Survival, Function, and Cognition After Hospitalization in Long-Term Acute Care Hospitals

Key Points Question What are the longer-term outcomes of middle-aged and older adults after hospitalization in long-term acute care hospitals (LTCHs)? Findings In this cohort study with 396 participants, 4 of 5 middle-aged and older adults hospitalized in an LTCH either died or survived with severe impairment characterized by dependencies in 2 or more activities of daily living or dementia within 2.5 years of hospitalization. Better survival prognosis and functional and cognitive status before hospitalization were associated with more favorable outcomes. Meaning These findings suggest that most middle-aged and older adults either die or survive with severe impairment (functional, cognitive, or both) within 2.5 years of hospitalization in an LTCH; prehospitalization health could guide patients, families, and clinicians in decisions about prolonged acute care.


Introduction
2][3] Despite recent changes in payment with an overall decline in the number of LTCHs in the US, 4 more than 70 000 fee-forservice (FFS) Medicare beneficiaries were hospitalized in an LTCH at a cost of $3.4 billion to Medicare. 1 Mortality rates are high in this medically complex population, with fewer than half of older adults surviving the year after LTCH hospitalization. 5,6However, little is known about functional and cognitive outcomes among survivors.These outcomes are important because patients with serious illnesses have goals besides living as long as possible, including maintaining independence. 7derstanding of functional and cognitive outcomes in addition to survival is needed to guide goals of care discussions and inform shared decision-making between clinicians and older adults and their families when faced with prolonged acute illness.][10][11][12] More accurate estimates of longterm functional and cognitive outcomes are necessary because families of patients cared for in an LTCH have overly optimistic expectations of recovery 9,13 and an unmet need for prognostic information to guide decision-making. 14,15 this study, we used data from a longitudinal, nationally representative survey of middle-aged and older adults, the Health and Retirement Study (HRS), 16 with biennial assessments of function and cognition and linkage to Medicare claims allowing identification of hospitalization in an LTCH.We aimed to describe survival, functional, and cognitive status after LTCH hospitalization and to identify factors associated with an adverse outcome.

Data Source and Study Population
We conducted a retrospective analysis of the HRS, a longitudinal cohort study of US adults aged 50 years or older. 16Core interviews were conducted biennially via telephone until death.We used data from the 2002 to 2020 HRS core interviews linked with FFS Medicare claims to identify participants with hospitalization in an LTCH between January

Cohort Assembly
We identified 632 LTCH hospitalizations among 480 participants (Figure 1).For participants with multiple stays during the study period, we selected the first hospitalization in the interval between 2 HRS interviews to avoid immortal time bias because survival was an outcome of interest.We excluded participants missing functional or cognitive assessment in the pre-LTCH (baseline) HRS interview, which occurred up to 2.5 years before the LTCH admission, resulting in a sample of 396 participants with an LTCH hospitalization.

Baseline Characteristics
We ascertained baseline sociodemographic and health characteristics from the HRS core interview immediately preceding the index LTCH hospitalization (hereafter, pre-LTCH hospitalization HRS interviews that asked participants to self-identify their race and ethnicity.These data are reported as Hispanic, non-Hispanic Black (hereinafter, Black), non-Hispanic White (hereinafter, White), or non-Hispanic other race.We considered all participants who self-identified as Hispanic to be Hispanic regardless of reported race.We combined participants who reported being American Functional status was ascertained from HRS interviews as the count of dependencies in activities of daily living (ADLs), determined as self-reported need for assistance in walking across a room, dressing, bathing, eating, transferring in or out of bed, or toileting (range, 0-6).Cognitive status was ascertained from HRS interviews using the Langa-Weir algorithm, which uses selfreported and proxy measures to classify participants into 3 groups: normal cognition, cognitive impairment nondementia (CIND), and dementia. 18We combined information on functional and cognitive status to define 3 categories of overall impairment: no impairment (no dependency in ADLs

Outcome
Our primary outcome was death or survival with 2 or more ADL dependencies or dementia, representing a combination of vital status, function, and cognition within 2.5 years of LTCH hospitalization, the time period for biennial HRS interviews.For the few survivors missing a post-LTCH interview (n = 8), we imputed functional and cognitive status from the prehospitalization interview assuming no change.Because the median time to recovery among older ICU survivors is nearly 3 months, 19 we only included assessments conducted 90 days or more after LTCH admission to ascertain outcomes.Death was ascertained from the HRS Tracker File, which determines mortality from next-of-kin interviews supplemented with the National Death Index. 20We additionally searched the Medicare Master Beneficiary Summary File for completeness of vital status information.

Statistical Analysis
We examined baseline characteristics of participants, stratified by primary outcome.We described outcomes by pre-LTCH baseline impairment status through a Sankey diagram representing the proportion of participants in each outcome category.Given our small sample size and event rate, we constructed a parsimonious multivariable logistic regression model to estimate associations between a priori selected risk factors available to clinicians and the primary outcome.Candidate factors identified from prior literature and the interdisciplinary expertise of our group included baseline impairment status, Lee prognostic index score, prolonged ICU stay, and mechanical ventilation during acute care or LTCH hospitalization. 5,10To assess the relative contribution of factors, we estimated a population attributable fraction.To facilitate shared decision-making and care planning for the subgroup of patients with a prolonged ICU stay and mechanical ventilation, we created a nomogram of estimated probability of an adverse outcome for factors with an association.We selected this subgroup because it is most representative of the LTCH population since implementation of the Centers for Medicare & Medicaid Services (CMS) site-neutral payment policy in 2016 that stipulates prolonged ICU stay, use of mechanical ventilation, or both for full reimbursement for LTCH hospitalizations. 21 conducted several sensitivity analyses to examine the robustness of our primary modeling approach.First, since we hypothesized that participants with severe impairment at baseline would be unlikely to improve after hospitalization, we repeated our analyses for the outcome of death only.
Second, because functional recovery can occur up to 6 months after critical illness, we conducted a sensitivity analysis excluding those who completed their post-LTCH interview before this period.
Third, since the measure of dependency in bathing in the Lee index overlaps with functional impairment in ADLs, 17 we repeated our multivariable model with a modified version of the Lee index omitting this factor.Fourth, given the variability in time between hospitalization and post-LTCH HRS interviews, we repeated our models accounting for time as a covariate.Fifth, because survey weights may not yield a nationally representative sample for our cohort, 22 we repeated our primary models using unweighted data.For all analyses, unless otherwise specified, we accounted for the complex survey design of the HRS and generated weighted means and percentages.P < .05(
Their median age was 75 (IQR, 68-82) years; 201 (51%) were women and 195 (49%) were men (Table 1).In the sensitivity analysis evaluating death only, the association with the Lee index was slightly attenuated (AOR, 2.8 [95% CI, 1.8 to 4.5]).Severe pre-LTCH impairment was no longer associated with death (eTable 2 in Supplement 1).When we excluded participants who completed the post-LTCH interview before 6 months after hospitalization, the magnitude of the association with the Lee index was similar; however, severe impairment was no longer associated with death or severe impairment (eTable 3 in Supplement 1).In the remaining sensitivity analyses removing the overlapping measure of bathing from the Lee index, accounting for time between HRS interview and outcome assessment, and using unweighted data, the associations were similar compared with the main analysis (eTables 4, 5, and 6 in Supplement 1).
To facilitate interpretation of the relative contribution of factors in our multivariable model, we estimated the probability of an adverse outcome for a given Lee index score across different baseline impairment status for participants with a prolonged ICU stay and mechanical ventilation (Figure 3).
Although the prevalence of an adverse outcome was high in this population overall, the probability was lower for patients with good baseline survival prognosis ( impairment (range, 56%-68%).For those with exceedingly poor survival prognosis (Lee index Ն14), the probability of an adverse outcome ranged from 91% to 98% and did not vary meaningfully by baseline impairment status.

Discussion
In this cohort study of middle-aged and older US adults enrolled in a nationally representative survey, we found that 4 of 5 participants died or survived with severe impairment, defined as dependencies in 2 or more ADLs or dementia within 2.5 years of hospitalization in an LTCH.The probability of dying or surviving with severe impairment was highly dependent on health status before admission.The patients most likely to survive with reasonably intact function and cognition were those who had a good survival prognosis before admission (Lee index <6) with either no or mild impairment (Յ1 ADL dependency and no dementia).Our findings highlight the importance of factoring in preillness health status to inform goals of care and decision-making for older adults with prolonged acute illness.
5][6] Among studies including adults across the lifespan, 8,[10][11][12]23 good functional status has been reported for as many as 50% of adults 12 and no or mild cognitive impairment in more than three-fourths 11 in the year after LTCH hospitalization. In trast, we observed a greater prevalence of death or survival with severe impairment (functional, cognitive, or both) in our cohort.This discrepancy is not surprising because prior prospective studies included patients who were younger and relatively functionally intact at baseline, excluding up to four-fifths of eligible participants and potentially yielding overly optimistic estimates.24 Because our study sample was derived from a longitudinal survey that prospectively assessed functional and cognitive status before and after LTCH hospitalization, we were able to characterize the composite outcome of survival, function, and cognition in a population of older adults who often have preexisting impairments.
Our findings have important clinical implications.First, the observation that 4 in 5 older adults died or survived with severe impairment (functional, cognitive, or both) after LTCH hospitalization Weighted percentages are reported for all categories.For the smallest categories of transitions, the values are as follows: 5%, from mild to no impairment; 1%, from severe to no impairment; and 5%, from severe to mild impairment.No impairment indicates no dependency in activities of daily living (ADLs) and normal cognition; mild impairment indicates dependency in 1 ADL, cognitive impairment nondementia, or both; and severe impairment indicates dependencies in 2 or more ADLs, dementia or both.A previous study reported that among Medicare beneficiaries hospitalized in an LTCH, only 1% received a specialist palliative care consultation during the acute care hospitalization preceding transfer or in the LTCH. 5 Although LTCHs provide multidisciplinary care, including a focus on ventilator weaning and interdisciplinary rehabilitation, only one-third offer palliative care. 3,25erefore, conversations about goals of care and involvement of specialist palliative expertise should occur during the acute care hospitalization preceding LTCH transfer.Second, the risk factors identified in our study can be used by clinicians to guide shared decision-making by providing a range of probabilities of a composite adverse outcome that older adults and their caregivers value more in treatment decisions and care planning than survival alone. 26For example, a 76-year-old male smoker with chronic lung disease and dependency in bathing and managing finances at baseline (Lee index = 14) is unlikely to survive without severe impairments after LTCH hospitalization.Thus, the care team should share this prognostic information and discuss hospice as an alternative to prolonged hospital care in an LTCH if aligned with the goals of both the patient and the caregiver.In contrast, a 66-year-old female nonsmoker with diabetes, no or mild cognitive impairment, and no dependency in ADLs but difficulty managing finances (Lee index = 5) has a 30% to 50% probability of surviving without severe functional and cognitive impairment.Depending on the goals of care, this patient could benefit from prolonged acute care in an LTCH setting.Our results suggest that assessing survival prognosis using the Lee index (available online 27 ), as well as ADL functioning and cognitive status before acute illness, can inform expectations and guide treatment choices among older adults with prolonged acute illness.

JAMA Network Open | Critical Care Medicine
Our study has several notable strengths that advance our understanding of prognosis after a prolonged acute illness.12]28 As such, our cohort had substantial representation of Black adults and those from lower socioeconomic backgrounds not commonly represented in prospective recovery cohorts of prolonged acute illness. 11,12,23Second, we included all patients with an LTCH hospitalization and prehospitalization HRS assessments, which circumvents issues of recall bias of prior health status and selection bias of including healthier patients who are more likely to consent to participate.Third, our observed rates of survival are consistent with prior literature, which provides external validity to our findings. 6,8,10,23urth, we examined functional and cognitive outcomes in addition to survival after an LTCH stay, patient-centered outcomes that matter to older adults. 29

Limitations
Our findings should be interpreted in the context of certain limitations.First, our assessment of illness severity was limited to characteristics that could be ascertained in claims records.
Nevertheless, we included a prolonged ICU stay and mechanical ventilation in our models, 2 clinical factors known to be important in prior work examining functional outcomes after acute and critical illness. 5,30,31Second, our study included patients with an LTCH hospitalization before implementation of the CMS site-neutral payment policy in 2016, designed to narrow LTCH focus to more severely ill patients with a prolonged ICU stay or mechanical ventilation.Thus, we anticipate the risk of an adverse outcome in a contemporary LTCH population to be even worse than estimates obtained in our study.Third, the relatively small cohort precluded evaluation of the following: individual components of the Lee index as risk factors; functional and cognitive impairments as separate outcomes; outcomes associated with LTCH facility-level characteristics; or disentanglement of outcomes by different care pathways after the LTCH stay, including postacute care delivered in skilled nursing facilities or rehabilitation hospitals or at home via home health services. 5,32Finally, although we used weights in our analyses as recommended by the HRS, our findings may not be nationally representative due to the relatively small sample size and to potential for differences in weights for participants with linked vs unlinked Medicare data. 22The similar estimates for risk factors in our unweighted models, however, increase confidence in our findings.
In this cohort study with 396 participants, 4 of 5 middle-aged and older adults hospitalized in an LTCH either died or survived with severe impairment characterized by dependencies in 2 or more activities of daily living or dementia within 2.5 years of hospitalization.Better survival prognosis and functional and cognitive status before hospitalization were associated with more favorable outcomes.Meaning These findings suggest that most middle-aged and older adults either die or survive with severe impairment (functional, cognitive, or both) within 2.5 years of hospitalization in an LTCH; prehospitalization health could guide patients, families, and clinicians in decisions about prolonged acute care.

Figure 2 .
Figure 2. Proportions of Participants Across Categories of Impairment and Survival After Long-Term Acute Care Hospital (LTCH) Hospitalization by Pre-LTCH Impairment Status 1, 2003, and December 31, 2016.The HRS was approved by the University of Michigan Institutional Review Board.All HRS participants provided informed written consent.The current study was approved by the University of California, San Francisco Institutional Review Board.This report adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

JAMA Network Open | Critical Care Medicine Survival
, Function, and Cognition After Hospitalization in Long-Term Acute Care Hospitals ). Sociodemographic characteristics included age, sex, race and ethnicity, education, marital status, JAMA Network Open.2024;7(5):e2413309.doi:10.1001/jamanetworkopen.2024.13309(Reprinted) May 28, 2024 2/12 Downloaded from jamanetwork.comby guest on 06/07/2024 living situation, and income.Because race and ethnicity are key sociodemographic characteristics and known determinants of health outcomes, information on race and ethnicity was obtained from

Table 1 .
Baseline Characteristics of Study Participants a Lee index <6) and no or mild JAMA Network Open | Critical Care Medicine Survival, Function, and Cognition After Hospitalization in Long-Term Acute Care Hospitals JAMA Network Open.2024;7(5):e2413309.doi:10.1001/jamanetworkopen.2024.13309(Reprinted)May28,2024 5/12 Downloaded from jamanetwork.combyguest on 06/07/2024 Abbreviations: ADL, activity of daily living; CCI, Charlson Comorbidity Index; CIND, cognitive impairment nondementia; COPD, chronic obstructive pulmonary disease; HRS, Health and Retirement Study; ICU, intensive care unit; LTCH, long-term acute care hospital; NA, not available (data not shown for cell sizes Յ25).aUnless indicated otherwise, values are presented as the unweighted No. (weighted %) of participants.bObtainedfrom HRS interviews that asked participants to self-identify their race and ethnicity.cAll who self-identified as Hispanic were considered Hispanic regardless of reported race.Due to small numbers, the "other" category combines participants who reported being American Indian or Alaska Native, Asian or Pacific Islander, or of other race.d Prognostic index for survival developed and g Patients who (1) spend 3 days or more in an ICU during the preceding acute care hospitalization, (2) require prolonged mechanical ventilation of 96 hours or longer, or (3) have a primary LTCH diagnosis other than psychiatric or rehabilitation.

Table 2 .
Multivariable Analysis Identifying Factors Associated With Death or Severe Impairment After LTCH Hospitalization b Includes functional impairment, cognitive impairment, or both.c Population Attributable Fraction (PAF) for Factors Evaluated in the Multivariable Model for the Outcome of Death or Severe Impairment After Long-Term Acute Care Hospitalization eTable 2. Multivariable Analysis Identifying Factors Associated With the Outcome of Death Following LTCH Hospitalization eTable 3. Multivariable Analysis Identifying Factors Associated With Death or Severe Impairment Following Hospitalization in an LTCH Among the Subgroup of Participants With Post-LTCH Interview or Death Beyond 6 Months of Hospitalization (n = 215) eTable 4. Sensitivity Analysis Replacing the Lee Index in the Multivariable Model With a Modified Version Removing the Factor of Dependency in Bathing That Overlaps With the Assessment of Functional Impairment eTable 5. Sensitivity Analysis With Additional Adjustment for Time Between Admission to an LTCH and Assessment of the Outcome of Either Death or Survival With Severe Impairment as Assessed in the Post-LTCH Health and Retirement Study (HRS) Interview eTable 6. Sensitivity Analysis Repeating the Multivariable Model for the Outcome of Death or Survival With Severe Impairment Without Survey Weights eFigure 1. Post-LTCH Outcome of Death Only, Stratified by Pre-LTCH Admission Impairment Status eFigure 2. Post-LTCH Outcome Stratified by Pre-LTCH Impairment Status Among the Subgroup of Participants Who Completed the Post-LTCH HRS Interview or Died Beyond 6 Months of Hospital Discharge (n = 215)