C-C1-03: Using the Electronic Health Record to Create Population Denominators: Optimization Using Insurance Enrollment

Abstract

Background and Aims: The HMORN includes research centers that are part of integrated delivery systems where the practice and insurance entities are independent (e.g., Geisinger, Marshfield). Rules for defining population denominators at these centers require acknowledging that patients in the practice population may not be members of the insurance entity. Methods for defining a population denominator for primary care patients from the electronic health record (EHR) can be validated in the subset of patients who are in the primary care system and members of the insurance entity. The aim of this study was to validate and optimize a method for calculating population denominators from the EHR. Methods: We proposed a method for defining population denominators from the EHR data at HMORN 2009. This method was based on describing utilization patterns of primary care patients’ overtime. For this study, the cohort was limited to the subset of Geisinger primary care patients who were enrolled in the Geisinger Health Plan. Survival analysis was used to minimize bias in person-time estimates and incidence estimates. The aims were to identify optimal times from insurance enrollment to first utilization and time from final utilization to insurance disenrollment. Since the results are likely dependent upon gender and age (i.e. a typical gap in utilization will tend to be longer for young males as compared to older males), the analyses were compared across patient demographics. Results: To define cohort entry, the time between initial encounter and insurance enrollment was estimated using survival analysis. EHR enrollment was considered active until the patient failed to have any encounters with a primary care clinic for an age/gender specific cutoff of time. If the patient became inactive, the end date was imputed forward in time based on optimization from survival analysis using the insurance enrollment. These estimates were used to create an age/gender specific algorithm for calculating population denominators from the EHR. Conclusions: EHR utilization can be used to define population denominators. Validation of the proposed method was conducted by comparing results to insurance enrollment spans. This application is limited to clinical areas where there is evidence of relatively complete capture.