Trajectories of Medical Costs Among Patients with Chronic Diseases from Health Plan and Provider-Delivered Care Management Programs: Four Years Follow-Up Results Using Bayesian Adaptive Splines
*Hsiu-Ching Chang, BlueCross BlueShield of MI
Keywords: Health policy, Provider-delivered care management, Physician organizations, Bayesian semiparametric, Adaptive knot selection, RJMCMC
In 2010, a Midwest health plan with an existing care management program (HPDCM) implemented a pilot study to assess care management delivered by providers (PDCM). This pilot occurred through collaboration with five physician organizations (POs). POs adopted different models to deliver care management: embedded or centralized. Our research is to compare different care management models in terms of their long-term medical savings. This study incorporated a long follow-up period to generate understanding of the program effects to help policymakers and employer groups on future investment in care management. Trajectories of costs were modeled through splines of truncated power polynomial functions. The number of knots and their locations are unknown and updated via RJMCMC. We found patients engaged by the embedded model constantly showed a lower chance of having medical utilization and their costs notably decreased one year after engagement as compared to HPDCM patients. However, savings between centralized model and HPDCM ran contrary and showed more subdued effects. We also carried out a simulation study to test the sensitivity of the misspecification on the overall inferences.