With increasing availability of electronic health data from EHRs, there has been expanding interest in using this data to identify populations for targeted health interventions. With a small proportion of patients using a majority of healthcare resources, the potential benefit of predicting high cost patients is clear, though early efforts have delivered few gains.
This project involves a manual review of a set of high-cost patients to characterize the ability of electronic health data to appropriately predict high utilization. Patient records will be reviewed for indications of declining health to provide an estimate of potential benefit for risk prediction. Information for prediction will also be characterized for its availability in structured electronic form. This project will be critical to inform one of the most invested areas of healthcare analytics.
Project can provide research opportunities for interested graduate students and post-docs.