The Comparative Advantage of Accountable Care Organizations in Structuring Post-Acute Care for Medicare Beneficiaries

Post-acute care providers, including skilled nursing facilities (SNFs), inpatient rehabilitation facilities (IRFs), home health agencies (HHA), long term care hospitals (LTCHs), and others cost over $60 billion dollars (10% of budget) for Medicare in 2016 and have spending growth that outpaces all other areas of Medicare spending. Little is known about best-practices to structuring post-acute care for an elderly population. Among all hospital discharges for Medicare beneficiaries, 40% require some form of post-acute care, 36% of which can enter multiple settings of rehabilitation during their post-acute episode (e.g. skilled nursing and home health post-discharge). Since 2012, Medicare has provided financial incentives for certain groups of physicians to improve quality, and reduce the cost of healthcare—known as Accountable Care Organizations (ACOs). Little is known about how ACOs structure post-acute care compared to other providers. We evaluate models of rehabilitation selected by ACOs compared to similar non-ACO groups, with an emphasis on the entire episode of rehabilitation.

We selected 618,574 hip fracture and 1,282,606 lower extremity joint replacement cases between 2013-2015, of which 14.0% and 16.0% were for ACO attributed Medicare beneficiaries. We identified ACO Beneficiaries using the Medicare Beneficiary Summary File linked to Medicare ACO Attribution files. Hospitalizations, SNF stays, IRF stays, and Home Health visits were identified using MedPAR and outpatient claims files.

We fit a linear probability model to estimate differences in where patients are sent post-discharge for either condition between ACO and non-ACO groups. To determine if ACOs structure post-acute care more effectively after formation than other providers, we use a propensity score matching approach, to compare outcomes of 30-day hospital readmission, 90-day post-acute spending, 90-day mortality, and 90-day rate of successful reentry to the community between ACO and non-ACO attributed beneficiaries. We apply propensity score matching of ACO attribution to balance on observables (demographics, hospital characteristics, and clinical indicators) and estimate treatment on treated effect estimates. We conduct sensitivity analysis for our propensity score specification by A.) Varying the model for selection on observables in forming our control populations of hip fracture and lower extremity joint replacement, and B.) Restricting our propensity model to constrain selection of controls within hospitals for which there was at least one treatment observation.

In unadjusted analysis, we find that between the years 2013-2015, ACOs directed their hip fracture patients to IRFs 3.4 percentage points more often upon discharge. For the group of hip fracture discharges, patients belonging to ACOs were home by 90 days 4.1 percentage points more often than non-ACO patients. For both the hip fracture and lower extremity joint replacement cohorts, ACOs use combination therapy (e.g. SNF combined with home health) more frequently. Our propensity score matching approach results suggest that the ACO hip fracture cohort experienced decreased 90-day mortality by 1.8 percentage points, and their rate of 90-day successful reentry to the community was 4.1 percentage points higher relative to controls. Persons attributed to an ACO discharged from the hospital with hip fracture were 1.46 percentage points less likely to be dead by 90 days post-discharge.