Nationally representative MEPS data were used to examine the relationship of age, gender and insurance status to quality of care among Americans with diabetes and cardiovascular conditions. Persons with missing education or insurance status data were excluded from the analysis (6 percent each for the diabetes and cardiovascular condition populations). Older adults who reported being "uninsured"; were excluded from analysis because they are eligible for Medicare and because there were few respondents who indicated that they were uninsured (0.1 percent each for the diabetes and cardiovascular condition populations). This study focused on respondents ≥45 with self-identified diabetes (N=4,067) or self-identified cardiovascular conditions (N=17,636). Refer to Chapter 3.2 for additional information about the
The dependent variables were 17 qualities of care indicators for diabetes and 15 qualities of care indicators for cardiovascular conditions. The indicators were from MEPS and addressed effectiveness, timeliness (delays) and patient-centeredness of care (Table 3). Of the 17 diabetes indicators, 7 measured effectiveness of care, 5 measured timeliness of care and 5 measured patient-centeredness of care. Of the 15 cardiovascular indicators, 5 measured effectiveness of care, 5 measured timeliness and 5 measured patient-centeredness items.
Six of the seven diabetes effectiveness of care indicators asked respondents if they had received HbA1c testing, an eye examination, a foot examination, blood pressure testing, blood cholesterol testing and influenza shot. The seventh indicator was a composite measure that examined whether respondents had received HbA1c testing and an eye and foot examination. The five cardiovascular effectiveness of care indicators asked respondents if they had received dietary advice, exercise advice, blood pressure testing, blood cholesterol testing and an influenza shot. All indicators asked about patient experiences during the past year.
The five timeliness indicators were the same for diabetes and cardiovascular condition patients. Three indicators asked all respondents if they had delays in obtaining medical care, dental care or prescription medications. A fourth indicator was a composite measure that asked respondents if they had a delay in any of these three types of care. The fifth indicator was concerned with whether patients who needed care for illness or injury actually received care as soon as they wanted it. All items asked about patient experiences during the past year.
The five patient-centeredness indicators were the same for diabetes and cardiovascular condition patients. Four items asked all respondents if their provider listened carefully, explained things clearly, showed respect and spent enough time with them. The fifth indicator was a composite measure that asked respondents if their provider did any of these four things. All indicators asked about patient experiences during the past year.
The main independent variables of interest were age, gender and health insurance status in the past year. Age was stratified by midlife adults (45–64 years) vs. older adults (≥65 years). Gender was coded as "male"; or "female."; Five mutually-exclusive insurance status categories were based on self-reported insurance coverage, differentiated by midlife adults vs. older adults. All midlife adults were coded according to whether they reported having private health insurance in the previous year (including health insurance through an employer or union or a private source that was not employment-related); having only public insurance in the previous year (including Medicare, TRICARE, Medicaid and other public hospital/physician coverage); or being uninsured for all of the previous year. Older adults were coded according to whether they reported having Medicare and private health insurance in the previous year, or whether they had Medicare alone or Medicare in combination with only public insurance in the previous year. Older adults who reported being uninsured were excluded because this category had very few
Covariates were race/ethnicity (non-Hispanic White, non-Hispanic Black or Hispanic of any race); education level (<high school or >some college); income level (poor, near poor or low [representing <200 percent federal poverty level (FPL)]) vs. middle or high income (representing >200 FPL); and self-rated health status (excellent, very good or good vs. fair or poor).
A multivariate analysis was performed using SAS version 9.2 (SAS Institute, Cary, NC) and SUDAAN Release 10.0.0 (RTI International, Research Triangle Park, NC). All estimates were weighted to reflect the complex survey sampling design. Variance was computed using the Taylor linearization method, taking sample design features into account using SUDAAN. The individual was the unit of analysis. The unadjusted associations of all quality indicators were compared across age, gender and insurance status groups. Because of data limitations, reliable unadjusted estimates (with minimum cell size criteria of 100 observations or relative standard error >0.3) were not possible for all age groups, insurance status and gender strata, and unreliable estimates were suppressed in the unadjusted tables (identified by * in Table 13). For further examination of the association of gender and insurance status with quality of care, logistic regressions were used to estimate each indicator separately for each age group, while controlling for potential confounding factors. A fixed reference group was used to assess group differences in quality of care. For example, males were the referent group for gender differences. Private insurance was the referent group for insurance status differences among midlife persons. Medicare and private insurance were the referent groups for insurance status differences among older adults. The multivariate analysis controlled for education, race/ethnicity, income and health status. Multivariate results were reported as adjusted odds ratios (OR) with 95 percent confidence intervals.
Because this analysis examined 17 simultaneous dependent variables, drawn from the same sample of diabetes patients, and 15 simultaneous dependent variables, drawn from the same sample of cardiovascular condition patients, a Bonferroni correction was used to interpret p-values. Thus, at the alpha testing level of 0.05, only p-values <0.002 (0.05/17) were considered significant for diabetes patients, and only p-values <0.003 (0.05/15) were considered significant for cardiovascular patients.