In this retrospective post-acute care cohort study linking Medicare claims to clinically detailed EHR data, we found that LTACH transfer for Medicare beneficiaries hospitalized to an inpatient medicine service in a high-LTACH use region was associated with similar clinical outcomes as SNF transfer (mortality, 60-day recovery, and days spent at home), but with greater healthcare spending (approximately $16,000 per transfer). When we focused on the subgroup of patients most representative of the contemporary LTACH population (i.e. meet site-neural payment criteria for full reimbursement), but did not receive mechanical ventilation after transfer, since these patients may not have been able to be cared for in SNFs, our findings remained consistent with our overall analyses. As payers and health systems are increasingly focused on maximizing value, limiting LTACH transfers for patients who truly need prolonged hospital-level care beyond the capabilities of SNFs, could result in comparable clinical outcomes at a much lower cost. This is especially pertinent during the current pandemic, as LTACHs may play an outsized role in managing the projected surges of patients with prolonged respiratory failure from COVID-19. 26,27
Our findings complement previous studies on the effectiveness of LTACHs which used different comparison groups, datasets, and analytic approaches.9,11,18,19 Our head-to-head comparison with SNFs, which is the principal post-acute care alternate to LTACHs, found similar effectiveness for LTACH transfer as studies that compared LTACH transfer to patients discharged to SNFs, IRFs, or home. Due to our limited sample size, we were unable to explore whether there were subgroups of patients transferred to LTACHs that would benefit compared to SNFs as previous studies have found, most notably chronically critically ill patients and those with multiple organ failure.18,19 We also found that adjustment for confounders available in EHR data, which included more information on clinical severity and complexity than Medicare administrative claims data, were necessary for examining mortality and recovery to better account for selection bias between patients transferred to LTACHs versus SNFs. However, the inclusion of EHR data did not meaningful change our findings for days at home or healthcare spending, which suggests that Medicare data is adequate for risk adjustment for these outcomes. Lastly, our sequential regression modeling approach approximated findings from studies that employed instrumental variable analytic techniques. In the absence of a randomized controlled trial of which post-acute care setting optimizes outcomes, it is reassuring that different approaches yield similar conclusions regarding the overall effectiveness of LTACH transfer.
While SNF transfer in theory may lead to comparable clinical outcomes, in practice, SNFs may lack the expertise, experience, and resources to provide care for this population with complex and serious illness.28 This is likely most pertinent in regions with high LTACH use, such as the South and Ohio Valley, than in the Pacific Northwest, North, and New England regions where LTACH use is scarce.7,8,12 In regions with high LTACH use, SNFs may not have developed or sustained the capabilities to adequately care for this population. Two recent federal policies have changed the financial incentives to shift less medically complex patients from LTACHs to SNFs. Beginning October 2020, the CMS site-neutral payment policy will considerably decrease the reimbursement to LTACHs for patients who did not have a qualifying ICU stay of 3 or more days prior to transfer or did not receive prolonged mechanical ventilation in the LTACH.29 Effective beginning October 2019, the CMS Patient-Driven Payment Model (PDPM) will increase the SNF per-diem reimbursement by 10–30% for patients who require non-rehabilitation care, such as antibiotic infusions, dialysis, wound care, and ventilator support.30,31
Our study should be interpreted in the context of certain limitations. First, our findings may not generalize to low LTACH use regions where the patients cared for in LTACHs are sicker than those in high-use regions, such as the Dallas-Fort Worth metroplex.29 Second, due to inclusion criteria of the original EHR cohort that we used for this study, we were only able to study patients admitted to an internal medicine service, so our findings may not apply to those with trauma, surgical, or neurologic conditions. Third, we were only able to compare outcomes available in Medicare data, so we do not know whether the greater intensity of care provided in LTACHs may have resulted in better cognitive or functional recovery or quality of life. However, patients had similar 60-day recovery and time spent at home, which may be surrogate measures of these more patient-oriented outcomes.