West Coast Ballroom
The future meets the past: Applying features of retrospective study design to improve prospective comparative effectiveness studies (306706)
*Jennifer H Lindquist, Department of Veterans Affairs
Valerie Anne Smith, Duke University; Department of Veterans Affairs
Courtney Harold Van Houtven, Duke University; Department of Veterans Affairs
Keywords: Prospective study design, exact matching, distance function matching, comparative effectiveness, minimizing bias, average treatment effect
Rigorous evaluation of programs or policies when randomization is impractical relies on careful study design. We use retrospective design principles to prospectively evaluate treatment effects, addressing confounding and allowing collection of patient reported outcomes (PROs). The exemplar study compares a geriatric primary care model (GeriPACT) to a standard primary care model (PACT) in VA medical centers to evaluate differences in experience and quality of care. After exact matching GeriPACT and PACT patients on key variables obtained via electronic health record (EHR), up to 25 matches were selected via distance function. Recruitment consisted of enrolling a GeriPACT patient then a matching PACT patient, using pre-assigned priority categories that minimized the distance function. Among the first 207 dyads enrolled, all standardized baseline differences were <0.2, suggesting our approach successfully balanced covariates. This strategy improves retrospective-alone designs to evaluate treatment effects in comparative effectiveness studies. This design harnesses nationally collected EHR data coupled with collection of PROs, an important perspective missed in secondary data studies.