Predicting long-term healthcare spending changes using longitudinal modeling methods in Medicare patients
Current approaches to predicting healthcare costs generally rely on a single composite value of spending, have modest predictive accuracy, or focus on short time horizons. By contrast, examining patients’ dynamic patterns of spending over longer periods may better discriminate between patients who have rising or falling spending and help target interventions to those at greatest need. We conducted a study using a random, nationwide sample of fully-insured Fee-For-Service beneficiaries at least 65 years of age. We used group-based trajectory modeling in medical and prescription claims data to classify patients by their spending patterns over a two-year period.
This project was funded by the National Institute for Health Care Management (NIHCM) (Lauffenburger PI).