Background/Significance: Diabetes mellitus is a chronic and progressive disease that causes significant morbidity and mortality and increases health care utilization and costs in both Veteran Administration (VA) and non-VA settings. 1. Diabetes and its complications are more prevalent in minority populations. Black Americans have two-fold increased age adjusted rates of diabetes, are more likely to develop and experience greater disability from diabetes complications compared to White Americans. 1. Black Americans with diabetes have higher rates of retinopathy, end-stage renal disease, lower limb amputations, and overall death rates. 2. Therefore, diabetes is a significant public health problem and Black American patients have disproportionately higher morbidity and mortality than their White American counterparts.
Several factors have been postulated to explain the disproportionately higher morbidity and mortality from diabetes in Black Americans and these include their mistrust of the health care system. 3. It is thought that distrustful patients are less likely to seek routine medical care, take prescribed medications consistently, adhere to treatments recommendations, and maintain continuity with health care providers and health care systems. 4. Recent studies show that Black Americans are less trusting of physicians and the health care system. 5. However, little is known about the association between trust and diabetes outcomes and whether distrust of physicians and the health care system contributes to the observed racial/ethnic differences in diabetes outcomes.
Theoretical Framework: The conceptual and theoretical framework of this study is the revised behavioral model of health services use (Andersen 1974, 1968, 1983, 1995). The model posits that people's use of health services is a function of their predisposition to use services, factors that enable or impede use, and their need for care (Andersen 1995). Trust in physicians and the health system falls under health beliefs (attitudes toward health services), which is one of the predisposing factors that is thought to predict health services utilization and health outcomes. Thus, people with high levels of trust in physicians and the health care system are expected to have more effective access, appropriate health utilization, and better health outcomes. The model has been revised to include veteran-specific variables such as level of service entitlement, period of service, duration in the VA system, and disability status and to measure both health services use and health outcomes.
Research Design and Methods: This is a prospective cohort study with five hypotheses organized under their specific aims as follows:
Specific Aim #1: Determine racial/ethnic differences in trust in physicians and mistrust of the health care system among veterans with Type 2 Diabetes.
Hypothesis #1: There is a difference in mean scores on the general trust in physician scale (GTIPS) between White and Black American veterans with Type 2 diabetes.
Hypothesis #2: There is a difference in mean scores on the Health Care System Distrust Scale between White and Black American veterans with Type 2 diabetes.
Specific Aim #2: Determine the predictive power of trust in physicians and mistrust of the health care system on personal health practices and health outcomes in a prospective cohort of veterans with Type 2 Diabetes
Hypothesis #1: Controlling for predisposing, enabling, need, and veteran-specific factors, diabetic veterans with lower trust scores or higher mistrust scores will be less likely to keep office appointments, take prescribed medications, and adhere to diabetes self-management recommendations after 12 months of follow-up.
Hypothesis #2: Controlling for predisposing, enabling, need, and veteran-specific factors, diabetic veterans with lower trust scores or higher mistrust scores will have higher mean hemoglobin A1C, blood pressure, and LDL cholesterol levels after 12 months of follow-up.
Hypothesis #3: Controlling for predisposing, enabling, need, and veteran-specific factors, diabetic veterans with lower trust scores or higher mistrust scores will be less likely to accept influenza vaccination after 12 months of follow-up.
Study site \& Subjects: Patients will be recruited from the Charleston VAMC. Equal number of White and Black American veterans aged 18 years and older with Type 2 Diabetes will be recruited. Race/ethnicity will be based on self-report. The diagnosis of type 2 Diabetes as well as health utilization and diabetes-specific health outcomes will be obtained from the VA electronic medical records system (CPRS). There are approximately 6,961 patients with Type 2 Diabetes at this site, of which 49.1% (3,417) are White Americans, 31.5% (2,189) are Black Americans, and 19.4% (1,355) are Hispanic or other. Approximately 97.5% are men and 90% are aged 50 years or older.
Sample size calculation:
Specific Aim #1: Sample Power V2.0 (SPSS) was used for sample size calculation based on the convention outlined by Cohen6. Overall experiment wise error was held to ?=0.05, and power to 80% using medium (0.25) effect sizes. Correction for multiplicity of tests (2 tests for primary hypotheses) involved using ?=0.025 (0.05/2). This yielded 125 patients per group. In addition, the sample was inflated to account for an estimated 20% attrition at 1 year of follow-up (death, relocation, or loss to follow-up). No more than 150 eligible patients need to be enrolled per group. Thus, 300 patients (150 Whites and 150 African Americans) will be recruited.
Specific Aim #2: The sample size determination for a reliable regression equation offered by Stevens7 is 15 subjects per predictor variable. Using this standard, a sample size of 300, as determined above, would allow the inclusion of 20 predictor variables. Because none of the hypotheses for Specific Aim 2 exceed 20 predictor variables, a sample of 300 will be adequate.
Survey Instruments: The GTIPS4 is a valid and reliable 11-item measure of general trust in physicians and the Health Care System Distrust Scale is a valid and reliable 10-item measure of mistrust of the health care system. Both instruments have been validated in Black and White Americans.
Statistical Analysis Plan: Descriptive statistics will be used to describe the characteristics of participants in the study.
Specific Aim #1: Mean scores on the trust and mistrust scales at baseline will be compared between White and Black Americans with the two-sample t-test and similar comparisons will be made while controlling for covariates (predisposing, enabling, need, and veteran-specific factors) using Analysis of Covariance (ANCOVA).
Specific Aim #2: Multiple linear regression will be used to test the effect of mean trust/mistrust scores on health utilization and mean hemoglobin A1C, blood pressure, and LDL cholesterol after 12 months of follow-up controlling for covariates. Similarly, multiple logistic regression will be used to test the effect of trust/mistrust on acceptance of the influenza vaccine controlling for covariates. STATA V8.0 will be used for data analysis and all tests will be two-tailed with overall p=0.05 for each hypothesis.
