Comparative effectiveness of long-term acute care hospital versus skilled nursing facility transfer

Table 3 Healthcare Spending after LTACH vs. SNF Transfer among Hospitalized Adults

	LTACH, $	SNF, $	Difference, $ (95% CI)	P-value
Medicare Spending
Index episode of care^{a, b}
Unadjusted	51,822	21,465	30,358 (27,985-32,730)	<.001
Medicare model	43,630	23,186	20,443 (18,511-22,375)	<.001
Medicare & EHR model	43,721	23,179	20,542 (18,603-22,481)	<.001
Post-Index episode of care^c,d
Unadjusted	32,724	21,843	10,881 (7325-14,437)	<.001
Medicare model	22,788	24,121	-1334 (− 4477-1810)	0.41
Medicare & EHR model	22,790	24,137	-1347 (− 4518-1825)	0.41
Total 1-year spending^b
Unadjusted	84,652	43,343	41,309 (36,304-46,313)	<.001
Medicare model	64,141	47,847	16,294 (11,859-20,728)	<.001
Medicare & EHR model	64,476	47,787	16,689 (12,216-21,162)	<.001
All Payer Spending
Index episode of care^a,b
Unadjusted	54,557	24,705	29,851 (27,267-32,435)	<.001
Medicare model	45,793	26,561	19,232 (16,966-21,498)	<.001
Medicare & EHR model	45,959	26,546	19,414 (17,137-21,690)	<.001
Post-index episode of care^c,d
Unadjusted	37,611	24,804	12,807 (8728-16,887)	<.001
Medicare model	26,336	27,333	− 997 (− 4623-2629)	0.59
Medicare & EHR model	26,348	27,338	− 991 (− 4656-2675)	0.60
Total 1-year spending^b
Unadjusted	92,310	49,541	42,769 (37,346-48,191)	<.001
Medicare model	69,798	54,461	15,337 (10,452-20,222)	<.001
Medicare & EHR model	70,193	54,399	15,795 (10,858-20,731)	<.001

Abbreviations: LTACH long-term acute care hospital, SNF skilled nursing facility
^a Includes the initial acute care hospitalization and subsequent LTACH or SNF stay
^b We computed marginal effects after gamma generalized linear regression model estimation with an identity link function to generate predicted spending for LTACH and SNF cohorts, and contrasts for the difference and robust standard errors using the delta method
^c Includes all hospitalizations and inpatient post-acute care stays (LTACH, SNF, IRF) after the index LTACH or SNF stay through 1-year after the date of transfer
^d We computed marginal effects after a two-part model, and contrasts for the difference and robust standard errors using the delta method. The first part of the model predicted whether spending for a patient was greater than zero. The second part, conditional on positive spending, predicted spending for LTACH and SNF patients using a gamma generalized linear regression model with a log link function

ISSN: 1472-6963