Abstract
Background/Aims Defining a population denominator is a major challenge in practice-based research. Because health plan membership is unknown, population denominators must be determined by alternate methods, e.g., healthcare utilization. Such an approach may introduce bias since the resulting population may only represent a small portion of individuals who are affiliated with a healthcare organization. Moreover, healthcare utilization is often confounded by demographic and socio-economic factors, potentially compounding bias in the population estimates.
Methods The majority of HMORN member organizations have health plans, whose enrollment information is captured in the Virtual Data Warehouse (VDW) Enrollment file. The health plan enrollment provides a well-defined population denominator. In order to participate in multi-site research, HMORN member organizations without health plans, such as Essentia Health (EH), need to determine an enrollment proxy. The purpose of this pilot study was to evaluate a utilization-based enrollment proxy algorithm. Four HMORN sites with Epic electronic medical records (three sites with and one site without standard VDW enrollment files) participated in the study. The utilization-based algorithm was applied in two steps: 1) creating the base table from Clarity (Epic) database; and 2) creating the utilization-based enrollment proxy using predetermined selection rules.
Results EH implementation of the enrollment proxy for years 2002–2012 contained 390,000 “enrollment” periods for 378,000 unique patients. The EH algorithm was applied to the data from the three sites with health plans; the resulting enrollment proxy files were compared to the standard enrollment files. Patient demographic data (age, sex) and place of residence categorized by rural-urban commuting area (RUCA) codes (urban, large rural, small/isolated rural), as well as race and socio-economic measures (Census file), were used to characterize the deviations between two population denominators.
Conclusions The results of this study will inform HMORN researchers about the comparability of the patient populations between standard health plan enrollment and the utilization-based proxy. This study will also evaluate patient characteristics likely to affect utilization, and therefore, accuracy of the utilization-based proxy as a population denominator. It may also be useful for further development of the algorithm.
