Effects of Vertical Integration and Competition on Length of Stay in Post-Acute Care Setting

Post-acute care facilities (PACs) are a significant source of geographic variation in spending with rising trends over the last 15 years; however, current research has lagged in explaining the variation or quality benefits, if any, of increased spending. Prospective payment systems (PPS) were put in place to curb spending but with limited effect on most PACs. The PPS for home health agencies (HHAs), which account for almost 40% of the PAC industry, consist of 60-day episode payments with unlimited continuous recertification for more episodes if the patient continues to be eligible, which is self-determined. This is unlike other provider group payments and can have perverse incentives for retaining patients.

In this paper, I study how changes in market structure or firm structure affect incentives to retain patients. While competitive markets should theoretically tend towards efficient allocations, regulations and imperfect markets can lead to exploitation of these payment structures—excessive recertification—as competition increases. Vertical integration can reduce these excessive lengths of stay by providing a steady patient flow, following organizational economic theory that vertical integration improves coordination and efficiency . Importantly, I also consider the resulting impact of changes in the rate of recertification on Medicare spending and patient outcomes.

I use 100% Medicare observational claims data linked with HHA claims from 2007 to 2013, allowing me to follow patients discharged from hospital to HHA over time. I augment this with hospital-level information from American Hospital Association, market-level demographic information from U.S. Census, and integration data from Hospital Cost Reports. The methodology consists of longitudinal regressions which test the effect of market structure and integration status on recertification rates, differences in spending and any adverse effect on readmissions. I use instrumental variables, hospital-agency pair fixed effects, and a conditional logit model to construct a Herfindahl-Hirschman index to overcome the main threats to identification: patient selection bias, provider-level selection bias, and measurement bias of competition. Thanks to the richness of the data, I control for many observable patient-level differences. Additionally, I rerun specifications using a matched sample of patients from integrated and non-integrated hospitals.

The results show that vertical integration reduces recertification rates by 3.4 percentage points (20%), while competition increases recertification rates by 1.1 percentage points per standard deviation of HHI (7%). Vertically integrated agencies are insulated from the competitive effects. Lowered recertifications translates to reduced Medicare spending by 5.8-7.5%. Competition resulted in 4-5.5% increase in HHA spending. No impacts on readmissions suggests that the extra recertifications under competition among the non-integrated agencies cannot be considered high-value care.