An Evaluation of Rhode Island's Statewide Multi-Payer PCMH Program: Effects on Utilization and Total Cost of Care
Objective: To assess how the CTC program has affected total cost of care and utilization of services for CTC patients, compared to similar, non-CTC patients.
Study Population: All covered lives who live in RI or who receive primary care services in RI, who receive care for at least one quarter in the study period (N= 2,291,863 person years). We exclude patients in the first five CTC practices, for whom pre-period data are unavailable.
Data source: July 2010-December 2014 RI claims data from commercial payers, Medicaid managed care, Medicaid fee-for-service, and Medicare.
Outcome measures: Cost: Total cost of care, inpatient costs, outpatient costs, professional costs, pharmacy costs; Utilization: ED Visits (all cause)/1000 member months, ED Visits (preventable)/1000 member months, Hospital Admissions (all cause)/1000 member months, Hospital Admissions (ACSC)/1000 member months, % Hospital 30-day Readmissions (all cause), Observation Stays (all cause)/1000 member months.
Study Design: All outcome measures are calculated on a quarterly basis, for each patient in the study population in quartern.
Patients are attributed to the practice on a monthly basis. If a patient is assigned to a PCP, then the patient is assigned to the PCP’s practice. If there is no assigned PCP, the attributed PCP is derived by looking back 27 months and assigning the patient to the PCP with whom they had the most recent preventive visit. If there are no preventive visits, the PCP with the most eligible visits during the previous 27 months is attributed to the member for that month.
A propensity score-matched difference-in-differences framework is used to estimate the effect of CTC on CTC patients, relative to a comparison group of non-CTC patients. Variables used in the propensity score model include age, sex, HCC score, Charlson index, payer type, plan type, zip code index, and the pre-period outcome.
For cost measures, we use negative two-part binomial generalized estimating equation (GEE) models to estimate the effect. Utilization is estimated using GEE models assuming negative binomial or logistic distributions with a log link.
Independent variables include a dummy for whether a patient is in a CTC practice, a dummy for the post period, and an interaction term between CTC status and post period. The interaction term represents the parameter of interest. The model also controls for a time trend and quarter. We use fixed effects at the practice level to control for unobservable differences between patients in CTC versus non-CTC practices.
Results are presented as difference-in-differences estimates with bootstrapped standard errors and clustering at the practice level.