Patient Outcomes After Long-Term Acute Care Hospital Closures

Key Points Question Were federal payment reform and the subsequent closure of long-term acute care hospitals (LTCHs) associated with upstream hospital care practices and outcomes for patients receiving mechanical ventilation? Findings In this cohort study of 8404 Medicare beneficiaries receiving mechanical ventilation for at least 96 hours at 45 closure-affected hospitals and 45 matched controls, LTCH closure was associated with decreased LTCH transfer and decreased spending; in the subgroup of patients receiving a tracheostomy, LTCH closure was additionally associated with increased use of do-not-resuscitate orders and transfer to skilled nursing facilities; there was no change in mortality. Meaning These findings suggest that discharge patterns and advanced directive decisions were sensitive to availability of postacute care options of mechanically ventilated patients.


Introduction
2][3] Like short-stay hospitals, LTCHs aim to provide advanced interdisciplinary care (eg, ventilator weaning by pulmonologists and respiratory therapists) and often have higher nurse-to-patient ratios than SNFs. 4,5LTCHs were created in the 1980s as an administrative exemption from the acute care hospital Prospective Payment System for 40 chronic disease hospitals and are currently defined by patients' mean length of stay (LOS; ie, Ն25 days).Because of longer LOS compared with short-stay hospitals, the Centers for Medicare & Medicaid Services (CMS) historically reimbursed LTCHs at higher rates, leading to rapid LTCH growth in the early 2000s. 6,7However, since 2005, CMS implemented a series of reforms designed to restrict growth and spending. 6Most recently, the Pathway for Sustainable Growth Rate Reform Act of 2013 reserved higher reimbursement rates for care of patients requiring more complex care, such as those receiving prolonged mechanical ventilation (MV) 6 ; reimbursement for patients not meeting complexity requirements was gradually reduced to hospital-level reimbursement rates. 8Due to resulting financial pressures, many LTCHs closed, 8 curtailing one option for postacute care for some short-stay hospitals.How CMS payment reform and subsequent LTCH closures might affect upstream hospital care patterns (eg, tracheostomy placement, palliative care, do-not-resuscitate orders) and overall outcomes (eg, spending and mortality) among patients at risk for prolonged MV is unclear.
The use of LTCHs for postacute care of hospitalized patients varies widely, depending on regional LTCH availability and transfer practices at individual hospitals. 9,10To estimate the association of LTCH closures with upstream short-stay hospital care patterns and patient outcomes, we leveraged hospital variation in use of closing LTCHs and conducted a difference-in-differences study among patients at risk for or receiving prolonged MV.As prior work has shown that earlier discharges to LTCHs (and by extension, earlier tracheostomy) are cost-saving for hospitals (which receive a single prospective payment regardless of LOS), 1 we hypothesized that LTCH closure would be associated with changes in upstream care patterns (ie, decreases in tracheostomy rates, increases in do-not-resuscitate orders, and shifts in discharge patterns) but not with patient outcomes.

Patient Cohort
This cohort study was deemed exempt from review and informed consent by Beth Israel Deaconess Medical Center and Boston University Institutional Review Boards because it was non-human participants research.This study is reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for observational studies.We used data from the Medicare Provider Analysis and Review File (MedPAR) and the Master Beneficiary Summary File from 2011 to 2019, which contain demographic and clinical data on hospital, LTCH, and SNF admissions.We linked to the CMS Chronic Conditions Warehouse 11 to obtain detailed comorbidity data from prior health care encounters.Additional hospital characteristics were obtained from the 2016 American Hospital Association annual survey database.

JAMA Network Open | Critical Care Medicine
Patient Outcomes After Long-Term Acute Care Hospital Closures We identified Medicare fee-for-service beneficiaries aged 66 years and older who were hospitalized between 2011 to 2019 with admission to an intensive care unit (ICU; per revenue center codes).Patients receiving prolonged MV, defined as those receiving MV for at least 96 hours and tracheostomy, 12,13 are known to have complex postacute care needs. 7,14We identified patients at risk for prolonged MV (ie, receipt of MV Ն96 hours; International Classification of Diseases, Ninth Revision [ICD-9] procedure code 96.72 or International Statistical Classification of Diseases and Related Health Problems, Tenth Revision [ICD-10] 5A1955Z, 15,16 who are known to have a high rate of progression to prolonged mechanical ventilation 15,16 ) and the subgroup receiving prolonged MV (ie, additional receipt of tracheostomy; ICD-9 codes 31.1 and 31.23;ICD-10 codes 0B11[0-4]F4 and 0B11[0-4]Z4). 12,13Evaluating patients receiving MV for 96 hours or longer allowed evaluation of associations between LTCH closure and treatment decisions made during acute respiratory failure (eg, tracheostomy placement).
LTCHs were defined by CMS Health Cost Reporting Information System reports (2011-2019). 5,17CHs that ceased operation between 2012 and 2018 were defined as closing LTCHs.We defined LTCH transfers as temporally adjacent hospital and LTCH admissions, where the hospital admission ended on date n and the LTCH admission began on date n or n + 1, a method that has been validated over the use of the discharge destination field in MedPAR data.17

Exposure
Difference-in-differences analyses compare dichotomous exposures, but hospitals vary on a continuum in the proportions of patients transferred to LTCHs and the numbers of different LTCHs used as transfer destinations.Our prior work showed that hospitals in the US transfer approximately 60% of patients receiving tracheostomies to LTCHs for postacute care. 3Therefore, to estimate outcomes associated with near-total loss of LTCH availability, we defined closure-affected hospitals as hospitals discharging at least 60% of patients receiving a tracheostomy to a closing LTCH in the year prior to closure, ie, hospitals that had relied on closing LTCHs for most or all postacute care of patients receiving a tracheostomy (a group with unchanged eligibility for higher CMS reimbursement rates).The exposure was defined at the hospital level because the use of LTCHs for postacute care is often dependent on institutional partnerships between hospitals and LTCHs and ecologic exposures can reduce patient-level confounding. 18

Outcomes
We examined 2 groups of outcomes.Upstream outcomes occurring during the acute short-stay hospitalization included do-not-resuscitate orders (ICD-9 code V49.86; ICD-10 code Z66), palliative care delivery (ICD-9 code V66.7; ICD-10 code Z51.5), tracheostomy rates, and disposition at the end of hospitalization (ie, LTCH transfer, SNF transfer, home discharge, facility or home hospice transfer, and in-hospital mortality).Patient outcomes included hospital LOS, alive and institution-free days within 90 days of admission (days not in a hospital, LTCH, or SNF 3,15 ), spending per days alive within 90 days of admission (sum of Medicare, out-of-pocket, and other payer payments, 19 divided by total days alive within 90 days of admission), and mortality within 90 days of admission.All outcomes, except for tracheostomy rates, were assessed for patients receiving MV for 96 hours or longer and the subgroup who received a tracheostomy.

Statistical Analysis
Difference-in-differences analyses can achieve unbiased effect estimates, even when intervention and control cohorts have differences in underlying characteristics.Still, we minimized possible confounding by 1:1 matching closure-affected hospitals to control, non-closure-affected hospitals (ie, hospitals that had not discharged any patients receiving a tracheostomy to a closing LTCH) on hospital characteristics 20  were identified, the 1 year before and after closure of an LTCH were used as the pre-exposure and postexposure years for exposed hospitals while excluding the year of LTCH closure; matched control hospitals each contributed the same years of data (Figure 1).
After matching hospitals, difference-in-differences analyses were performed at the patient level using multivariable hierarchical regression with 200-fold bootstrapping, adjusting for hospital-and patient-level characteristics to further reduce confounding between exposure and patient outcome.
Model variables included difference-in-differences variables estimating the association of LTCH closure with patient outcomes (absolute change) 21 ; hospital-and patient-level covariables as fixed effects (year and hospital characteristics plus rural vs urban status; patient sex, age, race and ethnicity, Social Vulnerability Index, 22 Medicare and Medicaid dual eligibility, surgical status, 23 chronic comorbidities 11  the Social Security Administration.We included race and ethnicity in fully adjusted models because these have been associated with differences in end-of-life care practices and outcomes among Medicare beneficiaries.Less than 1% of patients had missing data (Social Vulnerability Index); these patients were dropped in adjusted analysis.
To assess the validity of difference-in-difference analyses, we conducted 2 falsification tests using outcomes unlikely to be related to LTCH availability but associated with severity of illness: incidence of septic shock (ICD-9 codes 995.92 and 785.52;ICD-10 codes R6520 and R6521) and number of chronic comorbidities.As an alternative method to account for unobserved heterogeneity in hospital characteristics, we conducted a sensitivity analysis using hospital of admission as a fixed effect rather than random effect.We also conducted 2 sensitivity analyses assessing lower exposure The years before and after long-term acute care hospitals (LTCH) closure were identified as preclosure and postclosure years.In difference-in-difference analysis, control short-stay hospitals (SSHs) were matched to closure-affected SSHs by hospital characteristics; control hospitals contributed the same years of data for analysis as their matched closureaffected SSH.
strengths (eg, hospitals that may have alternate local LTCHs); we varied the definition of closureaffected hospitals to those discharging at least 30% or discharging more than 0% of patients receiving a tracheostomy to a closing LTCH in the year prior to closure.Finally, because standard 2-way fixed-effect difference-in-difference analyses involving staggered treatment timing can lead to bias, 26 we additionally estimated difference-in-differences using the Callaway and Sant'Anna method for staggered difference-in-differences, 26 which reduces bias through the estimation of group-time average treatment effects (ie, separate average treatment effects for each time period of treatment, which are then aggregated by dynamic time using a weighted mean approach).While aggregating by dynamic time produces average treatment effects for varying lengths of time after exposure (in this case, 1-7 years after LTCH closure), our estimate of interest was average treatment effect 1 year after LTCH closure to parallel our primary analysis.Staggered difference-in-differences models were adjusted for the same fixed (patient and hospital characteristics) and random (matched hospital clustering) effects as our primary static difference-in-difference models.Statistical testing was 2-tailed, with α = .05.Analyses were conducted using SAS version 9.4 (SAS Institute).Data were analyzed from October 2022 to June 2023.
In falsification testing, there was no association between LTCH closure and the occurrence of septic shock or the number of chronic comorbidities among patients receiving MV for at least 96 hours and patients receiving a tracheostomy (eTable 5 in Supplement 1).In sensitivity analysis with hospital of admission as a fixed effect, results were similar (eFigure 1 and eFigure 2 in Supplement 1).
In sensitivity analyses exploring hospitals with less prior reliance on closing LTCHs (103 hospitals discharging Ն30% of patients receiving a tracheostomy to a closing LTCH and 215 hospitals discharging >0% of patients receiving a tracheostomy to a closing LTCH), point estimates were generally reduced in magnitude (eg, LTCH referral) and lost statistical significance (eg, spending per day alive) (eTables 6-13, eFigure 3, and eFigure 4 in Supplement 1).In staggered difference-indifferences sensitivity analyses, there was no significant difference between closure-affected and matched control hospitals in any of the outcomes in any year prior to exposure (supporting the parallel trends assumption of our primary analyses), and average treatment effects 1 year after LTCH closure were similar to primary analysis estimates (eFigure 5 and eFigure 6 in Supplement 1).Hospice transfer -1.8 (-4.5 to 0.9) In-hospital death 1.7 (-2.2 to 5.6) Estimates and their 95% CIs of absolute change in outcomes were generated by difference-in-differences (DID) analysis adjusted for hospital-and patient-level characteristics, with matched exposure-control hospital pairs as random effect.DNR indicates do not resuscitate; LOS, length of stay; LTCH, long-term acute care hospital; SNF, skilled nursing facility.

Discussion
This cohort study used difference-in-differences analysis comparing LTCH closure-affected hospitals with non-closure-affected hospitals to estimate how CMS payment reform and subsequent LTCH closures might affect hospital practice patterns and patient outcomes.While most hospitals remained unaffected by LTCH closure, Closure-affected hospitals had decreased LTCH transfers and spending and increased do-not-resuscitate rates among patients receiving tracheostomy, but there were no significant changes in mortality.
Our study adds to prior literature investigating the association of LTCH use with patient outcomes.While prior studies have assessed the association between LTCHs and outcomes of other patient populations (and similarly found associations between LTCH use and increased spending) 6,27 using standard multivariable regression adjusting for hospital-and patient-level confounders 27,28 or instrumental variable analysis (with distance to nearest LTCH, 27 number of LTCHs in the admitting hospitals' hospital referral region, 25 or LTCH openings during prior market expansion as the instrument 6 ), we leveraged the closure of LTCHs in difference-in-difference analysis.By defining exposure hospitals solely by their prior reliance on a closing LTCH for postacute care of patients receiving a tracheostomy, we were able to observe how near-total loss of LTCH availability was associated with changing practices and outcomes (among patients with unchanged eligibility for higher CMS reimbursement rates) compared with otherwise similar but unaffected hospitals, while circumventing issues related to the strength of instruments used in prior instrumental variable analyses.Of note, we observed that while almost 20% of LTCHs closed, only approximately 3% of Hospice transfer -0.8 (-4.9 to 3.2) In-hospital death 0.5 (-7.0 to 8.0) Estimates and 95% CIs of change in outcomes were generated by difference-in-differences (DID) analysis adjusted for hospital-and patient-level characteristics, with matched exposure-control hospital pairs as random effect.DNR indicates do-not-resuscitate; LOS, length of stay; LTCH, long-term acute care hospital; SNF, skilled nursing facility; IFD, alive and institution-free days.The degree to which a hospital relied on a closing LTCH prior to LTCH closure and the degree to which its care patterns were subsequently altered may be related in part to local LTCH prevalence 9,29 : areas with multiple LTCHs 30 may be more insulated from the effect of a closing LTCH compared with those with zero or 1 LTCH. 31Our sensitivity analysis exploring hospitals with less prior reliance on closing LTCHs may suggest a dose-response association, whereby hospitals with higher prior reliance on a single closing LTCH experienced larger changes in LTCH referrals, and the sensitivity of certain outcomes upstream or downstream of LTCH referral (such as spending) may be dependent on the degree of change in LTCH referral.

JAMA Network Open | Critical Care Medicine
CMS's LTCH reimbursement reform, which reserved higher reimbursement for patients admitted to an ICU for 3 days or longer or receiving MV for 96 hours or longer on arrival to the LTCH, was intended to divert other, potentially less complex patients (who previously made up >40% of all LTCH patients 31 ) to lower-cost settings, such as SNFs. 6However, because reimbursement reform not only led LTCHs to adjust admission criteria but also led many LTCHs to close entirely, we found that closure-affected hospitals changed postacute care discharge practices in unintended ways.
Although our cohort would meet CMS criteria for higher LTCH reimbursement rates, we found that patients receiving a tracheostomy were discharged from closure-affected hospitals to SNFs with increased frequency.In the US, tracheostomies are the strongest factor associated with LTCH use over SNF use, as LTCHs are perceived to offer higher levels of interdisciplinary management for complex medical care. 10Like LTCHs, SNFs are known to vary in quality and capability of caring for complex patients.While studies comparing SNF and LTCHs in a heterogeneous cohort of inpatients who were not critically ill admitted to a general medical service found similar clinical outcomes between SNFs and LTCHs, 10,27 the ability of SNFs in the US to care for patients receiving or recovering from prolonged MV in the US is understudied.

Limitations
Our study has limitations.First, analyses are based on administrative data and may be vulnerable to misclassification.However, we used previously validated and published ICD-9 and ICD-10 algorithms 24,35 that have been verified for continuity using CMS General Equivalence Mappings.
ICD-9 and ICD-10 codes are known to slightly overestimate the true incidence of patients receiving MV for 96 hours or longer 36 and underestimate do-not-resuscitate orders and palliative care delivery, 37,38 but performance characteristics are unlikely to be differential across closure-affected and control hospitals.Our claims-based algorithm to adjust for acute severity of illness has also been validated against the Sequential Organ Failure Assessment score with strong performance characteristics. 24Second, we examined a number of outcomes that are known to be patient-centered 39 and/or important to understanding health system resource use

Figure 2 . 4
Figure 2. Change in Hospital Care Patterns and Patient Outcomes Associated With LTCH Closure at Closure-Affected Hospitals Compared With Matched Control Hospitals Among Patients Receiving Mechanical Ventilation for at Least 96 Hours

Figure 3 . 5
Figure 3. Change in Hospital Care Patterns and Patient Outcomes Associated With LTCH Closure at Closure-Affected Hospitals Compared With Matched Control Hospitals in the Subgroup of Patients Receiving a Tracheostomy

Table 1 .
Characteristics of Included Patients 2011 and 2019, 584 973 eligible patients receiving MV for at least 96 hours were admitted to 1287 hospitals and 99 454 patients admitted to 1261 hospitals were transferred to 459 LTCHS.Between 2012 and 2018, 84 LTCHs closed, which had received patients from 219 hospitals.Of these, 46 hospitals (3.6%) had discharged at least 60% of patients receiving a tracheostomy to a closing LTCH in the year prior to LTCH closure and were defined as closure-affected hospitals; 1031 hospitals (81.7%) had not discharged any patients to closing LTCHs; we matched 45 closure-affected hospitals to 45 control hospitals.eTable 1 in Supplement 1 shows characteristics of closure-affected and control hospitals before and after selection of hospital matches, and eTable 2 in Supplement 1 shows characteristics of the patients admitted to these hospitals between 2011 and 2019 before and after selection of hospital matches.After restricting to patients admitted in the year before and after an LTCH closure, a total of 8404 patients (mean [SD] age, 76.2 [7.2] years; 4419 [52.6%] men) were included in difference-in-difference analysis; patient-level baseline characteristics at closure-affected and control hospitals are shown in Table 1, and characteristics of the tracheostomy subgroup are shown in eTable 3 in Supplement 1. Unadjusted proportions of outcomes among patients receiving MV for at least 96 hours and patients receiving a tracheostomy at closure-affected and matched control hospitals are shown inTable 2 and eTable 4 in Supplement 1. Between preclosure and postclosure years, at both closureaffected and control hospitals, unadjusted rates of tracheostomy decreased among patients receiving MV for at least 96 hours; palliative care provision and do-not-resuscitate orders rates increased in patients receiving MV for at least 96 hours (Table 2) and the subgroup receiving a tracheostomy (eTable 4 in Supplement 1).Among the subgroup of patients receiving a tracheostomy, LTCH closure was similarly associated with decreased LTCH transfer (difference, −12.6 [95% CI, −21.7 to −3.6] percentage points) and decreased total spending per days alive within 90 days (difference, −$22 085.00 [95% CI, −$36 495.57 to −$7674.71]).Additionally, LTCH closure was also associated with increased do-notresuscitate rates (difference, 10.3 [95% CI, 4.2 to 16.3] percentage points) and increased SNF transfer JAMA Network Open | Critical Care Medicine Patient Outcomes After Long-Term Acute Care Hospital Closures JAMA Network Open.2023;6(11):e2344377. doi:10.1001/jamanetworkopen.2023.44377(Reprinted) November 21, 2023 5/14 Downloaded from jamanetwork.comby guest on 11/27/2023 JAMA Network Open.2023;6(11):e2344377. doi:10.1001/jamanetworkopen.2023.44377(Reprinted) November 21, 2023 6/14 Downloaded from jamanetwork.comby guest on 11/27/2023

Table 1 .
Characteristics of Included Patients (continued) c Prior LTCH use indicates the proportion of patients receiving tracheostomy discharged to any LTCH in the year prior to LTCH closure.
32,33Our findings suggest that discharge of a subset of patients recovering from prolonged MV (eg, after weaning or while receiving stable MV) to SNFs may result in similar mortality outcomes with lower spending compared with LTCHs, although functional lower total spending is likely driven by the lower prospective payment rate to hospitals and SNFs compared with LTCHs.However, the costs incurred by hospitals and SNFs (over the reimbursed prospective payment fee) may increase as LTCH use decreases; indeed, in the early 2000s, use of LTCHs for postacute care was suggested as a cost-saving measure from the perspective of hospitals. 1Downloaded from jamanetwork.comby guest on 11/27/2023 Future studies should investigate how costs to hospitals and SNFs and the overall economic impact on society have changed with LTCH closures.
2; however, other outcomes, such as attainment of goal-concordant care, patient and family satisfaction, and functional outcomes (including tracheostomy decannulation, mobility, cognitive, and psychiatric outcomes), will be important to assess via other study designs and databases.Third, our analysis was limited to Medicare beneficiaries; analyses of younger cohorts may yield different estimates.Fourth, LTCHs have a wide spectrum of relationships with short-stay hospitals; some may be colocated or co-owned, while others are completely independent.We grouped all sites defined by CMS as LTCHs together in our analysis (ie, LTCHs that are eligible for a higher prospective payment and are subject Keohane LM, Mart MF, Ely EW, et al.Establishing Medicaid incentives for liberating nursing home patients from ventilators.J Am Geriatr Soc.2022;70(1):259-268.doi:10.1111/jgs.1751334.Stelfox HT, Hemmelgarn BR, Bagshaw SM, et al.Intensive care unit bed availability and outcomes for hospitalized patients with sudden clinical deterioration.Arch Intern Med.2012;172(6):467-474. doi:10.1001/archinternmed.2011.231535.Vail EA, Wunsch H, Pinto R, et al.Use of Hydrocortisone, ascorbic acid, and thiamine in adults with septic shock.Am J Respir Crit Care Med.2020;202(11):1531-1539. doi:10.1164/rccm.202005-1829OCFonseca L, Walkey AJ, Ma X, Hua M. Validation of the V49.86 code for do-not-resuscitate status in hospitalized patients at a single academic medical center.Ann Am Thorac Soc.2018;15(10):1234-1237.doi:10.1513/AnnalsATS.201804-257RL 38.Hua M, Li G, Clancy C, Morrison RS, Wunsch H. Validation of the V66.7 code for palliative care consultation in a single academic medical center.J Palliat Med.2017;20(4):372-377.doi:10.1089/jpm.2016.036339.Groff AC, Colla CH, Lee TH.days spent at home-a patient-centered goal and outcome.N Engl J Med. 2016; 375(17):1610-1612.doi:10.1056/NEJMp1607206Characteristics of Closure-Affected and Control hospitals eTable 2. Characteristics of Patients Receiving MV Ն96 at Closure-Affected and Control Hospitals, Before and After Hospital-Level Matching eTable 3. Baseline Characteristics of Patients Receiving a Tracheostomy at Closure-Affected Hospitals and Matched Control Hospitals, After Hospital-Level Matching and Selection of Preclosure and Postclosure Years eTable 4. Unadjusted Outcomes of Patients Receiving a Tracheostomy at Closure-Affected and Matched Control Hospitals eTable 5. Falsification Testing eTable 6. Hospital Characteristics of Sensitivity Analysis, Defining Closure-Affected Hospitals as Those Discharging >30% of Patients Receiving a Tracheostomy to a Closing LTCH eTable 7. Hospital Characteristics of Sensitivity Analysis, Defining Closure-Affected Hospitals as Those Discharging >0% of Patients Receiving a Tracheostomy to a Closing LTCH eTable 8. Baseline Patient Characteristics of Patients Receiving MV Ն96 Hours in Sensitivity Analysis, Defining Closure-Affected Hospitals as Those Discharging Ն30% of Patients Receiving a Tracheostomy to a Closing LTCH eTable 9. Baseline Patient Characteristics of Patients Receiving a Tracheostomy in Sensitivity Analysis, Defining Closure-Affected Hospitals as Those Discharging Ն30% of Patients Receiving a Tracheostomy to a Closing LTCH eTable 10.Baseline Patient Characteristics of Patients Receiving MV Ն96 Hours in Sensitivity Analysis, Defining Closure-Affected Hospitals as Those Discharging Ն0% of Patients Receiving a Tracheostomy to a Closing LTCH eTable 11.Baseline Patient Characteristics of Patients Receiving a Tracheostomy in Sensitivity Analysis, Defining Closure-Affected Hospitals as Those Discharging Ն0% of Patients Receiving a Tracheostomy to a Closing LTCH eTable 12. Unadjusted Outcomes in Sensitivity Analysis, Defining Closure-Affected Hospitals as Those Discharging Ն30% of Patients Receiving a Tracheostomy to a Closing LTCH eTable 13.Unadjusted Outcomes in Sensitivity Analysis, Defining Closure-Affected Hospitals as Those Discharging Ն0% of Patients Receiving a Tracheostomy to a Closing LTCH eFigure 1. Sensitivity Analysis With Hospital of Admission as Fixed Effect, Among Patients Receiving Mechanical Ventilation Ն96 Hours eFigure 2. Sensitivity Analysis With Hospital of Admission as Fixed Effect, Among Patients Receiving a Tracheostomy eFigure 3. Adjusted Outcomes in Sensitivity Analysis Varying Definition of Closure-Affected Hospital, Among Patients Receiving MV Ն96 Hours eFigure 4. Adjusted Outcomes in Sensitivity Analysis Varying Definition of Closure-Affected Hospital, Among Patients Receiving a Tracheostomy eFigure 5. Adjusted Outcomes in Sensitivity Analysis Using the Callaway/Sant'Anna Staggered Difference-in-Difference Method Among Patients Receiving MV Ն96 Hours eFigure 6. Adjusted Outcomes in Sensitivity Analysis Using the Callaway/Sant'Anna Staggered Difference-in-Difference Method Among Patients Receiving a Tracheostomy to CMS payment reform); it is possible that subgroup analysis by LTCH ownership or colocation may lead different estimates.Fifth, a loss of an LTCH could have been accompanied by a gain of another LTCH, thereby biasing our results toward the null.Sixth, the latest year hospital practices and outcomes were assessed was 2019 (for LTCHs closing in 2018), although the full impact of payment reform led to LTCH closures through 2020.Future studies including hospital practices through 2021 (which may additionally be influenced by COVID-19) may be needed to update our findings.ConclusionsThis cohort study found that after CMS reimbursement reform, LTCH closure was associated with changes in upstream hospital care practices, including decreased LTCH transfer, increased do-notresuscitate rates, and decreased overall spending but was not associated with changes in mortality, LOS, or alive and institution-free days.Further studies are needed to understand how LTCH closure may be associated with other important outcomes, including patient and family satisfaction, functional outcomes, and attainment of goal-concordant care.33.