Quality and costs in narrow network skilled nursing facilities
Wednesday, June 15, 2016: 12:20 PM
419 (Fisher-Bennett Hall)
Hospitals increasingly bear the financial risk for health events and health care use occurring after their patients are discharged from the hospital, through payment reforms such as shared savings in accountable care organizations and bundled payment. In response, hospitals have sought closer alignment with a selective set of post-acute care providers such as skilled nursing facilities (SNFs) to have greater control over care received after hospital discharge. However, it remains unclear whether providers that are part of these preferred or narrow networks exhibit higher performance on quality measures and lower length of stay, and how such networks may affect overall Medicare spending and overall quality of care. In this paper, we focus on two prominent health systems with narrow skilled nursing facility (SNF) networks—Partners HealthCare in Boston and the Cleveland Clinic—and compare differences in mortality, readmissions, and SNF length of stay between the “preferred” SNFs and other SNFs within the same market using 2010 Medicare claims. First, we describe unadjusted differences in these outcomes between SNFs in preferred networks and other SNFs that are used by the hospital but are not part of the network. Next, we investigate differences controlling for a rich set of patient characteristics including demographics, socioeconomic status, and health during the preceding hospital stay. We also control for discharging hospital characteristics by including a fixed effect for each discharging hospital. Finally, we use a “differential distance” instrumental variables approach, using the relative distance of the closest preferred versus non-preferred SNF from a patient’s residence as an instrument for the actual choice of SNF. We use the estimates from these regressions to simulate how shifting more patients to preferred SNFs would affect overall Medicare spending and quality of care within included markets. Our preliminary results show that preferred SNFs are more likely to see healthier patients who have a lower probability of mortality and readmissions and who have lower length of stay. However, even after adjusting for this confounding we find that admission to a preferred SNF is associated with lower mortality and lower length of stay. The results are similar when we stratify the analysis by health system. Overall these results suggest that preferred SNF networks have the potential to simultaneously reduce health care costs and improve health outcomes.