Racial and ethnic disparities in post-acute care service utilization after Stroke

Abstract

Evidence examining disparities in post-acute care (PAC) utilization among various racial and ethnic groups after stroke and the influence of social determinants of health (SDOH) in these decisions is lacking. Thus, we searched the literature from January 2000 to November 2023 regarding PAC among individuals after stroke through: 1) Pubmed, 2) Scopus, 3) Web of Science, 4) Embase, and 5) CINAHL. We found 14 studies. Black individuals were more likely than White individuals to be discharged home with home health (HH) and skilled nursing facilities (SNF). Hispanic individuals were more likely than White individuals to be discharged home with HH, but less likely to be discharged to institutions. Lower socioeconomic status, Medicaid insurance, urban residence, area PAC supply and hospital characteristics were associated with increased institutional discharges among racial and ethnic minority individuals. Future policy should improve access to appropriate PAC commensurate with an individual’s medical/social complexity.

Introduction

Stroke is a leading cause of death,^1,2 affecting approximately 800,000 United States (U.S.) adults annually.³ Between 2011 and 2021, the number of stroke deaths increased from 128,932 to 162,890 and strokes remain one of the 10 most common reasons for hospitalizations in the U.S. Appropriate care during the post-acute care (PAC) period is vital for older adults who are generally affected by this acute illness and has been shown to decrease readmissions, falls, and mortality following hospital discharge after a stroke.^9–11

Despite the increasing diversity in the U.S.,¹² including in rural areas,¹³ racial and ethnic minority individuals continue to face disparities after having a stroke. Researchers have demonstrated that individuals from these minority groups have a higher risk of stroke and experience worse outcomes after a stroke, such as a recurrent stroke or death, compared to their White counterparts.^4,14 For instance, Black individuals have two times the risk of stroke compared to White individuals¹⁵ and the prevalence of stroke has increased significantly among racial and ethnic minorities in the past decade.¹⁶ Additionally, racial and ethnic minorities experience decreased rates of treatment and longer time-to-consult,¹⁵ resulting in worse outcomes and possibly death.

Health outcomes following acute illness among racial and ethnic minority individuals¹⁴ are frequently affected by social determinants of health (SDOH), such as access to care, employment, and living conditions.¹⁷ Researchers have previously highlighted the role of SES,¹⁸ insurance,¹⁹ education,¹⁸ geographic factors,^18,19 and PAC availability¹⁸ on discharge destinations following stroke. While the mechanism of how these factors interact with one another is unclear, appropriate utilization of PAC services can improve the effects of complex medical and social conditions inherent in this population.^9,10,18,20

Considering the critical role of SDOH in health access and outcomes among racial and ethnic minority individuals,^4,21 a review of the current evidence, with a focus on SDOH, may enhance our understanding of PAC utilization among these underserved populations after a stroke.

Previous reviews investigating factors associated with discharge destinations among patients hospitalized for stroke have shown mixed findings.^18,19 For instance, Van der Cruyssen et al (2014)¹⁹ found that Black individuals were more likely to be discharged to institutional settings compared to White individuals, whereas Mees et al (2016)¹⁸ found that Hispanic individuals were more likely to be discharged to home healthcare (HH) compared to White individuals. There was conflicting evidence with Black individuals either being more likely to be discharged to HH or skilled nursing facility (SNF) than White individuals. Notably, these reviews did not examine these disparities in discharge decisions in the context of SDOH.

Thus, this review aimed to fill this gap. Guided by the SDOH framework, which encompasses political, geographic, financial, and societal factors that drive health disparities,²² we examined disparities in PAC utilization among racial and ethnic minority individuals who were hospitalized for stroke.

Methods

Data Sources and Search Strategy

We used the updated Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines.²³ We conducted searches for stroke on November 8, 2023, combining key search term domains (e.g, stroke, discharge, post-acute care, disparities), in consultation with a trained informationist, using the following electronic databases: 1) MEDLINE (PubMed), 2) Scopus, 3) Web of Science, 4) Embase, and 5) Cumulative Index to Nursing and Allied Health Literature (CINAHL) (EBSCOhost). The original review was registered in Prospero (CRD42022306199), an international prospective register of systematic reviews, and the full search strategy is provided in Appendix 1. We used Covidence, a web-based systematic review management platform, to screen publications.

Eligibility Criteria

Studies eligible for inclusion were peer-reviewed, English-language publications that examined racial and ethnic disparities in PAC utilization after being hospitalized for stroke in the U.S. published between January 1, 2000, to November 8, 2023. We limited our search to studies after 2000 due to advancements in stroke care and critical changes in healthcare policy. Specifically, the introduction of the Prospective Payment System for post-acute care influenced financial incentives and shaped discharge destinations, which made post-2000 data/publications more relevant to current practices. We excluded articles that focused on non-U.S. populations, pediatric populations, and non-acute care hospitalizations. We excluded commentaries, case studies, abstracts, and literature reviews.

Screening and Data Extraction

Each step of the screening process was performed independently by at least two team members (Authors 1,2,3). Discrepancies were resolved through team discussion. Additionally, a team of research field (health disparities) and clinician experts provided confirmation of included studies. Then, data was extracted by a single team member with the first author verifying data for accuracy. We abstracted: 1) authors, 2) publication year, 3) sample size, 4) study design, 5) years of data collection, 6) whether the studies included time trends, racial and ethnic and/or urban/rural disparities, 7) datasets used, 8) sample characteristics, 9) discharge disposition type (i.e., home, home health (HH), skilled nursing facility (SNF)/Nursing home (NH), and/or inpatient rehabilitation facility (IRF)), 10) statistical analysis method, 11) relevant findings, and 12) limitations.

Risk of Bias Assessment

Three team members independently performed quality assessments of each study using the Joanna Briggs Institute (JBI) quality assessment tool and checklist for Cohort (11 questions) studies and Cross-Sectional (eight questions) studies²⁴. Responses of “yes” (1) or “no” (0) were recorded, ranging from 0–11 and 0–8, respectively; higher scores indicate higher quality studies. Discrepancies were resolved by team discussion.

Conceptual Framework and Data Synthesis

We used the SDOH framework²² to guide our examination of disparities related to PAC utilization among racial and ethnic minorities among included studies. This framework attempts to explain disparities in cardiovascular disease as being influenced by financial, geographic, societal, and political factors.²² As such, the following constructs are depicted to influence the PAC trajectory: 1) economic stability (i.e., income, employment status, wealth, or occupational category), 2) built environment (i.e., neighborhood crime and poverty, walkability, access to public transportation, pollutants, availability of green space, housing/food insecurity), 3) education (educational attainment, literacy), 4) food (i.e., access to healthy food, availability of groceries, higher fast food outlets), 5) community/social context (i.e., social support, network, and cohesion, community engagement, discrimination), and 6) healthcare system (i.e., hospital disproportionate share status, implicit provider attitudes).²² In all included studies, we focused on all constructs and corresponding variables from the SDOH framework.

Results

Study Selection

Figure 1 depicts the PRISMA flow diagram. Our searches for stroke-related discharge disposition studies yielded 424 articles, from which we removed six duplicates. Screening of abstracts and titles eliminated 363 studies, leaving 55 studies for full-text review. At this stage, we excluded 41 studies for various reasons as below.

Fig 1.

PRISMA Diagram

Source: Page MJ, et al. BMJ 2021;372:n71. doi: 10.1136/bmj.n71.

This work is licensed under CC BY 4.0. To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/

Study Characteristics

Table 1 outlines study characteristics. Included studies were published from 2006 to 2022. Sample sizes ranged from 72 to 1,322,162 with mean ages of 52 to 80 years. Study designs included cohort (n=5),^25–29 cross-sectional (n=8),^30–37 and experimental (n=1).³⁸ Data sources included data from 1996 to 2020 and were from the following: 1) national database,^27,29,34,37 including Medicare^28,33 (n=6), 2) state database (n=6)^{25,26,30–32,39} (3) multi-centers (n=3)^26,30,33 and 4) single center (n=2).^35,36 For statistical analyses, multivariable regression was used in all studies (n=14).^25–39 Regarding types of stroke, 13 studies included stroke classifications as the following: 1) ischemic only (n=4)^27,33,36,37, 2) hemorrhagic only (n=1)³⁴, 3) ischemic and hemorrhagic (n=5)^{25,26,31,32,35}, and 4) ischemic, hemorrhagic and other (n=2)^29,30. The following PAC settings were studied: home (n=8),^{26,29–34,38} 2) HH (n=5),^{29–31,33,38} 3) SNF/NH (n=11),^{25,27–33,35,37,38} and 4) IRF (n=12).^{25,26,28–36,38}

Table 1.

Summary of each study (n=14)

Author, year	Sample Size	Study design			Data years	Database					Statistical analysis			Home	HH	SNF/NH	IRF
		CH	CS	Other		Medicare	National	State	Multi-site	Single Inst	REGR	PSM	Other
Gregory, 2006	n=12,208		•		2000			•			•			•		•	•
Gregory, 2009	n=7,810		•		2002–2004			•			•			•	•	•	•
Sandel, 2009	n=11,119	•			1996–2003			•			•					•	•
Kind, 2010	n=60,164		•		1999–2000	•			•		•			•	•	•	•
Freburger, 2011	n=187,188		•		2005–2006			•	•		•			•	•	•	•
Holmes, 2012	n=187,188 (stroke only)	•			2005–2006			•	•		•			•		•	•
Jaja, 2013	n=31,631		•		2005–2010		•				•			•			•
Agarwal, 2015	n=775,905	•			2003–2011		•				•					•
Bettger, 2015	n=849,780	•			2003–2011		•				•		Cochrane-Armitage test	•	•	•	•
Morgenstern, 2017	n=72			Pilot	2015			•			•			•	•	•	•
Mehta, 2021	n=1,322, 162		•		2006–2016		•				•					•
Thau, 2021	n=507		•		2019–2020					•	•						•
Chavez, 2022	n=1,717		•		2016–2019					•	•					•	•
Kim, 2022	n=543,556	•			2012–2016	•					•	•				•	•

^*Notes: CH=Cohort; CS=Cross-sectional; REGR=Regression; PSM=Propensity Score Matching; HH=Home Health; SNF/NH=Skilled Nursing Facility/Nursing Home; IRF=Inpatient Rehabilitation Facility

Table 2 provides study details for each study. Table 3 and 4 present quality appraisals. Of the cohort studies,^25–29 three studies scored eight out of 11^26,27,29 and two studies scored 11 out of 11.^25,28 Of the cross-sectional studies, most studies score eight out of eight^30–34,37 and two studies scored seven out of eight.^35,36 We included all studies regardless of their JBI scores to provide a comprehensive review of literature of this topic.

Table 2.

Detailed summary of each study (n=14)

Author, year	Database	Sample characteristics	Statistical analysis	Relevant findings	Limitations
Gregory, 2006	Maryland Health Services and Cost Review Commission database	Blacks (n=3,144, 26%): Age = 66+/−14.9 years, Females=59% Whites (n=9,064, 74%): Age = 73+/−12.5 years, Females=56%	Logistic regression	Blacks > Whites were more likely to be discharged to NH or IRF; Blacks were less likely to be discharge to SNF or home In terms of urban/rural, Blacks = Whites discharges to SNF; both urban Blacks and Whites had increased IRF discharges; for rural populations, there were no differences by race in disposition status; rural vs urban populations had 1/2 times lower rates of discharges to IRF; in rural settings, Blacks > Whites had lower home and SNF discharges; compared with urban-dwelling Whites, urban-dwelling Blacks were more likely to be discharged to IRF	Unable to measure stroke severity; no information on other determinants of IRF such as education or income
Gregory, 2009	State hospital inpatient discharge data	Blacks (n=1,688): Age >=60 years, 51%; Females, 57% Other (n=319): Age >=60 years, 41%; Females, 52% Whites (n=5,803): Age >=60 years,71%; Females, 48%	Logistic regression	Overall trends showed that there were 7% IRF discharges, 22% SNF discharges, and 72% home discharges; no significant differences in discharge patterns regarding race; IRF discharges were associated with urban location and less poverty; home discharges were associated with rural location and poor county; home discharges were associated with living in a poor county or Whites residing in areas of rich counties; others had two times higher likelihood of IRF discharges; however, others that lived in high poverty counties had lower likelihood of IRF discharges	High level of missing data for race (33%), which needed to be imputed; other unmeasured confounders (i.e., education, income, marital status); no information on stroke severity
Sandel, 2009	Kaiser Permanente Northern California (KPNC) claims database, California state mortality data, and US census data	White (n=8,038, 70.2%): Age = 71.6+/−12.3 years, Females = 51.6% Black (n=1,232, 10.8%): Age = 65.4+/−13.2 years, Females = 53.9% Asian (n=1,006, 8.8%): Age = 63.7+/−12.9 years, Females = 47.2% Hispanic (n=843, 7.4%): Age = 64.9+/−13.9 years, Females = 46.1%	Logistic regression	From 1996 to 2003, the following changes in discharges occurred: 1) increase in IRF (2.4% vs 4.3%), 2) decrease in SNF (43.6% vs 32.6%), 3) decrease in HH (18.9% vs 13.5%), and 4) increase in home (33.7% vs 46.6%) Hispanics and Whites had similar patterns of PAC discharges; Blacks and Asians had greater level of PAC than Whites; Asian and Blacks were more likely than Whites to be discharged to IRF. White females were more likely to be treated in SNFs; Those more likely to go to an IRH were individuals from higher socioeconomic groups, urban areas, and geographic areas close to the hospital	Findings not generalizable for fee-for-service patients or those from other geographic areas; lacking variables (i.e., stroke severity or comorbidities) or missing data (i.e., lacking ethnic designations) may have resulted in misclassification
Kind, 2010	Medicare claims (Fee-for-service data), HMO data from a large national managed care organization, Census 2000 data	Whites (n=52,306): Age = 80+/−7.4 years, Males= 39% Black (n=9015): Age = 78+/8 years, Males = 34% Hispanic (n=2268): Age=79+/−7.5 years, Males = 39%	Predicted probabilities	Blacks vs Whites had higher discharges to HH (21% vs 16%, respectively) and lower discharges to SNFs (26% vs 33%, respectively); Hispanics vs Whites had higher discharges to HH (19% s 16%, respectively) and lower discharges to SNFs (19% vs 16%, respectively); Blacks, Hispanics, and Whites had similarly likely to be discharged home and to IRFs	Lack of stroke severity, social support measures, poststroke functionality
Freburger, 2011	State Inpatient Databases (hospitaldischarge data from short-term acute-care hospitals in 4 states: AZ, FL, NJ, WI), AHA 2006 Annual Survey Database, CMS 2006 Provider of Services File, CMS 2006 Hospital Costs Reports, 2006 Demographic Update of the Census 2000 data, 2006 Area Resource File	Mean age: 72.6+/−11.9 years, Females: 52.4%, Whites: 79.5%, Blacks: 11.4%, Hispanics: 9.1%	Logistical regression	Blacks were more likely to receive institutional care; Hispanics and racial/ethnic minorities on Medicaid were less likely to receive institutional care; no geographic differences were observed among different race/ethnicities for institutional care receipt Racial/ethnic minorities were more likely to receive HH. Rural residing individuals were less likely to receive HH; racial/ethnic minorities vs Whites in Wisconsin were more likely to receive HH Blacks, urban-dwelling individuals, and Florida -residing individuals were more likely to receive SNF care; New Jersey residents were more likely to receive IRF care; Racial/ethnic minorities vs Whites in Wisconsin were more likely to receive IRF care	Findings not generalizable to states not included in this study; no information on outpatient therapy, functional status, patient preference, provider characteristics
Holmes, 2012	State Inpatient Databases (4 states: AZ, FL, NJ, WI), AHA 2006, CMS 2006 Provider of services file, Census 2000 data, 2006 Area Resource File	White: Mean age = 73.7 years, Blacks: Mean age = 67.1 years, Hispanics: Mean age = 70.3 years	Logistic regression, Blinder-Oaxaca decomposition approach, matching	About 28.12 percent of stroke patients received institutional care; in terms of race, Black > White > Hispanic received institutional care; Blacks vs Whites were more likely to use institutional PAC; after adjusting for gender, the disparities between Blacks vs Whites decreases; geography was a significant factor in racial/ethnic differences in PAC use; Whites were treated in hospitals that had a lower tendency of sending patients to institutions Percent discharged to SNF vs IRF: White > Black > Hispanic; Age is a particularly important factor in explaining differences in White vs Black rate of SNF vs IRF; hospital characteristics contributed to the White-Hispanic disparity; SES had modest effect; PAC supply was significant but had only minimal effect; after adjustment, Whites were less likely to be discharged to SNF vs IRF and racial/ethnic minorities are more likely to be discharged to SNF vs IRF For discharge to institution vs home: observed characteristics (Table 3) explain only 29% of difference in institutional discharge between Black/White, 42% difference between White/Hispanic (i.e., 71% of the difference in PAC/discharge to institution vs. home between Black/White is unexplained; these could be cultural differences in preferences for care, unobserved social differences (e.g., social support), differences in condition severity, or discrimination PAC supply in the community had the least effect; SES and PAC supply had modest effect on racial/ethnic disparities in PAC use	Limited to four states (e.g., NJ had higher rate of referral to institution compared to other three states; if it’s an outlier, could limit generalizability); limited data on functional status which is an important determinant of PAC; LOS may be confounded by condition severity or could reflect waiting time for institutional PAC; limitations in variable measurement given datasets
Jaja, 2013	Nationwide Inpatient Sample	White (n=15,376): Age = 60.27+/−16.57, Females = 62% Blacks (n=3,412): Age = 53+/−15.42, Females = 65% Hispanic (n=3,133): Age = 52.46+/−17.15, Females = 58% Asian/Pacific Islander (n=1,098): Age = 59.80+/17.79, Females = 65% Native American/Other (n=1,069): Age = 54.85+/17.19, Females = 59%	Multinomial logistic regression	Overall, 42% of patients were discharged to an institution; Race/ethnicity were significantly associated with institutional discharge (p<0.001); Blacks vs Whites were more likely to be discharged to institutions (OR 1.27; 95% CI: 1.14–1.40); no significant difference seen for Hispanics, Asian/Pacific Islanders, and Native Americans/Others	did not adjust for subarachnoid hemorrhage severity; multiple imputation used for 20% of data because some states do not submit race information to the database, which may limit generalizability
Agarwal, 2015	Nationwide Inpatient Sample	SES Quartile 1: Mean age=69.8+/−14.6 years, Females = 54.6% SES Quartile 2: Mean age=71.5+/−14.5 years, Females = 54% SES Quartile 3: Mean age=72.1+/−14.4 years, Females = 53.7% SES Quartile 4: Mean age=73.3+/−14.4 years, Females = 53.2%	Multivariable hierarchical logistic regression	Significant trend toward reduced discharge to nursing facilities among those from high-income SES quartiles compared with those from low-income SES quartiles (P-trend<0.001); no evidence of interaction between race and SES categories Progressive decrease in the adjusted odds of discharge to nursing facilities across SES quartiles in each race subgroup (i.e., White, Nonwhite) In the Discussion they argue that, while race is a convenient way to label people and previous literature has evaluated the impact of racial disparities on outcomes after stroke, factors related to SES (income, education, housing) are probably more important in health-related outcomes	The database subject to error in delineating admission and patient status so patients may have been represented more than once; subject to traditional bias of retrospective observational studies (e.g., selection bias); outcomes may be affected by variables not consistently available in the database (e.g., stroke severity, prehospital stroke care delivery); used median household income of entire ZIP code to “impute” the SES of each patient
Bettger, 2015	National (i.e., hospitals participating in Get With The Guidelines-Stroke initiative), AHA	Home (n=371,092, 43.7%): Age = 64.3+/−14.5 years, Females = 45.6%, Race = White (68.9%), Black (16.4%), Asian (2.8%), Hispanic (7.0%), Other (4.7%) Home health (n=97,471, 11.5%): Age= 71.9+/−13.6 years, Females = 55.8%, Race = White (68.1%), Black (17.9%), Asian (2.7%), Hispanic (7.2%), Other (3.9%) SNF (n=165,411, 19.5%): Age = 77.8+/−12 years, Females = 60.3%, Race = White (73.2%), Black (14.4%), Asian (2.8%), Hispanic (5.2%), Other (4.3%) IRF (n=215,806, 25.4%): Age = 69.9+/−13.9 years, Females = 50.9%, Race = White (70.1%), Black (17.9%), Asian (2.6%), Hispanic (5.5%), Other (3.8%)	Logistic regression, Cochran-Armitage test to assess trends	PAC use over 8 years rose 2.1% overall in unadjusted analyses; increase in IRF discharges (6.9%) and HH discharges (3.6%); decrease in SNF discharges by 8.3%; PAC use increased for all types of stroke severity, but higher severity was associated with PAC use (the greatest change over the 8-year period was for patients with severe stroke, who had a 9.4% mean increase to IRF and 12.4% decrease to SNF) In the adjusted analysis, PAC use did not increase per year (OR 1.00, 95% CI: 0.99–1.00) Regarding race/ethnicity, both Whites (OR 1.20, 95% CI 1.13–1.28) and Blacks patients (OR 1.34, 95% CI 1.24–1.45) were more likely to be discharged to any PAC (IRF, SNF, or HH) vs others; there were no statistically significantly different odds of PAC use among urban vs rural stroke patients	Data from a clinical registry, hospital characteristics obtained from public sources; their smallest geographical unit of analysis was multistate region, and previous studies found that PAC availability varies at the county level; hospitals that participate in GWTG-Stroke are more often larger/urban, may not represent overall US hospital population; NIhSs assessment was often missing; other factors influencing discharge disposition not available in these data
Morgenstern, 2017	State population-based stroke study (Brain Attack Surveillance, Corpus Christi, Texas)	Whites: Age, median = 71 years, Females = 46% Hispanics: Age, median = 63 years, Females = 50%	Logistic regression	48/72 (67%) received any type of rehab (classification of rehab service in this study was IRF, SNF, long-term acute care, outpatient rehab, home rehab) 25% of all patients discharged to IRF No significant differences between MA and Whites in terms of discharge with vs without rehab Among those who had rehab, Whites (100%) were more likely to go to IRF or SNF compared with MA (49%), who were more likely to have rehab as outpatient or home (p=0.002) After adjusting for age and Medicare insurance by dichotomizing the group to <65 and >65, Whites were had 4.6 higher odds (95% CI: 0.94–22.5) of being discharged to IRF compared to MAs	A pilot study and was underpowered to control for important covariates (e.g., insurance status, stroke severity)
Mehta, 2021	Nationwide Inpatient Sample	Age: 72 (61.0–82.0) years Females: 52.2% Race: White (70.2%), Black (16.4%), Hispanic (7.6%), Asian (2.4%), Native American (0.6%), Other (2.7%)	Logistic regression	Black race demonstrated higher odds of PAC facility discharge (OR=1.328, 95% CI: 1.298–1.358)	No information on disease severity or type of disease; coding errors and misclassifications; no readmission data so lacking long-term outcomes
Thau, 2021	Single-center	Age: 67 (57–77) years, Females: 44.5%, White: 52.3%,	Logistic regression	There was 9% decrease in IRF discharges during COVID-19 periods compared to pre-periods, which translated to 38% decreased odds of IRF versus home discharge; discharge rates to IRF were higher in May, 2020 then declined; White race was associated with IRF versus home discharge	Single-center study limits generalizability; did not evaluate effects of insurance on discharge disposition/planning
Chavez, 2022	Single-center	Home (n=863): 63.51 +/− 15.01 years, males 48.67%, race (White, 53.88%; Hispanic, 9.97%; Black, 22.48%; Asian, 2.32%; Other, 11.36%) IRF (n=663): 63.14 +/− 14.54 years, males 49.32%, race (White, 53.39%; Hispanic, 11.46%; Black, 23.68%; Asian, 2.87%; Other, 8.75%) SNF (n=191): 72.11 +/− 13.4 years, males 40.31%, race (White, 51.31%; Hispanic, 9.42%; Black, 27.23%; Asian, 0%; Other, 12.04%)	Lognormal and multinomial regression	Hispanic versus White individuals had higher SNF discharges (p=0.01); no statistically significant differences for Black and Asian versus White patients for SNF	Small sample of Asian individuals; more urban versus rural representation, which limits generalizability
Kim, 2022	Medicare Provider Analysis and Review files (fee-for-service)	MSSP: Age, 80.5 years; Males, 41.0%; Race (White, 84.3%; Black, 11.5%; Other, 4.2%) Non-MSSP: Age, 80.4 years; Males, 42.6%; Race (White, 84.4%; Black, 11.1%; Other, 4.5%)	PSM, logistic regression, linear probability models	The use of PAC facility was higher in MSSP vs non-MSSP hospitals; MSSP participation after the 3rd year was associated with increased discharges to PAC facilities of 1.5% (95% CI: 0.00–0.3) vs non-MSSP hospitals; racial/minorities had higher PAC facility use compared to White individuals	Limited to metropolitan areas; administrative data, which lacks individual-level clinical and SES information

Notes of abbreviations: Intensive care unit (ICU); propensity score matching (PSM); home health (HH); skilled nursing facility (SNF); nursing home (NH); inpatient rehabilitation facility (IRF); post-acute care (PAC); Mexican Americans (MA); Arizona (AZ); Florida (FL); New Jersey (NJ); Wisconsin (WI)

Table 3.

Quality Assessments for Cohort Studies (n=5)

Author, year	Groups: similar-recruited from same population?	Exposures: measured similarly?	Exposures: measure d in a valid & reliable way?	Confounding factors: identified?	Confounding factors: strategies to deal with them stated?	Participan ts: free of the outcome at start of the study?	Outcomes: measured in a valid & reliable way?	Follow-up: sufficient?	Follow-up: complete? Reasons of loss described & explored?	Incomplete follow-up: strategies utilized?	Statisticalanalysis: appropriate?	JBI Total (max=11)
Sandel, 2009	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	11
Holmes, 2012	Yes	Yes	Yes	Yes	Yes	Yes	Yes	N/A	N/A	N/A	Yes	8
Agarwal, 2015	Yes	Yes	Yes	Yes	Yes	Yes	Yes	N/A	N/A	N/A	Yes	8
Bettger, 2015	Yes	Yes	Yes	Yes	Yes	Yes	Yes	N/A	N/A	N/A	Yes	8
Kim, 2022	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	11

Table 4.

Quality Assessments for Cross Sectional Studies (n=8)

Author, year	Inclusion criteria: clearly defined?	Study subjects & setting: described in detail?	Exposure: measured in a valid & reliable way?	Measurement of condition: objective, standard criteria used?	Confounding factors: identified?	Confounding factors: were strategies to deal with stated?	Outcomes: measured in a valid & reliable way?	Statistical analysis: appropriate usage?	JBI Total, Max=8
Gregory, 2006	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	8
Gregory, 2009	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	8
Kind, 2010	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	8
Freburger, 2011	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	8
Jaja, 2013	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	8
Mehta, 2021	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	8
Thau, 2021	Yes	Yes	Yes	Yes	Yes	N/A	Yes	Yes	7
Chavez, 2022	Yes	Yes	No	Yes	Yes	Yes	Yes	Yes	7

Disparities in PAC utilization

Table 5 outlines racial and ethnic differences in PAC utilization. The following sections are divided into the following sequence according to the intensity of PAC: 1) home, 2) HH, 3) SNF/NH, and 4) IRF.

Table 5.

Racial/ethnic Disparities and Social Determinants of Health in each study (n=14)

		Racial/Ethnic Disparities				Social Determinants of Health
Author, year	Racial/ethnic category	Home	HH	SNF/NH	IRF	Economic stability	Built Environment	Education	Community/Social Context	Healthcare system	Food	Analysis
Gregory, 2006	Blacks (B), Whites (W)	W>B		W>B	B>W	Insurance status	Urban/rural status		Marital status			***
Gregory, 2009	Blacks (B), Others (O) (American Indians, Asian/Pacific Islanders), Whites (W)	W>B; W>O		B=W;O=W	W>O*	Insurance status	Residential poverty**; urban/rural status**					Used binary logit model to examine independent effect of poverty on discharge destination; used an interaction term for race and urban/rural OR poverty level
Sandel, 2009	Blacks (B), Hispanics (H), Asians (A), Whites (W)			B>W; A>W H=W	A>W B=W; H=W		Urban/rural status; median household income level	Education level				***
Kind, 2010	Blacks (B), Hispanics (H), Whites (W)	B=W; H=W	B>W; H>W	W>B H=W	B=W; H=W	Insurance status; Medicaid/Medicare	Percentage below poverty	Percentage of >24 years with college degree				***
Freburger, 2011	Blacks (B), Hispanics (H), Whites (W)		B>W; H>W	B>W W>H	B>W W>H	Socioeconomic status**; uninsured or Medicaid status	Metropolitan status of residence (large, medium/sm all, micro/non-micropolita n)**; median household income			Hospital stroke volume, teaching affiliation, nurse, physical/occupational/speech therapist fulltime equivalents for 1000 admissions; profit status		Included a model with an interaction between race, SES, and geographic (metropolitan) variables
Holmes, 2012	Blacks (B), Hispanics (H), Whites (W)	W>H		B>W; H>W	B>W; H>W	Insurance status	Median household income; metropolita n status (large, medium, micro/nonm etropolitan)			Hospital stroke volume, teaching/profit status, nurse/therapist full-time equivalents; rehabilitation facility affiliation; number of physical/occup ational therapists, home health agencies, number of skilled nursing facility beds, number of inpatient rehabilitation facilities per county population		Decom position approach was used by performing regression for each racial/ethnic subgroups
Jaja, 2013	Blacks (B), Hispanics (H), Asians (A), Native Americ ans (NA), Whites (W)		B>W; A>W H=W NA=W	B>W; A>W H=W NA=W	B>W; A>W H=W NA=W	Insurance status	Neighborhood median income			Hospital bed size (large, medium, small); teaching status		***
Agarwal, 2015	Non-Whites (NW), Whites (W)			NW=W		Insurance status	Location (urban/rural), median income**			Hospital bed size (large, medium, small); teaching/owner ship status		Included the interaction between race and median income
Bettger, 2015	Blacks (B), Others (O)+, Whites (W)	O>B; O>W	B>O;W>O	B>O;W>O	B>O;W>O	Insurance status	Urban/rural status			Number of hospital beds, teaching status		***
Morgenstern, 2017	Hispanics (H), Whites (W)	H>W	H>W	W>H	W>H
Mehta, 2021	Blacks (B), Hispanics (H), Asians (A), Native Americans (NA), Others (O), Whites (W)			B>W	H>W; NA>W W>B; W>A; W>O	Insurance status	Median household income			Hospital size (large, medium, small)		***
Thau, 2021	Blacks (B), Whites (W)				W>B
Chavez, 2022	Blacks (B), Hispanics (H), Asians/Others (A/O), Whites (W)			H>W B=W; A/O=W	B=W; H=W; A/O=W	Insurance status, employment			Marital status			***
Kim, 2022	Blacks (B), Others (O)+, Whites (W)			B>W; O>W	B>W; O>W	Insurance status**; Medicaid/Medicare;				Participation of Medicare Shared Savings Plan**; number of hospital beds; profit/teaching status; nursing staff per daily census; hospital case mix; disproportionat e patient percentage		Included interaction terms between Medicare Shared Saving Program status/hospital, race/ethnicity, and insurance status

Bolded when racial/ethnic minority groups have higher probability/odds of discharge to certain destinations;

^*denotes for conditions of poverty; +denotes undetermined;

^**denotes inclusion of interaction terms with race/ethnicity status;

^***SDOH variables included as covariates

Home

Six studies examined racial and ethnic differences in home discharges after stroke.^{26,29–33,38} Except one study that found Hispanic individuals more likely to go home than White individuals,³⁸ most studies found White individuals to have higher home discharges than other racial and ethnic groups (i.e., Black and/or Hispanic individuals).^26,29,31,32

HH

Five studies examined racial and ethnic differences in HH discharges.^{29,30,33,34,38} In four studies that included Black individuals, all studies showed that Black individuals were more likely than White individuals to go home with HH.^29,30,33,34 In four studies that included Hispanic individuals,^30,33,34,38 most studies showed that Hispanic individuals were more likely than White individuals to go home with HH.^30,33,38

SNF/NH

Thirteen studies examined racial and ethnic differences in SNF/NH discharges.^25–35,38 Among 12 studies that included Black individuals,^{25,26,28–37} more than half of the studies demonstrated that Black individuals were more likely to be discharged to SNF/NH than White individuals.^{25,26,28–30,34,37} However, among seven studies that included Hispanic individuals,^{25,26,30,33–35,37,38} only two studies showed that Hispanic individuals are more likely to be discharged to SNF/NH than White individuals.^26,35 Among four studies that included Asian individuals,^25,34,35,37 only one study showed that Asian individuals are more likely than White individuals to be discharged to SNF/NH.³⁴

IRF

Thirteen studies investigated racial and ethnic differences in IRF discharges.^{25,26,28–38} Among 12 studies that included Black individuals,^{25,26,28–37} half of the studies demonstrated that Black individuals are more likely to be discharged to IRF than White individuals.^{26,28–30,32,34} Among eight studies that included Hispanic individuals,^{25,26,30,33–35,37,38} only two studies showed that Hispanic individuals are more likely than White individuals to be discharged to IRF.^26,37

Racial and ethnic disparities in PAC utilization in the context of SDOH

Table 5 shows SDOH variables included in the studies. Most studies considered SDOH (n=12).^25–35,37 All of these studies included health insurance status as a variable representing economic stability,^25–35,37 with one study also including employment status.³⁵ Urban/rural status,^25,27,29,32 metropolitan status,^26,30 and neighborhood poverty^{25–27,30,33,34,37} were included as variables representing the built environment construct. Level of education was only considered in two studies.^25,33 Marital status was considered as a variable representing community/social context in only two studies.^32,35 Hospital volume^26,30/beds,^{27–29,34,37} teaching status,^{26,27,29,30,34} profit status,^26,30 availability of healthcare providers (i.e., nurses, therapist)^26,28,30 and facilities²⁶ were included as variables representing the healthcare system construct. Variables representing the construct of food were not considered in any studies. All studies controlled for included SDOH variables, and less than half of studies (n=5) considered interactions of race and other characteristics.^{26–28,30,31} Four studies included an interaction of SES and race and ethnic status^27,28,30,31 and two studies included an interaction of urban/rural status and race and ethnic status.^30,31

In some of these studies, racial and ethnic minority individuals were of lower income (Blacks^25,31,34,37 and Hispanics^25,31,37),³³ low education levels (Blacks and Hispanics),³³ rural residence (Hispanics),²⁵ or urban residence (Blacks).³² In the context of SDOH, studies demonstrated that the following were determinants of institutional PAC utilization: 1) economic factors (i.e., low socioeconomic status (SES)^26,31,34,37 or no insurance or Medicaid insurance,^{25,29,30,34,35,37} 2) urban residence,^30,31 3) the presence of PAC supply,^25,26 and 4) hospital characteristics.^26,30 Only one study demonstrated the positive association between marital status and going to IRF versus SNF.³⁵

Trends in PAC utilization

Table 6 includes studies (n=4) that discussed general time trends of discharge destinations after hospitalization from stroke. Only one study included home as a discharge destination and showed an increasing trend for home discharges from 1996 to 2003.²⁵ For HH, two studies covered different time frames, and one study showed a decreasing trend of HH from 1996 to 2003,²⁵ whereas the other study showed an increasing trend of HH from 2003 to 2011.²⁹ Similarly, three studies covered different time periods of 1996 to 2003,²⁵ 2003 to 2011,²⁹ and 2012 to 2016.²⁸ There was a decreasing trend for earlier years (before 2011)^25,29 but an increasing trend after 2011.²⁸ There was an overall increasing trend of IRF discharges over time from studies covering years 1996 to 2020.^25,28,29,36

Table 6.

Time trends in Post-acute Care (n=4)

Author, year	Data periods	Home	HH	SNF/NH	IRF
Stroke (n=9)
Sandel, 2009	1996–2003	↑	↓	↓	↑
Bettger, 2015	2003–2011		↑	↓	↑
Thau, 2021	2019–2020				↑
Kim, 2022	2012–2016			↑	↑

Notes: (−) denotes “no change”

Discussion

We found racial and ethnic disparities in discharge destinations after hospitalization from stroke. Additionally, lower SES status, no insurance or Medicaid insurance, urban residence, presence of PAC supply and hospital characteristics (i.e., IRF affiliation) were associated with increased institutional PAC utilization among racial and ethnic minority individuals. Hispanic individuals mostly resided in rural areas and had limited access to institutional PAC services.

Our study found that Black individuals had higher SNF/NH discharges than White individuals among patients hospitalized for stroke. In a systematic review that explored predictive factors for discharge destinations after stroke, Van der Cruyssen and colleagues found that Black individuals were more likely to be discharged to institutional PAC settings (i.e., SNF/NH or IRFs) compared to White individuals.¹⁹ Another similar systematic review by Mees and colleagues reported that Hispanic individuals were more likely to be discharged to HH.¹⁸ Our results were similar to both the reviews’ findings in that racial and ethnic minority groups had higher likelihood of being discharged to SNF/NH or HH; however, we did not necessarily find that to be the case for IRF discharges. In fact, we found that results were mixed for Black individuals’ utilization of IRF,^{26,28–30,32,34} and Hispanic individuals were less likely to be discharged to IRFs^26,37 than White individuals. Since our review included articles that were published after the aforementioned reviews,^18,19 our results may reflect more recent trends that need to be confirmed through more studies.

By applying the SDOH framework, we found that there was limited integration of SDOH variables among included studies. For instance, while the SDOH framework considered various variables of neighborhood crime and poverty, walkability, access to public transportation, pollutants, availability of green space, housing insecurity for the built environment construct, included studies only considered geographic variability in a crude form of urban/rural differences. Furthermore, important SDOH such as community/social context, education, and food security were rarely considered in studies. Similarly, included studies did not properly represent the variables related to the healthcare system construct such as implicit provider biases or hospital disproportionate share status. Instead, studies only included descriptive hospital characteristics, which were not necessarily those that would affect the care of racial and ethnic minority individuals.

Our results found lower SES status, no insurance or Medicaid insurance, urban residence, PAC availability and hospital characteristics (i.e., IRF affiliation) were associated with increased institutional PAC utilization among racial and ethnic minority individuals. This is consistent with previous literature which has demonstrated the influence of SES,¹⁸ insurance,¹⁹ education,¹⁸ geographic factors,^18,19 and PAC availability¹⁸ on PAC utilization after stroke although these factors were not reviewed in the context of patients’ racial and ethnic minority status. Additionally, it is also worth considering the intersection of age and race and ethnic status on PAC utilization after stroke. While older age is associated with increased rates of stroke,¹⁶ the racial and ethnic minority individuals in the included studies were younger than White individuals,^{25,26,31–34,38} demonstrating a possible influence of age on decisions for discharge destinations, likely because the disparity in stroke incidence is greatest in younger compared to older individuals.^40,41

Our findings regarding PAC utilization trends over time for individuals hospitalized after stroke may have been influenced by policy changes. The increasing trend in home discharges over time is unsurprising given policy shifts such as bundled payments⁴² and prospective payment systems for PAC in 2002, both of which may have influenced decisions to lower costs by limiting the use of institutional PAC.⁴³ Conversely, the IRF 75% rule in 2004, which required IRFs to have a certain percentage of their patients be in need of intensive multidisciplinary inpatient rehabilitation as well as have one or more of 13 medical conditions,⁴⁴ may have led to increases in IRF discharges.

Our findings have several policy implications. Ideally, PAC services after hospital discharge functions to transition individuals from the inpatient setting to medically appropriate settings, thus minimizing care disruptions that could lead to adverse events and hospital readmissions.⁴⁵ What is considered equitable is when an individual, regardless of their sociocultural and medical background, can reach discharge destinations that are suitable for his/her/their required level of care. In a systematic review that compared patients with stroke who were discharged to IRF versus SNF, Pattath and colleagues (2023) found that individuals identifying as non-Hispanic Black or Hispanic were more likely to be discharged to SNF than IRF compared to non-Hispanic White individuals, while controlling for functional and medical status (i.e., comorbidities, acuity) as well as other individual-level (i.e., age, insurance) and hospital-level (i.e., bed size) characteristics.⁴⁶ In other words, racial and ethnic minority individuals with the same clinical profile were less likely than White individuals to access IRF, which provides a higher level of care and is associated with improved functional outcomes compared to SNF.⁴⁶ Considering the higher rates and severity of stroke⁴ and the SDOH constraints experienced by racial and ethnic minority individuals compared to White individuals, our findings of limited access to SNF/NH and IRF among Hispanic individuals and mixed results regarding IRF access for Black individuals compared to White individuals is concerning and require further attention from a policy perspective. Historically, not only have racial and ethnic minority individuals been subjected to health policies that have limited their economic opportunities and led to worse health outcomes, their needs and voices have not been heard.⁴⁷ Future policy regarding PAC for racial and ethnic minority individuals hospitalized for stroke requires the consideration of SDOH that may influence PAC decisions and their outcomes following stroke. Researchers should identify the most prominent SDOH elements that affect this population by eliciting input and feedback from those who have been affected by stroke.

Our findings also have several research implications. The finding that PAC decisions are, in part, mediated by SDOH suggests the need for additional research in other aspects of SDOH. This knowledge could be used by healthcare leaders and policymakers to help vulnerable individuals better access high-quality PAC care. This will also enable a more tailored SDOH framework in the context of PAC utilization among racial and ethnic minority individuals hospitalized after acute medical events. Also, future research efforts should focus on improving screening of SDOH among racial and ethnic minority individuals and intervening with appropriate care coordination to allow equitable access to appropriate PAC services. Finally, additional studies are necessary to further explore the impact of these changes in Medicare policy and PAC utilization on long-term outcomes among hospitalized patients who had stroke.

Limitations

This review has several limitations. First, there were inconsistent groupings and designations of different PAC services and race and ethnicity, which prevented direct comparisons among studies. Second, our findings have limited generalizability beyond the U.S., as countries have different health systems and policies. Third, since our main objective was not to understand the general trends in discharge destinations, our depiction of time trends among studies covering disparities among different racial and ethnic groups may have limited utility. Fourthly, included studies were inconsistent with their inclusion of stroke types (i.e., ischemic, hemorrhagic, both), which may indicate differing degrees of stroke severity and residual deficits, which not only made comparisons of discharge destinations difficult, but it also made it hard to determine whether an appropriate level of PAC service was assigned. Finally, we acknowledge that we have not included end-of-life care options such as hospice or palliative care. With high mortality rates observed among those who experience stroke¹ and increasing utilization of end-of-life care in this population,^48,49 studies have found disparities in the provision of end-of-life care among minority racial and ethnic groups.^48,50 Future reviews may examine disparities in the provision of end-of-life care services among various racial and ethnic groups. Despite these limitations, this review was methodologically rigorous, adhering to the most recent guidelines in its conduct and reporting, consulting with a knowledgeable informationist for search design, and appraised the quality of included studies with a validated and well-accepted quality appraisal tool. We also made a unique contribution by synthesizing findings through the lens of a SDOH framework, and highlighting important findings and areas for future investigation to improve health outcomes.

Conclusion

While we found that Black and Hispanic individuals were less likely to be discharged home following hospitalization from stroke compared to White individuals, they were more likely than White individuals to be discharged home with HH. Black individuals were more likely to be discharged to SNF than White individuals; however, there were mixed results for discharges to IRF. Hispanic individuals were less likely to be discharged to SNF and IRF compared to White individuals. Lower SES status, no insurance or Medicaid insurance, urban residence, area PAC supply and hospital characteristics (i.e., IRF affiliation) were associated with increased institutional PAC utilization among racial and ethnic minority individuals. Future policy and research should incorporate the perspectives of community stakeholders in identifying the most salient aspects of SDOH that affect PAC decisions among racial and ethnic minority individuals. Furthermore, more research is needed to explore the impact of these changes in Medicare policy and PAC utilization on long-term outcomes among hospitalized patients who had stroke.

Highlights.

• Disparities exist in post-acute care (PAC) use among racial and ethnic groups

• Socioeconomic status and Medicaid insurance were associated with PAC use

• Urban residence and hospital characteristics were associated with PAC use