the adverse outcome) is the larger. Thus, as explained in references 2, 6, and 7 (and their corrections), as well as 10-12, with regard to the appraisal of the directions of changes in the size of disparities over time, the approach of the National Healthcare Disparities Report tends toward reaching exactly the opposite conclusions of those reached by researchers who rely on absolute differences (allowing, of course, that meaningful changes may counter the patterns that are solely statistically driven)(10-12).
Second, the authors also analyze the size of disparities in terms of continuous measures, including differences between average systolic blood pressure levels and average hemoglobin A1c levels. Such an approach may well avoid the problems associated with the ways that absolute differences and other standard measures of differences between outcome rates tend to be affected by the overall prevalence of an outcome. Where available data permit, such an approach -- making use, as it does, of the actual distributions of risk factors rather than a derived difference between means of a hypothetical risk distribution -- may well be much superior to the approach discussed in references 9 and 10.
But the most useful way to measure differences between means is by deriving an effect size from the arithmetic difference between means divided by the standard deviation (either the pooled standard deviation or that of one group or the other). The authors instead simply discuss the changes in terms of changes in the arithmetic differences between means. The authors highlight what they term a "0.7%" reduction in the race or ethnic difference and a "0.5%" reduction in the education difference for subjects over 65 compared with those under 65. The former reduction was from an absolute difference between mean levels of 0.9 percentage points to one of 0.2 percentage points; the latter was from 0.6 percentage points to 0.1 percentage points. Thus, allowing that the standard deviation serving as the denominator in the effect size fractions may differ somewhat for the 45 to 64 group and the over-65 group (NHANES 2005 data, which happen to be at hand, show a standard deviation of 1.23 for the former group and 0.94 for the latter group), these figures suggest a very substantial reduction in the disparity. The magnitude of such reduction is hardly reflected in the "0.7%" and "0.5%" figures used by the authors, even if such figures reflect valid measures.
Third, while the authors generally rely on percentage point difference and percentage point changes in figures, they almost invariably describe an "x percentage point" change as an "x%" change. Whether or not the latter usage is even technically correct, most readers regard an "x%" change as a percentage change not a percentage point change. The authors themselves in endeavoring to be clear in their discussion of the implications of differential increases in insurance coverage refer to a "6.6 percentage point" difference in increases in coverage rates. The entire article would be much clearer if discussion of absolute differences between percents were invariably described as percentage point differences rather than percent differences (13).
1. McWilliams JM, Meara E., Zaslavsky AM, Ayanian JZ. Differences in control of cardiovascular disease and diabetes by race, ethnicity, and education, U.S. trends from 1999 to 2006 and effects of Medicare coverage. Ann Int Med. 2009;150:505-515.
2. Scanlan JP. Effects of choice measure on determination of whether health care disparities are increasing or decreasing. Journal Review May 1, 2007 (responding to Trivedi AN, Zaslavsky AM, Schneider EC, Ayanian JZ. Trends in the quality of care and racial disparities in Medicare managed care. N Engl J Med 2005;353:692-700, and two other articles in the same issue): http://journalreview.org/v2/articles/view/16107620.html
3. Scanlan JP. Understanding patterns of absolute differences in vaccination rates in different settings. Journal Review Apr. 22, 2008 (responding to Schneider EC, Cleary PD, Zaslavsky AM, Epstein AM. Racial disparity in influenza vaccination: Does managed care narrow the gap between blacks and whites? JAMA 2001;286:1455-1460): http://journalreview.org/v2/articles/view/11572737.html
4. Scanlan JP. First learn to measure healthcare disparities. Health Affairs Mar. 12, 2008 (responding to Casalino LP, Elster A, Eisenberg A, et al. Will pay-for-performance and quality reporting affect health care disparities? Health Affairs 2007;26(3):405-414):: http://content.healthaffairs.org/cgi/eletters/26/3/w405
5. Scanlan JP. Inclusion of healthcare disparities issues in pay-for -performance programs should await development of reliable means of measuring changes in disparities over time. Journal Review Feb. 16, 2008 (responding to Casalino LP, Elster A, Eisenberg A, et al. Will pay-for- performance and quality reporting affect health care disparities? Health Affairs 2007;26(3):405-414): http://journalreview.org/v2/articles/view/17426053.html
6. Scanlan JP. Understanding the ways improvements in quality affect different measures of disparities in healthcare outcomes regardless of meaningful changes in the relationships between two distributions of factors associated with the outcome. Journal Review Aug. 30, 2007 (responding to Sequist TD, Adams AS, Zhang F, Ross-Degnan D, Ayanian JZ. The effect of quality improvement on racial disparities in diabetes care. Arch Intern Med 2006;166:675-681): http://journalreview.org/v2/articles/view/16567608.html
7. Scanlan JP. Understanding patterns of correlations between plan quality and different measures of healthcare disparities. Journal Review Aug. 30, 2007 (responding to Trivedi AN, Zaslavsky AM, Schneider EC, Ayanian JZ. Relationship between quality of care and racial disparities in Medicare health plans. JAMA 2006;296:1998-2004):: http://journalreview.org/v2/articles/view/17062863.html
8. Houweling TAJ, Kunst AE, Huisman M, Mackenbach JP. Using relative and absolute measures for monitoring health inequalities: experiences from cross-national analyses on maternal and child health. International Journal for Equity in Health 2007;6:15: http://www.equityhealthj.com/content/6/1/15
9. Solutions sub-page of Measuring Health Disparities page of jpscanlan.com http://www.jpscanlan.com/measuringhealthdisp/solutions.html
10. Can We Actually Measure Health Disparities?, presented at the 7th International Conference on Health Policy Statistics, Philadelphia, PA, Jan. 17-18, 2008: Oral Presentation: http://www.jpscanlan.com/images/2008_ICHPS_Oral.pdf; PowerPoint Presentation: http://www.jpscanlan.com/images/2008_ICHPS.ppt
11. Measurement Problems in the National Healthcare Disparities Report, presented at American Public Health Association 135th Annual Meeting & Exposition, Washington, DC, Nov. 3-7, 2007: PowerPoint Presentation: http://www.jpscanlan.com/images/APHA_2007_Presentation.ppt; Oral Presentation: http://www.jpscanlan.com/images/ORAL_ANNOTATED.pdf; Addendum (March 11, 2008): http://www.jpscanlan.com/images/Addendum.pdf
12. Scanlan JP. Study illustrates ways in which the direction of a change in disparity turns on the measure chosen. Pediatrics Mar. 27, 2008 (responding to Morita JY, Ramirez E, Trick WE. Effect of school-entry vaccination requirements on racial and ethnic disparities in Hepatitis B immunization coverage among public high school students. Pediatrics 2008;121:e547-e552): http://pediatrics.aappublications.org/cgi/eletters/121/3/e547
13. Percentage Points sub-page of Vignettes page of jpscanlan.com: http://www.jpscanlan.com/vignettes/percentgepoints.html
Cardiology, Endocrine and Metabolism, Diabetes, Healthcare Delivery and Policy, Coronary Risk Factors.
Results provided by:
Copyright © 2016 American College of Physicians. All Rights Reserved.
Print ISSN: 0003-4819 | Online ISSN: 1539-3704
Conditions of Use
This PDF is available to Subscribers Only