A Diabetes Report Card for the United States: Quality of Care in the 1990s FREE

–The Editors

Diabetes, a typical example of chronic disease, currently affects 16 million people in the United States, causes considerable morbidity and mortality, and costs the nation almost $100 billion per year (1–2). Fortunately, several efficacious treatment strategies to prevent or delay diabetes complications have emerged during the past decade, including control of glycemia, lipids, and hypertension; early detection and treatment of diabetic retinopathy, nephropathy, and foot disease; therapy with aspirin and angiotensin-converting enzyme inhibitors; and influenza and pneumococcal vaccines (3–12). Many of these treatments are cost-effective (13–17). However, their implementation in the United States remains suboptimal and varied (18–22). There is considerable pressure on U.S. health care systems to improve this situation and to deliver high-quality care while controlling costs. Influential bodies, such as the Institute of Medicine, have recently emphasized the need for strategies to improve the current quality of all medical care in the United States (23). Federal agencies are also increasingly responding with initiatives to address quality of care (24–25).

One powerful quality initiative in the United States is the Diabetes Quality Improvement Project (DQIP), which is designed to influence the care of millions of U.S. patients with diabetes (26). The standard measures of quality of diabetes care proposed by DQIP were incorporated into the National Committee for Quality Assurance (NCQA) Health Plan Employer and Data Information Set. In contrast to guidelines or clinical goals for individual care, DQIP measures are designed to assess the performance of health care systems for a population, and they offer a wayto make comparisons across health care systems. Several public and private health care systems, including the Indian Health Service (27), the Veterans Administration, the U.S. Department of Defense, and numerous managed care organizations, have adopted these indicators.

Currently, there is no national reference for assessing the quality of diabetes care using a standard set of measures. Such a reference could help health care systems using DQIP measures to compare their own performance against population norms rather than the norms of other health systems. This reference could be a benchmark for assessing population changes in quality of diabetes care after implementation of national quality improvement initiatives recommended by the Institute of Medicine and NCQA. Furthermore, the use of a standard set of measures to assess the quality of diabetes care at a national level can facilitate international comparison. For this report, we used U.S. national data to provide a reference and benchmark of the quality of diabetes care as measured by the DQIP indicators.

We used two federally funded, nationally representative population-based surveys, the Third U.S. National Health and Nutrition Examination Survey (NHANES III) (1988–1994) and the Behavioral Risk Factors Surveillance System (BRFSS) (1995). Data from these two surveys were analyzed separately. Some measures came exclusively from NHANES III, and others came from BRFSS. We included adults 18 to 75 years of age who reported receiving a previous diagnosis of diabetes from a physician. Women with gestational diabetes were excluded. The two survey groups were similar in demographic characteristics, household income, health insurance coverage, and prevalence of diabetes. In NHANES III compared with BRFSS, more persons did not have a high school education (41.5% vs. 27.1%) and fewer persons used insulin (30.9% vs. 39.8%) (Table 1).

The methods for NHANES III are described elsewhere (28). Briefly, the NHANES III survey used a nationally representative sample of the civilian noninstitutionalized population obtained through a complex multistage cluster sampling design, with oversampling of non-Hispanic black persons, Mexican-American persons, and elderly persons. During a household interview, data were collected on sociodemographic characteristics and medical and family history. Within 4 weeks, this was followed by a clinical examination at a mobile examination center. The procedures for blood collection and processing have also been described elsewhere (29–30). Low-density lipoprotein (LDL) cholesterol level was calculated by using the Friedwald equation for persons who fasted more than 8 hours (n = 302). Data from NHANES III were self-reported (demographic and clinical variables) or were obtained during clinical examination (hemoglobin A_1c level, cholesterol level, triglyceride level, and blood pressure). The survey included 16 705 participants 18 to 75 years of age, 1026 of whom reported receiving a diagnosis of diabetes from a physician before the survey.

The BRFSS is an annual random-digit telephone survey of state population–based samples of the civilian noninstitutionalized population of adults (≥ 18 years of age) in each of the 50 states and the District of Columbia. Its purpose, methods, and data analyses have been described in detail elsewhere (31). A diabetes-specific BRFSS module was used to collect data on clinical characteristics and preventive care practices from respondents with diabetes. Data from BRFSS were self-reported (dilated eye examination, lipid test, foot examination, and demographic and clinical variables). The survey included 103 929 participants 18 to 75 years of age, 3059 of whom completed the diabetes module and reported receiving a diagnosis of diabetes from a physician. We used the 1995 BRFSS so that the time frame would correspond with NHANES III.

We assessed the quality of diabetes care using indicators from the DQIP measurement set for which data were available. The DQIP began under the sponsorship of a coalition that included the American Diabetes Association, the Foundation for Accountability, the Health Care Financing Administration, and NCQA. The American Academy of Physicians, the American College of Physicians, the Veterans Administration, and the Centers for Disease Control and Prevention later joined the coalition. A committee of experts in diabetes care was responsible for developing the DQIP measure set. The DQIP classified the proposed indicators into three categories: accountability indicators, quality improvement indicators, and indicators under field-testing. The accountability measures, which are well grounded in evidence, have received consensus support from the scientific and medical community and have been extensively field-tested in a variety of health care settings. These measures are used to compare health systems and plans or providers. The quality improvement measures are not considered appropriate for comparing systems and plans or providers but were recommended by the NCQA for assessing internal performance. Table 2 shows all of the analyzed quality indicators and their data sources. The DQIP measures are a combination of process and outcome measures. In this paper, we use DQIP terminology and DQIP accountability and quality improvement measures.

All analyses were conducted by using SAS software (SAS Institute, Inc., Cary, North Carolina) for data management and SUDAAN (Research Triangle Institute, Research Triangle Park, North Carolina) to account for the complex sampling scheme, the unequal probability of selection, the oversampling of certain demographic groups, and nonresponse factors (32–33). Level of care was estimated by the percentage, with corresponding 95% CIs, of respondents who reported each preventive care practice. We estimated prevalence by groups of sex, age, ethnicity, education, health insurance, type of treatment, and duration of diabetes. Multiple logistic regression and computation of predictive margins were used to estimate the probability of receiving DQIP indicators when we controlled for all other independent variables. Predictive margins are a type of direct standardization in which the predicted values from the logistic regression models are averaged over the covariate distribution in the population (34). Predictive margins and their standard errors from logistic regression models are provided by the current version of SUDAAN (33). Taylor linearization is used in SUDAAN for calculating these standard errors. Odds ratios are usually used to display the result from logistic regression models. However, predictive margins are easier to interpret than odds ratios, and they do not require designating one of the groups as the referent group.

Among adults 18 to 75 years of age with diabetes, 28.8% had had hemoglobin A_1c levels tested in the past year, and 18.0% had a hemoglobin A_1c level greater than 9.5%. Biannual lipid testing was done in 85.3% of participants, but only 42.0% had an LDL cholesterol level less than 3.4 mmol/L (<130 mg/dL). Only 65.7% had blood pressure less than 140/90 mm Hg, 63.3% had an annual dilated eye examination, and 54.8% had an annual foot examination (Table 3).

The distribution of the quality improvement measures (hemoglobin A_1c and LDL values and systolic and diastolic blood pressure) are shown in the Figure. The median hemoglobin A_1c level was 7.5% (CI, 7.2% to 7.9%). Less than half of the population (42.9%) had hemoglobin A_1c levels less than 7%, and 14.9% had hemoglobin A_1c levels greater than or equal to 10%. The median systolic blood pressure was 130.9 mm Hg (CI, 129.6 to 133.6 mm Hg). One fifth of participants had a systolic blood pressure of 140 to 159 mm Hg, and 9.1% had a systolic blood pressure of 160 to 179 mm Hg. The median LDL level was 3.5 mmol/L (CI, 3.4 to 3.8 mmol/L) (134.4 mg/dL [CI, 130.1 to 144.9 mg/dL]). More than half of the participants had an LDL level greater than 3.4 mmol/L (>130 mg/dL), and almost one fourth had an LDL level greater than 4.1 mmol/L (>160 mg/dL).

Regarding other preventive care measures, which may be considered as future quality indicators, only 45.7% of persons had received the influenza vaccine in the past year and only 26.5% had ever received the pneumococcal vaccine (Table 3). Only 38.0% monitored their blood glucose level at least once daily, and 58.0% had had a dental examination in the past year.

In unadjusted analysis, a higher proportion of persons at least 65 years of age received biannual lipid tests and annual dilated eye examinations compared with persons 18 to 44 years of age (88.9% vs. 74.0% [P = 0.001] and 70.2% vs. 50.4% [P = 0.009], respectively). Compared with non-Hispanic white persons, non-Hispanic black persons were more likely to have a hemoglobin A_1c level greater than 9.5% (29.9% vs. 15.7%; P = 0.002) and were less likely to have controlled blood pressure (<140/90 mm Hg) (56.7% vs. 66.3%; P = 0.04). Persons with education beyond high school were more likely to have controlled blood pressure than those who did not have a high school education (74.3% vs. 59.0%; P = 0.06). Insured persons were less likely than uninsured persons to have a hemoglobin A_1c level greater than 9.5% (17.5% vs. 34.8%; P = 0.05) and were more likely to have lipid tests and annual eye examinations (86.9% vs. 74.1% [P = 0.008] and 66.5% vs. 43.2% [P = 0.006], respectively). In the BRFSS data, persons who did not see a physician because of cost (although they had health coverage) were less likely to have annual dilated eye examinations (46% vs. 67%; P = 0.01) and were less likely to have their cholesterol checked (76% vs. 87%; P = 0.01) compared with those who did not have cost considerations. Compared with nonusers, persons who used insulin were more likely to have a hemoglobin A_1c level greater than 9.5% (27.5% vs. 15.1%; P = 0.002) and to have received an annual dilated eye examination (72.2% vs. 57.6%; P = 0.006) and foot examination (67.3% vs. 47.1%; P = 0.003). Persons who used insulin were not as likely to have LDL levels less than 3.4 mmol/L (<130 mg/dL) (18.8% vs. 43.9%; P = 0.07) and were more likely to self-monitor blood glucose level at least once per day (60.3% vs. 23.4%; P = 0.001).

We obtained results after controlling for age, sex, ethnicity, education, health insurance, insulin use, and duration of diabetes (Table 4). Insured persons were more likely than uninsured persons to have had a dilated eye examination (66.5% vs. 43.2%; P = 0.001) and were less likely to have a hemoglobin A_1c level greater than 9.5% (17.7% vs. 26.9%; P = 0.2), although the latter difference did not reach statistical significance. Persons who used insulin were more likely than nonusers to have had an annual dilated eye examination (72.2% vs. 57.6%; P = 0.001 for the difference) and foot examination (67.3% vs. 47.1%; P = 0.001); however, those who used insulin were also more likely to have poor glycemic control (24.2% vs. 15.5%; P = 0.02). Persons 65 to 75 years of age were more likely than persons 18 to 44 years of age to have biannual lipid testing (90.6% vs. 71.9%; P = 0.001), annual dilated eye examination (69.7% vs. 53.4%; P = 0.003), and controlled blood pressure (48.1% vs. 93.1%; P = 0.001). Finally, non-Hispanic black persons were more likely than non-Hispanic white persons to have an elevated hemoglobin A_1clevel (27.1% vs. 15.9%; P = 0.002) and poor blood pressure control (56.6% vs. 64.8%; P = 0.08).

Using DQIP, a standard set of measures, we found that a large proportion of persons in the United States receives less than optimal diabetes care. Overall, 18.9% of participants had a hemoglobin A_1c level greater than 9.5%, 58.0% had poor lipid control, 34.3% had poor blood pressure control, 36.7% did not receive an annual dilated eye examination, and 45.2% did not have a foot examination. When applied to the estimated 7.9 million persons 18 to 75 years of age in the United States who have diagnosed diabetes, these statistics translated to an estimated 1.4 million persons with hemoglobin A_1c levels greater than 9.5%, 2.7 million with uncontrolled blood pressure, 2.9 million without annual dilated eye examinations, and 3.6 million without annual foot examinations. Diabetes care in the United States can be vastly improved, and such improvement may yield substantial health benefits (35).

Clinical trials indicate that decreasing hemoglobin A_1c levels by 1 percentage point would reduce microvascular complications by 25% to 30% (3–4) and that reducing blood pressure by 10 mm Hg would decrease macrovascular and microvascular complications and mortality rates by 35% (6). In addition, good lipid control can reduce the risk for coronary heart disease by 25% to 55% and the risk for death by 43% (8, 36). These results, together with our findings, suggest that improvement in quality of care can substantially improve the health of persons with diabetes nationwide.

Previous reports have indicated that quality of diabetes care in the United States is poor (18–19, 22). Harris recently used a limited set of variables to describe the quality of national diabetes care (37–38). Our study, which involved two large national data sets, adds to the existing literature through specific and comprehensive use of DQIP indicators to provide a practical benchmark for diabetes quality of care process and outcome indicators. Our study also provides a benchmark for quality of diabetes care in several important subgroups, such as age, sex, and health insurance status.

We found that receipt of preventive care varied consistently by two factors across groups. First, not having health insurance was highly associated with poorer status for several indicators. Previous reports have indicated marked ethnic differences in health insurance coverage and the risk for diabetes complications (39), and uninsured adults report greater unmet health needs than insured adults (40). Second, consistent with previous studies, insulin users in our study were more likely to receive preventive care (41–42). This is probably because insulin use is linked to disease severity, and providers are more likely to follow recommendations when managing patients with severe disease. Although complications are more common among persons using insulin, the long-term risk for complications and the number of patients not taking insulin are both substantial. Health care providers and patients must better address preventive care if insulin is not used.

We found that non-Hispanic black persons had worse glycemic control than other ethnic groups; other reports also support this finding (43–45). Of interest, we found that non-Hispanic black persons did not differ significantly from participants from other ethnic groups in receipt of preventive services. Future health services research may need to identify reasons for poor glycemic and blood pressure control in black persons independent of access to care.

Several new measures (influenza vaccine, pneumococcal vaccine, self-monitoring blood glucose, and dental examination) may be considered for future national quality indicator sets. Many reports have demonstrated the efficacy of influenza vaccination among persons with diabetes. In a study by Nichol and colleagues, hospitalization decreased by 64% and respiratory disease decreased by 32% among vaccinated patients (46–47). In our study, only 45% of participants had received the influenza vaccine. Reports have also demonstrated an association between diabetes and periodontal disease and missing teeth (48–51). These oral health problems can also complicate overall diabetes management and increase the risk for poor glycemic control (50). In our study, only 58% of persons with diabetes had visited a dentist in the past year.

The DQIP indicators were initially developed for managed care organizations, which are becoming the main source of health care services for persons with diabetes in the United States. Even in 1990, 95% of health care benefits provided by U.S. employers involved some form of managed care (52–53). Clearly, these indicators have considerable potential to accelerate the translation of research into clinical practice. Decisions to use specific values and distributions of hemoglobin A_1c levels, blood pressure, and LDL cholesterol levels as quality-of-care indicators represent a migration to measurement of intermediate outcomes. Such decisions are prompted by convincing research linking management of these indicators to reduced risk for microvascular and macrovascular complications of diabetes (3–7).

Our study has several limitations. Data were self-reported, which can result in recall bias. However, the BRFSS responses related to diabetes status and sociodemographic characteristics have a good to excellent reproducibility (54–55) and validity (56). Reports have shown that the rates of a performance measure obtained from a member survey are usually higher than rates based on administrative data or abstracts of medical records (57). Even with this limitation, however, we still found suboptimal levels of care. Furthermore, we reported the distribution of hemoglobin A_1c level, blood pressure, and lipid levels based on examination and not on chart review. We used two different population-based surveys for this study, but sociodemographic characteristics and other important factors were similar in each. Practices have probably changed since these surveys were conducted in 1988–1994 and in 1995, but the extent of this change is unknown. This is especially important because many clinical trials, such as the Diabetes Control and Complication Trial and the United Kingdom Prospective Diabetes Study, released their findings in the late 1990s. The major value of our data is as a benchmark of the quality of diabetes care before the results of these major trials were released. Thus, the impact of national programs stimulated by these trials can be assessed.

Although health insurance was associated with several preventive measures, we could not explore the association further. The NHANES III and BRFSS surveys used similar yes/no questions to determine whether participants had health coverage, but they did not request more detailed information. In addition, neither survey assessed nephropathy status, self-management education, interpersonal care from provider, or participant satisfaction, all of which are important components of care that should be considered for inclusion in future national surveys.

Many U.S. adults with diabetes are receiving suboptimal care. We need to identify mechanisms at the patient, provider, and health care system levels to improve the quality of diabetes care (58). Health insurance was a strong correlate of several indicators, further supporting a large body of evidence indicating that more thorough coverage increases use of preventive care. Our data on the quality of diabetes care provide a national benchmark and can serve as a basis for focused clinical and public health efforts to reduce diabetes complications. These data also offer a normative set of the distribution of quality of care in the United States that will enable various organizations to compare and assess their own performances. When these methods are used, it may be possible to identify best care practices, and dissemination of these practices across settings could follow. When quality of diabetes care is described by using a standard set of measures, such as DQIP, future international comparison may be facilitated. It is clear from these data that a wide gap exists in the United States between our knowledge of effective diabetes interventions and their implementation in practice. This gap must be narrowed, and the recent emphasis placed on improving quality of care and translation research (23–25) is encouraging.