Impacts of the Maryland Hospital System Financing Model on Hospital-based Psychiatric Care

Background: In calendar year 2014, a new hospital system financing model that covers all hospital inpatient and outpatient services and all payers of hospital services was implemented statewide in Maryland. Under this Maryland model, a quasi-public regulatory organization assigns each hospital a prospective annual budget, sets performance targets for each hospital’s rate of 30-day inpatient readmission, and determines financial incentives or penalties based on past performance. The broad goals of the model are to limit increases in health care costs and improve care coordination and care quality for individuals with complex illness. This study examines the impacts of the Maryland model on psychiatric inpatients’ services utilization outcomes.

Method: We used Medicare claims data to examine four services utilization outcomes—inpatient length of stay, post-discharge outpatient mental health follow-up within 7 days, 30-day all-cause inpatient readmission, and 30-day psychiatric inpatient readmission. A differences-in-differences approach was used to identify the effects of the Maryland model during three policy implementation periods starting in January 2014 (the first 6 months, months 7 to 18, and months 19 to 30), where the difference-in-differences contrasts were between the post-policy periods and the pre-policy period (2012-2013) and between Maryland and inpatients in two other contiguous states (Pennsylvania and Virginia) during the same time periods. A total of 221,919 psychiatric inpatient stays were identified using a combination of principal discharge diagnosis and provider identification numbers. Consistent with similar previous studies, exclusion criteria included discontinuous Medicare enrollment, transfer to another acute care hospital, planned readmission, discharge against medical advice, enrollment in a health maintenance organization, and death (n=54,132, 24.4%). An additional 3,158 stays (1.4%) with missing covariate values were also excluded, leaving a final sample of N=164,629. Inverse probability weights were constructed to ensure acceptable (<.25 SD) covariate balance across twelve study sub-groups defined by unique combinations of state and policy period. Double-robust inverse probability weighted regression models were then estimated with covariates for demographic characteristics, discharge diagnoses, and hospital characteristics.

Results: Most inpatients had principal diagnoses for psychotic disorders (51%), depressed mood disorders (24%), or drug and alcohol disorders (13%), and most (69%) had discharge diagnoses indicating 2 or more co-occurring chronic conditions. Regression estimates indicate that the new financing model was associated with significant (P<.05) reductions in mean length-of-stay during the index inpatient admission and with post-discharge reductions in all-cause readmissions, psychiatric readmissions, and outpatient mental health follow-up visits. Marginal effect sizes ranged from approximately 5% to 20%, depending on outcome and policy period.

Conclusion: Implementation of the Maryland model was associated with reductions in all-cause readmissions and psychiatric readmissions among psychiatric inpatients in Maryland hospitals relative to similar cohorts of psychiatric inpatients at hospitals in two comparison states. However, the finding that the model also resulted in a lower probability of post-discharge mental health outpatient follow-up was an unintended consequence suggesting that the model has not led to overall improvements in care coordination for individuals with complex psychiatric illness.