Research in a fragmented healthcare system can be challenging when one seeks to follow patients within a care episode. Breast cancer treatment can continue for months or years and a care trajectory may reflect many decisions. Women may seek evaluations or specialty care at more than one medical center, either through physician referrals or "shopping" behavior. Claims data can track where patients are seen, but lack clinical detail. This study aims to better understand treatment location decisions among women seen at more than one healthcare facility. Methods: The OncoShare database combines EHR data from two large healthcare facilities in the same catchment area-a multisite community practice and an academic medical center-for all women treated at either site for breast cancer from 2000-2011. We use descriptive statistics and longitudinal modeling strategies to characterize treatment trajectories and define predictors of treatment facility decisions. Results: In a previous cross-sectional analysis of these data, the 16% of women who received treatment at both institutions had comparable prognostic factors, but far more diagnostic and treatment interventions than those treated at only one facility. We use a longitudinal analysis to characterize the care trajectories of these women, and test predictors of their treatment decisions such as geographic location, previous diagnostic tests performed, and timing of interventions. Conclusions: Preliminary results from this data revealed a marked difference in treatment intensity among women who sought care at both institutions, despite comparable prognoses. The richness of the EHR now allows the extraction of predictors such as patterns of medical care use before the breast cancer diagnosis.
