We identified over 10,000 insulin-treated patients at the three participating medical centers. We randomly selected approximately 3000 subjects from this list and invited 589 eligible subjects to participate. We enrolled 359 (61%) subjects, but subsequently excluded 21 with incomplete data, leaving a cohort of 338 for the analysis. Their mean age was 65.1 ± 9.7 years, 96% were men, and 59% were married. Based on self-description, the cohort was comprised of 226 (67%) non-Hispanic white, 72 (21%) Hispanic, and 35 (10%) African-American subjects. Over two-thirds of the subjects had at least one microvascular disease complication and an equivalent number had a macrovascular disease complication. Although average daily insulin doses were substantial (66 units), only one-third of the subjects were concurrently treated with an oral hypoglycemic medication. Most subjects had an elevated baseline A1c value suggesting poor glycemic control, including 99 (29%) with a value between 7.0% and 8.0%, and 148 (44%) with a value ≥ 8.0%. The median level of activity was 7 met-hours of activity per day (equivalent to 2.5 hours of light home activities plus 0.5 hours of moderate walking). Sixty-two percent had a BMI ≥ 30 and 22.2% were current smokers. However, average lipid and blood pressure measurements were close to the target values recommended by the American Diabetes Association (ADA) [30
Clinical and socioeconomic characteristics stratified by race/ethnicity are presented in Table . African-American subjects had the highest A1c and were most likely to have poor glycemic control. Fifty-one percent of African Americans had an A1c ≥ 8% compared to 49% for Hispanics and 40% for non-Hispanic whites (p = 0.27). Mean ± SD A1c levels were 7.9 ± 1.4% for non-Hispanic whites, 8.2 ± 1.6% for Hispanics, and 8.8 ± 2.9% for African Americans (P = 0.05). We found significant differences in the daily units of insulin, with non-Hispanic whites receiving 70.6 ± 48.8 units compared to 58.4 ± 32.6 units for Hispanics and 53.1 ± 36.2 units for African Americans (p < 0.01). However, we found no differences between minorities and non-Hispanic whites in the daily number of insulin injections, the number of different insulin preparations used, or the use of oral hypoglycemic medications. Body mass index, amount of exercise, smoking status, lipid levels, and blood pressures were also similar between groups.
Sociodemographic and clinical features by race and ethnicity
We also evaluated potential barriers to care. African-American subjects considered themselves less disabled for work (Table ). Hispanic subjects had the most dependents, and reported the highest psychosocial barriers with respect to language preference, education, depression, diabetes knowledge, and performance on the MMSE (Table ). African-Americans perceived the fewest problems with glycemic control, had the fewest negative attitudes about diabetes, and had the highest self-ratings for self-care abilities and dietary adherence. They also tended to have the strongest convictions about the importance of self-care and the fewest perceived barriers to exercise.
Psychological features by race and ethnicity*.
We found that higher depression scores (r = 0.11, p = 0.049), greater work hours per week (r = 0.17, p = 0.002), greater number of household dependents (r = 0.13, p = 0.023), being employed (p = 0.004), and age (r = -0.21, p = 0.0001) were significantly associated with a higher baseline A1c. After adjusting for these covariates with a multivariate linear regression analysis, baseline A1c was significantly higher for African-American subjects (+0.93%, p = 0.002), though not for Hispanics (+0.25%, p = 0.29), compared to non-Hispanic whites.
We stratified daily insulin dose by race and ethnicity and level of glycemic control (Table ). We found that unadjusted daily insulin units increased monotonically with A1c in non-Hispanic whites, but not in Hispanics or African Americans. We found the most significant racial/ethnic differences in insulin doses among subjects with poorly controlled diabetes (A1c ≥ 8.0%), with non-Hispanic whites receiving approximately 22 daily units more than Hispanics and 26 daily units more than African Americans (p < 0.01). At follow-up, the treatment patterns remained essentially the same. Overall, 63% of each minority group and 70% of the non-Hispanic whites returned for the 26-week follow-up visit. For non-Hispanic whites, Hispanics, and African Americans, the mean ± SD daily insulin units were 71.1 ± 51.2, 63.1 ± 36.9, and 51.4 ± 26.6, respectively (p = 0.14). Among subjects with a baseline A1c ≥ 8.0%, the mean ± SD daily insulin units at follow-up also differed significantly (p = 0.04): non-Hispanic whites = 80.3 ± 51.3, Hispanics = 63.0 ± 34.0, and African Americans = 47.8 ± 23.1.
Daily insulin units by baseline hemoglobin A1c and race/ethnicity
We found that race/ethnicity, BMI, baseline A1c, and use of any other oral hypoglycemic medication were significant predictors of daily insulin units on multivariate linear regression analyses (Table ). African-Americans received an insulin dose that was an average of 17.8 units less (P = 0.01) and Hispanics an average of 10.5 units less (P = 0.04) than non-Hispanic whites. No significant interactions were observed. When the adjusted model was limited to those with baseline A1c ≥ 8%, African-Americans received an insulin dose that was an average of 26.5 units less (P = 0.01) and Hispanics an average of 15.9 units less (P = 0.03) than non-Hispanic whites. These differences persisted after adjusting for practice site.
Regression model predicting daily insulin units.