Identifying Patterns of Geographic Clustering in VA Enrollment and Active Utilization. National Rural Evaluation Center (NREC)

Bollinger M, Mader M, Hudson TJ, Wong ES, Finley EP, Fortney JC, Pyne JM, Drummond KL, Abraham TH, Townsend J, Lee J, Batten A, Bosworth HB, Boyko EJ, Littman A. Identifying Patterns of Geographic Clustering in VA Enrollment and Active Utilization. National Rural Evaluation Center (NREC). Poster session presented at: VA Rural Health State of the Science Conference; 2016 Sep 12; Washington, DC.

Objectives: VA rates of enrollment and healthcare utilization vary spatially and demographically, influenced not only by individual patient characteristics but also by the characteristics of communities and the placement of VA facilities. Understanding the spatial patterning of enrollment and healthcare utilization is important information for planners and policymakers and helps identify areas where targeted interventions may be needed to either increase or reduce capacity. The objectives of this presentation are to (1) discuss the methodology used to develop our estimates of enrollment and utilization, and (2) to share the results of the spatial analyses we conducted. Methods: VA administrative data and the American Community Survey (ACS) (2010-2014) were used to estimate enrollment and utilization. County-specific enrollment was calculated as VHA enrollees divided by the number of Veterans in each county from the ACS. County-specific utilization rates were calculated as active users (i.e., VA use in the last 2 years) divided by the number of enrollees. To account for variation due to differences in age and sex, we calculated age- and sex-standardized rates, referred to as Standardized Enrollment Ratio (SER) and a Standardized Utilization Ratio (SUR). Finally, we assessed geographic clustering of the SER and SUR using the global and local Moran's I statistic. All results were mapped in a geographic information system (GIS) with ARCGIS 10.2. Results: Both enrollment and utilization had strongly positive (Moran's I > 0.6) autocorrelation indicating substantial geographic clustering. High enrollment and utilization clustering were observed in the upper mid-West and Mid-Atlantic States. Low enrollment clustering was observed in the along the coastal areas of the U.S while clustering of low healthcare utilization was seen in various parts of the country but especially in the North Atlantic states - primarily in rural areas. Implications: Enrollment and utilization varies geographically by rurality. Moran's I statistics, combined with GIS, provided important context for visualizing enrollment and utilization. Next steps include determining which community-level factors are associated with the observed patterns using Bayesian multilevel spatial modeling. Impacts: Spatial data analyses represent an important tool for ensuring VA provides needed services to Veterans. This study suggests that spatial analyses have the potential to provide greater understanding and more accurate investigations of enrollment and utilization rates particularly as they relate to social determinants of health.