Lead/Presenter: Kelly Hunt,
COIN - Charleston
All Authors: Hunt KJ (COIN-Charleston), Davis ML (Charleston Health Equity and Rural Outreach Innovation Center), Shoaibi A (Charleston Health Equity and Rural Outreach Innovation Center, Medical University of South Carolina) Neelon B (Charleston Health Equity and Rural Outreach Innovation Center, Medical University of South Carolina)
Objectives:
To examine the extent to which access measures impact spatial differences in glycemic control among Veterans with diabetes across the U.S.
Methods:
A national retrospective cohort was constructed of 1,447,319 patients with type 2 diabetes receiving primary care within the VA during 2015 who had at least 1 valid HbA1c measure in 2015. Poor glycemic control was defined as a mean HbA1c ? 8%. Using a Bayesian hierarchical spatial model, we examined spatial patterns in poor glycemic control across 125 VAMC catchment areas in a random sample of 100,000 patients from out cohort. We modeled the odds of poor glycemic control while adjusting for demographic factors including race-ethnic group, comorbidities, medications and two access measures; namely, number of outpatient visits per year and distance from the nearest VA primary care facility.
Results:
Our results indicate major variability in glycemic control across catchment areas even after adjustment for access measures. Prevalence of uncontrolled diabetes ranged from 17.5% to 32.9% across catchment areas. Spatial precision estimates increased from 12.3 (8.3, 17.7) in a model containing only catchment area, to 19.0 (11.4, 30.9) when demographics, comorbidity and medication use was included, to 20.3 (12.1, 33.4) in our model that included access measures. In our final model hot spots of poor glycemic control were observed in Texas, Maine, Oregon and California, while spots of better than average control were found in Florida, Michigan, Missouri, Kansas and North Carolina. Patients within catchment areas with poor control had up to 1.29 (1.08, 1.54) times the odds of having uncontrolled diabetes as those within a catchment area with average control. Patients with fewer than 4 outpatient visits in the year were more likely to have uncontrolled diabetes than patients with 24 or more outpatient visits per year [OR = 1.40 (1.31-1.51)].
Implications:
Regional variation in poor diabetes control is evident among this national sample of veterans after adjustment for demographics, comorbidity burden, medication use and access metrics.
Impacts:
This analysis allows the VA to target resource allocation and behavioral interventions not only at the Veteran level, but at the geographic contexts in which Veterans live.