Changes in discharge location and readmission rates under bundled payment

TO THE EDITOR

Patients are often referred for post-acute care after hospitalization to improve outcomes and reduce readmissions. The use of post-acute care has grown rapidly, with costs doubling since 2001.^1,2 Though it can take place at home, skilled nursing facilities, or inpatient rehabilitation facilities, facility-based care is more expensive,^3,4 and it is unknown if it is more effective.

In 2013, NYU Langone Medical Center (NYULMC) joined the national Bundled Payment for Care Improvement (BPCI) model for patients with Medicare fee-for-service insurance undergoing cardiac valve replacement, major joint replacement of the lower extremity, or spinal fusion.⁵ The BPCI model held NYULMC accountable for costs occurring from the index admission to 90-days after discharge. To control these costs, NYULMC attempted to shift referrals from facility-based to home-based post-acute care. In the context of this shift in referrals, we examined the change in readmission rates.

METHODS

The study used complete claims data provided by Medicare. We divided the study period into three phases. Medicare provided baseline period data from 7/1/2009–5/30/2012. Of note, NYULMC was closed from 10/29/2012–12/27/2012 because of superstorm Sandy. In the preparation period (1/1/2013–9/30/2013), NYUMC began preparations for BPCI, but cost incentives were not in effect. In the risk-bearing period (10/1/2013–8/31/2014), cost incentives took effect. Inclusion criteria were determined by the BPCI model.

For each condition, we examined whether BPCI period was associated with discharge to post-acute care. We also examined for each condition whether the BPCI period was associated with 30-day readmission. For all models, we used generalized estimating equation models, controlling for age, race, sex and major complications or comorbidity (as determined by a “major complication or comorbidity” principal diagnosis related group) during index admission, with clustering at the physician level.

RESULTS

Patient characteristics appear in Table 1. Adjusted discharge rates to post-acute care facilities fell from 71% in the baseline period to 22% in the risk-bearing period (OR=0.10, 95% CI 0.07, 0.15) for cardiac valve surgery, from 68% to 34% (OR=0.26, 95% CI 0.22, 0.31) for major lower extremity joint replacement, and from 40% to 30% (OR=0.71, 95% CI 0.50, 1.01) for spinal fusion surgery (Figure 1). In these same cohorts, readmission rates were unchanged in cardiac valve surgery patients (17% vs. 15%, OR=0.92, 95% CI, 0.63, 1.35), significantly decreased in joint replacement patients (7% vs. 5%, OR=0.68, 95% CI 0.49, 0.96) and unchanged in spinal fusion patients (9% vs. 10%, OR=1.10, 95% CI 0.65, 1.87). Average length of stay decreased for all conditions from the baseline to the risk-bearing period (results not shown).

Table 1.

Characteristics of Bundled Payment Episodes by Type

Characteristics	Baseline Period	Risk-Bearing Period	P-value
Cardiac valve, No.	644	315
Female, No. (%)	317 (49)	140 (44)	0.16
Age, mean (SD), y	78.1 (8.4)	76.3 (8.0)	0.001
DRG with MCC, No. (%)	415 (64)	175 (56)	0.008
Major lower joint replacement, No.	1908	1028
Female, No. (%)	1286 (67)	673 (66)	0.29
Age, mean (SD), y	74.6 (8.9)	72.0 (8.6)	<0.0001
DRG with MCC, No. (%)	120 (6)	38 (4)	0.003
Spinal fusion, No.	518	229
Female, No. (%)	300 (58)	134 (59)	0.88
Age, mean (SD), y	73.1 (9.0)	70.7 (9.3)	0.001
DRG with MCC, No. (%)	35 (7)	18 (8)	0.59

Abbreviations: DRG, diagnosis related group; MCC, major complication or comorbidity.

Figure 1.

COMMENT

We achieved an absolute 49% and 34% reduction in rates of discharge to post-acute care facilities in patients undergoing cardiac valve surgery and lower extremity joint replacement surgery, respectively, with no corresponding increase in readmission rates. Readmission rates were similarly stable in patients undergoing spinal fusion, in whom post-acute care facility rates were unchanged.

Our findings suggest that institutions may be able to shift some patients from facility-based to home-based post-acute care without adversely affecting hospital readmission rates or length of hospital stay. Although we did not investigate the mechanism for the major decrease in rates of referral to facility-based post-acute care, it is important to note that the inpatient rehabilitation facility owned by NYULMC closed in the risk-bearing period. Such a dramatic decrease may not have occurred without the change in availability of inpatient beds.

A limitation of the study is that we did not examine outcomes such as functional status or quality of life, which are worth further investigation. A further limitation is that the disease burden in the population may have changed during the study period, though we adjusted for the presence of major complications or comorbidities.

Our findings raise questions about the value of providing post-acute care services in facilities, where care is more costly and potentially more disruptive to the lives of patients and families.