Social Determinants of Health and Delivery of Rehabilitation to Older Adults During ICU Hospitalization

Key Points Question Are social determinants associated with differential delivery of skilled rehabilitation services to older adults during hospitalization with a critical illness? Findings In this cohort study of 1618 hospitalizations in older adults, after accounting for prehospitalization disability and acute illness characteristics, dual Medicare and Medicaid eligibility and rural residence were associated with a lower likelihood of delivery of any rehabilitation, whereas limited English proficiency was associated with reduced amount of rehabilitation services delivered during a critical illness hospitalization. Meaning These findings suggest social determinants of health should be taken into consideration in efforts to enhance equitable delivery of skilled rehabilitation to older adults who are critically ill.


Introduction
Surviving a critical illness, an increasingly common occurrence among older adults, [1][2][3] is frequently accompanied by new or worsening disability. 4,5Skilled rehabilitation with physical therapy (PT) and occupational therapy (OT) during hospitalization facilitates mobilization of patients recovering from critical illness and is known to prevent functional decline [6][7][8] and help to identify post-acute care needs. 9Therefore, inequitable delivery of rehabilitation services to older adults with social or economic disadvantage may contribute to downstream disparities in disability. 10Whether social determinants of health (SDOH) are associated with differences in delivery of skilled rehabilitation during a critical illness hospitalization is unknown.
5][16] However, these studies did not investigate associations between SDOH and delivery of in-hospital rehabilitation.
8][19] For example, limited English proficiency (LEP) could be associated with lower rehabilitation delivery due to perceived barriers to engagement with therapy services or differences in management of sedation and delirium by clinicians. 20Furthermore, variation in resources and practices at hospitals caring for a higher proportion of patients with socioeconomic disadvantage could lead to reduced delivery of skilled rehabilitation services, as has been observed for other care processes. 21,22A common challenge in evaluation of health care disparities is the absence of granular information on SDOH, such as LEP, income, and education, and preexisting health status, that could influence in-hospital treatment needs, such as prehospitalization disability in the case of rehabilitation services.We leveraged a nationally representative longitudinal study of aging with detailed information on SDOH and prehospitalization geriatric risk factors, linked with administrative claims, to investigate whether SDOH are associated with differences in the delivery of skilled rehabilitation to older adults during hospitalization in an intensive care units (ICU).

Methods
The protocol for the National Health and Aging Trends Study (NHATS) was approved by the Johns Hopkins University institutional review board, and our cohort study using these data was approved by the Yale University institutional review board.All participants provided informed consent.We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Study Population
Data were drawn from the National Health and Aging Trends Study (NHATS), a longitudinal, nationally representative survey of community-dwelling Medicare beneficiaries ages 65 years and older living in the contiguous United States. 23The initial sample was drawn from the Medicare enrollment database on September 30, 2010. 24The survey collected information on sociodemographics, including race and ethnicity, insurance, education, income, English proficiency, rural residence, and clinical characteristics, through annual in-person interviews starting in 2011.If a participant was unavailable for interview, a proxy knowledgeable about their health was interviewed.
Race and ethnicity were categorized as Hispanic, non-Hispanic Black, non-Hispanic White, and other (including participants reporting race as American Indian, Alaska Native, Asian, Native Hawaiian, Pacific Islander, or other race).Race and ethnicity were included in descriptive analyses because they are key SDOH.

JAMA Network Open | Critical Care Medicine
Social Determinants of Health and Delivery of Rehabilitation to Older Adults During ICU Hospitalization

Ascertainment of ICU Admissions and Acquisition of ICU Hospitalization Data
ICU admissions were identified through linked inpatient claims files for Medicare fee-for-service and Medicare Advantage participants using critical care revenue codes indicating admission to general, specialty, or coronary care units but excluding psychiatric and intermediate care units. 25Information on mechanical ventilation and organ dysfunction was obtained using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) and International Statistical Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnosis and procedure codes (eTable 1 in Supplement 1). 26,27ICU length of stay was determined based on days with a critical care revenue code.

Outcome Ascertainment
Our primary outcomes were delivery of any PT or OT, as determined by revenue center codes 042.X and 043.X respectively, and amount of PT or OT, determined as number of units of evaluation or treatment delivered during ICU hospitalization.In general, PT and OT are billed in 15-minute increments; therefore, 1 billed unit represents 15 minutes of intervention by a therapist.These units were modeled as units per day to account for differences in hospital length of stay.Since services between 8 and 22 minutes are aggregated as a single unit, the observed rate of therapy is a rounded assessment of the actual delivered amount of therapy.

Assessment of SDOH
Given its previously reported associations with increased risk for functional decline following critical illness and reduced delivery of in-hospital rehabilitation to older adults, our primary exposure was dual eligibility for Medicare and Medicaid. 10,28We additionally explored associations of other SDOH that have been linked to rehabilitation delivery in other populations or settings [29][30][31] and were available in NHATS or Medicare claims data.SDOH assessed in this study included income, education, LEP, and rurality.Dual-eligibility for Medicare and Medicaid was assessed using the dual Medicare-Medicaid status indicator in the Medicare Master Beneficiary Summary File at any time during the year preceding the ICU hospitalization.Information on other SDOH was derived from the participant's NHATS survey immediately preceding ICU hospitalization.Income and assets were assessed using a composite of income from Social Security; Department of Veterans Affairs; pension; retirement plans; funds, stocks, and bonds; and checking and savings accounts and operationalized as quartiles in our sample. 23The only missing data were on household income; missing data were imputed using values provided by NHATS. 32Education was characterized as less than high school vs more.LEP was operationalized as a response of not well or not at all, as opposed to well or very well, to questions about how well respondents understand or speak English. 23,33Residence was classified as rural (nonmetropolitan) vs urban (metropolitan), as assigned by NHATS based on the Office of Management and Budget classification of county of residence.We considered but did not evaluate the exposure of race and ethnicity because of small proportions of participants who identified as a race other than Black or White in our sample.

Assessment of Covariates
We selected covariates that could be potential confounders in rehabilitation delivery based on prior research and clinical relevance.We included age categorized into intervals based on proportions in our sample (65-74, 75-79, 80-84, 85-89, and Ն90 years), sex, count of disabilities in the NHATS interview preceding ICU hospitalization (defined as need for help or inability to perform activities of daily living, including 4 self-care activities [eating, bathing, using the toilet, and dressing] and 3 mobility activities [getting outside, getting around inside one's home, and getting out of bed]), 10,34 use of mechanical ventilation (eTable 1 in Supplement 1), and severity of illness (determined as count of organ dysfunction). 27

JAMA Network Open | Critical Care Medicine
Social Determinants of Health and Delivery of Rehabilitation to Older Adults During ICU Hospitalization

Assembly of the Analytic Sample
Assembly of the analytic sample is presented in the eFigure in Supplement 1.We identified 2832 NHATS participants from 2011 to 2018 who had a hospitalization with an ICU admission for at least 1 day.Participants could contribute multiple observations; however, we restricted our sample to 2299 first ICU hospitalizations in the interval between consecutive annual NHATS interviews to allow updating model covariates.After excluding hospitalization from 681 participants who were not community-dwelling at the pre-ICU NHATS interview, our sample included 1618 ICU hospitalizations.

Statistical Analysis
We describe demographic and clinical characteristics of our sample using means and SDs or medians and IQRs for continuous variables and counts and weighted percentages for categorical variables, as appropriate.For each person-year of NHATS data, we used specific analytic weights that adjust for differential probabilities of selection and nonresponse within each strata (region) and cluster (zip code within county); this allows generalization to the 2011 Medicare population. 35,36For income, NHATS provided 5 imputed data sets that were used only in the models testing this exposure. 32ong other exposures considered in the models, only education had any missing data (0.7%).
Hence, our models were based on complete case data.We separately fit multivariable logistic regression models for the binary outcome of delivery of any in-hospital PT or OT on each of the 5 exposures with adjustment for covariates.For exposures with a significant association with this outcome, we calculated risk differences.For the rate of PT or OT delivered per day of hospital stay, we fit multivariable Poisson regression models on each of the exposures with adjustment for the same covariates.We calculated least square means of the outcomes significantly associated with this outcome.Because the provision of PT or OT could be prioritized to patients presumed to be discharging to a facility, we conducted sensitivity analyses excluding participants admitted from a nursing home or with a stay 100 days or more between pre-ICU NHATS interview and index ICU hospitalization.We used SAS software version 9.4 (SAS Institute) for descriptive analyses and SAS-callable SUDAAN software version 11 (RTI International) for all models.To account for the small number of participants who contributed multiple hospitalizations, we used generalized estimating equations with an exchangeable covariance structure based on its minimization of quasilikelihood under the independence model criterion.In all analyses, significance was defined as a 2-tailed P < .05.Data were analyzed from August 2022 to September 2023.
a Values represent characteristics for the unweighted sample.For income, imputed data sets provided by NHATS were used.Among the remaining variables, missingness was 0.8% for race and ethnicity and 0.7% for education.Among the individual indicators of frailty, missingness ranged from 0.1% to 1.9%.
b Categories represent a combination of race and ethnicity as self-reported in the NHATS interview.The category Hispanic includes participants of all races who reported Hispanic ethnicity; other includes participants reporting race as American Indian, Alaska Native, Asian, Native Hawaiian, Pacific Islander, or other race.
d The frailty score is derived from the composite of 1 point for each of the 5 frailty criteria (range, 0-5): weight loss, muscle weakness, exhaustion, slow gait speed, and low physical activity.
e Count of disabilities was characterized as the need for help or inability to perform 4 activities of daily living (eating, bathing, using the toilet, and dressing) and 3 mobility activities (getting outside, getting around inside one's home, getting out of bed) (range, 0-7).
f Ascertained from hospitalization record in linked Medicare claims data.residents of rural vs urban areas (Figure 2).For the rate of total therapy, LEP was associated with lower rates of in-hospital PT or OT compared with not having LEP (adjusted rate ratio [aRR], 0.55 [95% CI, 0.32-0.94])(Figure 1B).On the absolute scale, a mean rate of 0.7 (95% CI, 0.4-1.2) units/d of PT or OT was delivered to participants with LEP, compared with 1.3 (95% CI, 1.1-1.5)units/d to those proficient in English (Figure 3).Over a 5-day hospital stay, this would translate into 24 to 66 fewer minutes of therapy for patients with LEP compared with those proficient in English.Having an income between $12 000 and $22 000 was also significantly associated with a lower rate of therapy  delivered (aRR, 0.71 [95% CI, 0.52-0.95])compared with the highest quartile of income; the other income categories did not demonstrate significant associations (Figure 1B).
In sensitivity analyses excluding hospitalization for participants who were admitted from a nursing home or had a nursing home stay of at least 100 days between their pre-ICU NHATS interview and the index ICU hospitalization, the magnitude and direction of association between the exposures and both outcomes were similar, albeit with wider CIs (eTable 2 and eTable 3 in Supplement 1).

Discussion
In this nationally representative cohort study of older adults, we found that select SDOH were associated with reduced delivery of skilled rehabilitation services during hospitalization for critical illness.After accounting for prehospitalization disability and severity of acute illness, older adults who were dually eligible for Medicare and Medicaid and those who resided in rural areas had 30% to 35% lower odds of receiving any PT or OT during an ICU hospitalization than patients who were not dually eligible or who were proficient in English.Patients with LEP received lower amounts of therapy than patients proficient in English.Given the well-documented value of in-hospital rehabilitative PT and OT in preventing functional decline and identifying care needs at discharge following critical illness, [6][7][8][9] our findings of reduced delivery of these services to older adults by dual eligibility status, rural residence, and LEP warrant consideration of targeted efforts to mitigate inequities.
The prevalence of in-hospital rehabilitation delivery in our study was comparable with estimates from contemporary studies of acutely hospitalized patients. 28,37,38Among patients who are critically ill, clinician surveys 11,12,39 and cohort studies of hospitals participating in trials or quality reporting initiatives 14,16 have reported wide variability in the use of PT and OT.  more underresourced and may not be adequately staffed by professionals essential to delivering rehabilitation. 40Similar to our findings, dual Medicare and Medicaid eligibility was associated with a lower likelihood of skilled rehabilitation among patients with acute stroke. 28Second, percolation of evidence-based strategies, such as the ABCDEF bundle (assess, prevent, and manage pain; both spontaneous awakening and spontaneous breathing trials; choice of analgesia and sedation; assess, prevent, and manage delirium; early mobility and exercise; and family engagement and empowerment), 6,41 to promote rehabilitation may be lower at these hospitals.Study collaboratives promoting adoption of the bundle have typically engaged urban, academic hospitals 15 ; whether and how this guidance is translated into practice in rural hospitals should be explored.Third, the association between LEP and reduced PT and OT suggests that interpersonal communication barriers or implicit biases can underlie suboptimal delivery of rehabilitation, as described for other care processes. 20,42Consistent with our observations, in a 2020 study at a safety-net hospital in Texas, speaking a language other than English or Spanish was associated with fewer minutes of therapy delivered to older adults hospitalized with prolonged acute illness. 29Among patients who are critically ill, the reduced amount of rehabilitation services delivered to those with LEP could be due to differential prescription of rehabilitation services by physicians because of perceived barriers to engagement or differences in management of sedation and delirium that might influence a patient's ability to participate in rehabilitation. 17,19,20r findings have important implications.7][8] While mobilizing hospitalized patients who are acutely ill can be achieved by personnel other than rehabilitation therapists, the complex feasibility and safety considerations for older adults who are critically ill usually warrant an interdisciplinary approach. 13,43,44Evaluation by physical or occupational therapists is a part of recommended ICU mobilization protocols 6,7,45 and associated with its delivery in observational studies. 13,14Therefore, while not equivalent to it, the lower delivery of PT or OT services by SDOH in our study likely represents underlying differences in ICU mobilization.Second, evaluation by physical and occupational therapists, usually on the wards, is important to identify postacute rehabilitation needs before hospital discharge.Reduced delivery of any PT or OT during hospitalization suggests that postacute rehabilitation needs are assessed less frequently among older adults with Medicaid and those in rural areas.In-hospital skilled rehabilitation is associated not only with higher mobility at discharge 9,46,47 but also improved long-term function among adults recovering from critical illness. 48Therefore, reduced delivery of any rehabilitation during ICU hospitalization may represent a missed opportunity to improve long-term functional outcomes after critical illness and inequities in this practice can be a mechanism underlying disparities in post-ICU disability. 10r study has several strengths.First, we used a nationally representative sample of older adults with ICU hospitalization.Second, we included granular assessment of prehospitalization disability that could influence in-hospital rehabilitation needs.Third, we had information on SDOH using instruments specifically tailored for older adults, 23 Fourth, we used rigorous methods to determine delivery of PT and OT using claims data. 28,49,50

Limitations
Our findings should be interpreted in the context of a few limitations.First, we could not distinguish the contribution of hospital-level effects or evaluate hospital factors, such as supply of physical therapists, due to the limited number of observations per hospital and hospital-level information in our claims-linked NHATS data.This should be investigated in future work to understand the role of structural factors in driving differences in rehabilitation delivery.Second, we could not distinguish the delivery of PT and OT in the ICU from that delivered elsewhere in the hospital; future studies evaluating this could guide efforts to mitigate differences.Third, we could not evaluate nurse-driven mobility.While this practice exists, an interdisciplinary approach, including at least an evaluation by rehabilitation therapists, remains usual practice in most US ICUs, 13,44 suggesting that in-hospital

JAMA Network Open | Critical Care Medicine
Social Determinants of Health and Delivery of Rehabilitation to Older Adults During ICU Hospitalization skilled rehabilitation, although not a surrogate for ICU mobilization, is likely associated with differences in this practice.Fourth, we did not have information on severity of illness scores, such as the SOFA score and Acute Physiology and Chronic Health Evaluation scale score, which are known to be associated with rehabilitation in the ICU. 51Nevertheless we used a validated claims-based organ dysfunction algorithm to account for severity of illness. 27Fifth, because of the small number of participants with self-reported race and ethnicity other than non-Hispanic Black or non-Hispanic White in our sample, we could not evaluate it as an exposure; future work should investigate this.
Furthermore, because LEP and rurality were evaluated as exploratory exposures in our study, the observed associations should be further investigated in future studies.

Conclusions
In this nationally representative cohort study of older adults, dual eligibility for Medicare and Medicaid and rural residency were associated with lower likelihood and LEP was associated with reduced rate of delivery of skilled rehabilitation therapy during hospitalization with critical illness.
Our findings highlight the need to consider these SDOH in efforts to enhance equitable delivery of skilled rehabilitation services during hospitalization.Future research is needed to distinguish individual-vs structural-level factors underlying differences in in-hospital rehabilitation delivery by SDOH.

Figure 2 .B
Figure 2. Absolute Risk Differences in the Adjusted Percentages of Intensive Care Unit Hospitalizations With Delivery of Any Physical Therapy (PT) or Occupational Therapy (OT) Despite this known variability, to our knowledge, prior work has not evaluated the role of SDOH in the delivery of skilled rehabilitation services among Medicare beneficiaries.The availability of patient-level information on SDOH in NHATS, beyond those usually available in administrative data, allowed us to evaluate this important question.Our findings of reduced delivery of skilled rehabilitation to older adults who are dually eligible for Medicare and Medicaid, live in rural areas, or have LEP may have explanations rooted in structural-and individual-level factors underlying in-hospital care delivery.First, factors related to resources and organization of rehabilitation services at hospitals deserve consideration.Staffing by physical therapists and nurses is associated with rehabilitation among patients who are critically ill.

Table .
Characteristics of ICU Hospitalizations Overall, and by Delivery of PT or OT a 11,18Hospitals serving more dually eligible patients and those located in rural areas are generally