Abstract
Background and Objectives
Post-acute care is critical for patient functional recovery and successful community transition. This study aimed to understand the current racial, ethnic, and regional disparities in post-acute service utilization after stroke.
Methods
This retrospective cross-sectional study included patients hospitalized for ischemic stroke and intracerebral hemorrhage in 2017–2018 using the National Inpatient Sample. Discharge destinations were classified as follows: (1) facility including inpatient rehabilitation, skilled nursing facility, and facility hospice; (2) home health care (HHC), including home health and home hospice; and (3) home without HHC. Multinomial logistic regression was used to study the odds of discharge to a facility over home and HHC over home without HHC by race, ethnicity, insurance, and census division, adjusting for clinical factors and survey design.
Results
Among the 1,000,980 weighted ischemic stroke admissions, 66.9% were White, 17.6% Black, 9.5% Hispanic, 3.1% Asian American/Pacific Islander, and 0.4% Native American. Relative to private insurance, uninsured patients had the lowest adjusted odds of facility over home discharge (0.44; 95% CI 0.40–0.48) and HHC discharge over home without HHC (0.79; 95% CI 0.71–0.88). Compared with White patients, only Hispanic patients with Medicare/Medicaid insurance or self-pay had lower odds of facility over home discharge (adjusted OR 0.80 and 0.75, respectively; 95% CI 0.76–0.84 and 0.63–0.93). Uninsured Hispanic patients also had lower odds of HHC discharge over home without HHC than White patients (0.74; 95% CI 0.57–0.97). Facility discharge rate was the highest in East North Central (39.2%) and lowest in Pacific (31.2%). HHC discharge rate was the highest in New England (20.2%) and lowest in West North Central (10.3%), which had the highest home without HHC discharge (46.1%). Compared with New England, other census divisions had lower odds of facility over any home discharge with Pacific being the lowest (adjusted OR, 0.66; 95% CI 0.60–0.71) and HHC over home without HHC discharge with West North Central being the lowest (adjusted OR, 0.33; 95% CI 0.29–0.38). Similar patterns were observed in intracerebral hemorrhage.
Discussion
Significant insurance-dependent racial and ethnic disparities and regional variations were evident in post-acute service utilization after stroke. Targeted efforts are needed to improve post-acute service access for uninsured patients especially Hispanic patients and people in certain regions.
Introduction
Stroke is the leading cause of long-term disability.1 Stroke survivors should receive appropriate post-acute care to enhance recovery and minimize disability.2 However, the decisions of discharge disposition are heavily influenced by nonclinical factors such as race, sex, insurance, and geographic location.3-10 Women have more home health care (HHC) enrollment but less outpatient rehabilitation treatment.3 Previous studies about racial and ethnic disparities in post-acute service utilization reached mixed results, especially about Black patients in the United States.7-10 A study of 72 patients in Texas showed that compared with White patients, Mexican Americans were more likely to receive home-based rehabilitation and less likely to receive inpatient rehabilitation, although a proportion of patients sent home without any rehabilitation were comparable.7 A study of 1999 and 2000 Medicare data showed that Black and Hispanic patients were more likely to be discharged to HHC and less likely to be discharged to skilled nursing facilities (SNFs) than White patients.8 By contrast, a study using Tennessee state data showed that Black patients were more likely to be discharged to a facility than White patients.9 It is worth noting that the interaction of insurance status with race and ethnicity was not examined in these studies. In addition, regional variations of post-acute service availability and utilization have been previously reported,6 but the recent patterns have not been studied. It is possible that racial, ethnic, insurance status, and regional disparities underlie the rising concerns about the unexplained variations in post-acute service utilization across the country in which patients with the same clinical characteristics may be discharged to very different post-acute settings.4,11
In addition, health policies and regulations have strong influence on the accessibility of post-acute care.4,12-15 The Bundled Payment for Care Improvement Initiative (BPCI) established by the Affordable Care Act have resulted in reductions in inpatient rehabilitation accessibility.12-14,16 For providers and facilities that participated in BPCI, total episode payments were reconciled against a target price. When episode payments were less than the target, participants were eligible for additional amounts; when payments were more than the target, participants may have to repay CMS.14 The BPCI has resulted in significant decline in Medicare payments primarily because of reductions in institutional post-acute care in both access and duration.17 Unfortunately, the reduction in post-acute service has been associated with significant decline in functional outcomes and increase in readmission after stroke, different from other conditions such as orthopaedic surgeries.17 It remains unclear whether these new health care policies have differentially affected post-acute care among stroke patients of different races, ethnicities, and geographic locations. Further actions to improve health care equality and outcomes are only possible with more in-depth understanding of the current practice patterns. This study aimed to rigorously examine the current racial and ethnic disparities in post-acute service utilization and their interactions with insurance, specifically in facility over home discharge and HHC over home without HHC discharge. We also examined the regional variations in post-acute service utilization.
Methods
Data Source and Study Cohort
This retrospective cross-sectional study used 2017 and 2018 National Inpatient Sample (NIS) from the Healthcare Cost and Utilization Project (HCUP).18-22 NIS is the largest all-payer inpatient dataset in the United States, approximating a 20% stratified sample of patient hospitalizations in all the community hospitals in the United States.23 To obtain nationwide estimates, discharge weights were developed to extrapolate NIS discharges to the discharge universe.24 The discharge weights were calculated by first stratifying the NIS hospitals by census division, urban/rural location, teaching status, bed size, and ownership. A weight was then calculated for each stratum by dividing the number of universe discharges (actual count of discharges) in that stratum, obtained from HCUP and the American Hospital Association data, by the number of NIS discharges in the stratum. Weights have been assigned to each discharge and are stored in each record. When the discharge weights are applied to the unweighted NIS data, the result is an estimate of the number of discharges for the entire universe.25
The approach with stratified random sampling ensures that the database content is representative of the US population and can be weighed to produce national estimates. The data from Alabama, Idaho, and New Hampshire were not available in the 2017 and 2018 NIS data. However, the NIS sampling and weighting strategies were adjusted accordingly. This study included patients who were admitted directly from the emergency department for ischemic stroke and ICH with complete inpatient data in the NIS. The inclusion and exclusion algorithms are presented in Figure 1. The primary discharge diagnoses were abstracted using the International Classification of Diseases 10th Revision (ICD-10) as provided in eTable 1. The modified Elixhauser comorbidity index was calculated using HCUP comorbidity software.26,27 Original race and ethnicity data were provided by the data source State Inpatient Data.28 The collection methods for race and ethnicity varied by states using patient self-identification, electronic health record, or hospital staff reporting.28 HCUP launched several initiatives to mitigate misclassification or missing data for race and ethnicity, including crosslinking of administrative databases and hospital staff training.28,29 NIS combines “race” and “ethnicity” provided by the data source into one data element. If both race and ethnicity are available, ethnicity takes precedence over race in setting the HCUP value.
Figure 1. Inclusion and Exclusion Algorithms.
Left against medical advice referred to the patients who declined the recommendations to stay in the hospital. Home with HHC included organized home health service and home hospice. Facility discharge included discharge to inpatient rehabilitation, skilled nursing facility, hospice facility, long-term care facility, and psychiatric hospital. The presented values may not add up to the weighted total because of cell suppression policy of the Agency for Healthcare Research and Quality (hcup-us.ahrq.gov/team/NationwideDUA.pdf). The discharge weights were calculated by first stratifying the NIS hospitals by census division, urban/rural location, teaching status, bed size, and ownership. A weight was then calculated for each stratum, by dividing the number of universe discharges (actual count of discharges) in that stratum, obtained from HCUP and the American Hospital Association data, by the number of NIS discharges in the stratum. Weights have been assigned to each discharge and are stored in each record. When the discharge weights are applied to the unweighted NIS data, the result is an estimate of the number of discharges for the entire universe. ED admissions = admissions directly from the emergency department; HHC = home health care.
Outcomes
The outcomes were 3 main discharge destinations: (1) facility, including inpatient rehabilitation facility (IRF), SNF, hospice facility, psychiatric hospital, and long-term care hospital; (2) HHC, including organized home health service and home hospice; and (3) home or self-care without HHC. Detail HCUP definitions of each discharge destination are provided in eTable 2.
Statistical Analysis
Observed patient and hospital characteristics for each discharge destination were presented as weighted frequency (percentage and 95% CI) for categorical variables and median and interquartile range for continuous variables. The proportion of patients being discharged to each post-acute care pathway by census division were depicted on a heat map. The differences in discharge destination for each patient and hospital factor were compared using Rao-Scott chi-square tests for categorical variables and ANOVA with log transformation for continuous variables, adjusted for survey design. A bivariate survey-weighted logistic regression model was used to test the unadjusted relationship between patient and hospital characteristics and each discharge destination.
A multinomial logistic regression model with a generalized logit link was used to examine the associations between each factor and discharge to one pathway vs others (facility over any home discharge, facility over HHC, and HHC over home without HHC). The multinomial generalized logistic regression model is a model designed to predict the probabilities of multiclass categorically distributed dependent variables (more than 2 outcomes), given a set of independent variables.30 The interactions between race, ethnicity, and insurance were assessed by comparing each non-White racial and ethnic group with White patients within each insurance stratum. The models were adjusted for potential confounders including patient demographics and socioeconomic status (age, sex, race, ethnicity, primary insurance, zip code–level median household income by quartile, and rural-urban residency), comorbidities, time of stroke admission (weekend admissions) and proxies of stroke severity (length of stay, hemiplegia or paralysis, aphasia, intubation or mechanical ventilation, pneumonia, urinary tract infection, deep vein thrombosis, IV thrombolytic therapy, and endovascular thrombectomy), and hospital characteristics (hospital bed size, annual stroke volume, and census division). We combined 2 recent years of NIS data and adjusted the stratum identifiers based on the year to ensure proper handling of variance elements. We did not adjust for year because secular differences were not part of our hypotheses.
For comorbidities, we adopted most of Elixhauser comorbidity indices, with additional stroke-specific variables that have been used in previous studies (hypertension, myocardial infarction, congestive heart failure, atrial fibrillation, COPD, renal failure, liver disease, depression, peripheral vascular disease, diabetes mellitus, rheumatoid disease, coagulopathy, obesity, weight loss, fluid and electrolyte disorders, anemia, alcohol abuse, drug abuse, psychosis, malignancy, hypothyroidism, neurologic disorders, and smoking).26,31-33 The NIH Stroke Scale (NIHSS) was not adjusted in the modeling analyses because they were reported in only 31.9% of patients with ischemic stroke and 20.5% of patients with ICH in our cohort (eTable 3), and a previous study showed that NIHSS missingness were not random.34 Previous studies have shown that the selected proxies of stroke severity from the administrative database were either directly correlated with NIHSS or associated with stroke outcomes.33,35,36 Odds ratios and 95% CIs were reported with significance values for the exponentiated parameter estimates based on reference cell parameterization.
Handling of missing data: The data missing rates in the NIS database were low (eTable 4), and imputations were not necessary.37,38 Patients with discharge status unknown or missing were excluded from the analysis (335 patients in ischemic stroke study population and 55 patients in ICH population). Patients with missing race, ethnicity, and insurance data were excluded from the modeling analyses. The remaining records all had positive discharge weights and were included in the analyses using the not missing completely at random (NOMCAR) option as recommended by HCUP.39 The NOMCAR function computes variance estimates by analyzing the nonmissing values as a domain (subpopulation), where the entire population includes both nonmissing and missing domains.39
The analyses were performed using SAS survey procedures (version 9.4, SAS Institute, NC). All hypothesis tests were 2-sided, with p < 0.05 considered statistically significant.
Standard Protocol Approvals, Registrations, and Patient Consents
The study was approved by the Institutional Review Board of the Cleveland Clinic Foundation under the policy of exempted human subject research. Patient consents were deemed not needed.
Data Availability
The data used in this study will not be shared by the authors because of our Data Use Agreement with the HCUP, which prohibits data redistribution. The HCUP Central Distributor is the entity that accepts, processes, and fulfills applications for the databases and manages data use agreements for all data users. Investigators may request access to anonymized patient data at distributor.hcup-us.ahrq.gov/.
Results
Discharge Destination by Patient and Hospital Characteristics
The inclusion and exclusion algorithms are presented in Figure 1. A total of 200,196 unweighted ischemic stroke and 18,287 ICH admissions were observed, representing 1,000,980 weighted ischemic stroke and 91,435 weighted ICH admissions. Discharge destinations after ischemic stroke based on patient age, sex, and socioeconomic status available in the NIS database (race, ethnicity, primary insurance, and zip code–level median household income) are shown in Table 1, and by clinical factors and hospital characteristics in eTable 5. Among ischemic stroke admissions, 49.1% were male, 66.9% White, 17.6% Black, 9.5% Hispanic, 3.1% Asian American/Pacific Islander, and 0.4% Native American. Overall, 2.6% died in the hospital and 42.7% were discharged to home without HHC, 15.2% to HHC, 35.7% to a facility, and 1.2% left against medical advice (declined the recommendations to stay in the hospital). Among all the racial/ethnic groups, Hispanic patients had the highest rate of discharge to home without HHC (49.1%) and lowest facility discharge rate (29.1%). Patients with Medicare insurance had the highest rate of facility discharge (42.0%) and lowest rate of home without HHC discharge (34.1%). Similar patterns existed after ICH (eTables 6 and 7).
Table 1.
Discharge Destination of Ischemic Stroke by Patient Characteristics
| Total admission, n (%) | In-hospital mortality, % (95% CI) | Home without HHCa, % (95% CI) | Home health care, % (95% CI) | Facilityb, % (95% CI) | Left AMA, % (95% CI) | p Value | |
| Overall | 1,000,980 | 26,140 2.6 (2.5, 2.7) |
427,385 (42.7%) | 152,575 (15.2%) | 357,125 (35.7%) | 11,935 (1.2%) | |
| Age, y | <0.001 | ||||||
| 18–49 | 77,630 (7.8) | 1.0 (0.9–1.2) | 67.3 (66.5–68.1) | 8.0 (7.6–8.5) | 16.5 (15.9–17.1) | 3.5 (3.2–3.8) | |
| 50–64 | 254,490 (25.4) | 1.6 (1.5–1.7) | 56.7 (56.2–57.3) | 11.4 (11.0–11.7) | 25.4 (25.0–25.9) | 2.0 (1.9–2.2) | |
| 65–79 | 368,425 (36.8) | 2.3 (2.1–2.4) | 43.2 (42.8–43.7) | 15.3 (15.0–15.7) | 35.6 (35.2–36.0) | 0.8 (0.8–0.9) | |
| ≥80 | 300,100 (30.0) | 4.3 (4.1–4.5) | 23.8 (23.4–24.2) | 20.3 (19.9–20.6) | 49.4 (48.9–49.9) | 0.3 (0.3–0.4) | |
| Sex | <0.001 | ||||||
| Male | 491,635 (49.1) | 2.5 (2.4–2.6) | 46.8 (46.4–47.1) | 13.0 (12.7–13.3) | 33.4 (33.1–33.8) | 1.5 (1.4–1.6) | |
| Female | 509,010 (50.9) | 2.8 (2.6–2.9) | 38.8 (38.4–39.2) | 17.4 (17.1–17.7) | 37.9 (37.5–38.2) | 0.9 (0.8–0.9) | |
| Race and ethnicity | <0.001 | ||||||
| White | 654,550 (66.9) | 2.8 (2.7–2.9) | 42.1 (41.8–42.5) | 14.9 (14.6–15.2) | 36.5 (36.2–36.9) | 1.0 (1.0–1.1) | |
| Black | 171,930 (17.6) | 2.0 (1.9–2.2) | 41.3 (40.7–42.0) | 16.2 (15.7–16.7) | 36.5 (35.9–37.1) | 1.5 (1.4–1.7) | |
| Hispanic | 92,955 (9.5) | 2.0 (1.8–2.2) | 49.1 (48.1–50.1) | 15.6 (14.9–16.3) | 29.1 (28.3–30.0) | 1.6 (1.4–1.8) | |
| AAPI | 30,330 (3.1) | 3.2 (2.7–3.7) | 41.7 (40.3–43.1) | 17.4 (16.2–18.5) | 34.3 (33.1–35.6) | 0.8 (0.6–1.1) | |
| Native American | 3,540 (0.4) | 2.3 (1.3–3.3) | 47.5 (43.6–51.5) | 12.3 (9.8–14.7) | 32.7 (29.2–36.3) | — | |
| Other | 25,205 (2.6) | 2.9 (2.4–3.4) | 44.8 (43.3–46.3) | 16.2 (15.0–17.3) | 31.5 (30.1–33.0) | 1.9 (1.5–2.3) | |
| Race, ethnicity, and sex | <0.001 | ||||||
| White male | 323,105 (33.0) | 2.6 (2.4–2.7) | 47.1 (46.6–47.5) | 12.7 (12.4–13.0) | 33.5 (33.1–33.9) | 1.4 (1.3–1.5) | |
| White female | 331,445 (33.9) | 3.0 (2.8–3.1) | 37.4 (36.9–37.8) | 17.0 (16.7–17.4) | 39.6 (39.1–40.0) | 0.73 (0.66–0.80) | |
| Black male | 80,090 (8.2) | 2.1 (1.8–2.3) | 43.7 (42.8–44.5) | 13.9 (13.3–14.5) | 35.8 (35.0–36.6) | 2.0 (1.7–2.2) | |
| Black female | 91,840 (9.4) | 2.0 (1.8–2.2) | 39.3 (38.5–40.1) | 18.3 (17.6–18.9) | 37.1 (36.3–37.9) | 1.1 (0.95–1.3) | |
| Hispanic male | 47,385 (4.8) | 2.0 (1.7–2.3) | 50.4 (49.3–51.6) | 13.6 (12.8–14.4) | 29.2 (28.2–30.3) | 1.9 (1.6–2.2) | |
| Hispanic female | 45,570 (4.7) | 2.1 (1.8–2.4) | 47.8 (46.4–49.1) | 17.7 (16.8–18.6) | 29.0 (27.9–30.1) | 1.3 (1.06–1.6) | |
| AAPI/NatAm/other male | 29,915 (3.1) | 2.6 (2.2–3.0) | 46.5 (45.2–47.9) | 14.1 (13.1–15.1) | 32.3 (31.0–33.6) | 1.5 (1.2–1.8) | |
| AAPI/NatAm/other female | 29,160 (3.0) | 3.4 (2.9–3.9) | 40.2 (38.8–41.5) | 19.1 (18.0–20.2) | 33.8 (32.5–35.0) | 1.3 (0.96–1.6) | |
| Primary insurance | <0.001 | ||||||
| Medicare | 661,570 (66.2) | 3.0 (2.9–3.2) | 34.1 (33.8–34.5) | 17.6 (17.3–17.9) | 42.0 (41.7–42.4) | 0.8 (0.8–0.9) | |
| Medicaid | 90,565 (9.1) | 1.9 (1.7–2.1) | 51.8 (51.0–52.6) | 13.0 (12.4–13.5) | 26.9 (26.1–27.6) | 3.3 (3.1–3.6) | |
| Private insurance | 184,235 (18.4) | 1.6 (1.5–1.7) | 60.7 (60.1–61.4) | 9.7 (9.3–10.1) | 23.9 (23.3–24.4) | 1.0 (0.9–1.1) | |
| Self-pay | 40,395 (4.0) | 2.1 (1.8–2.4) | 70.9 (69.8–72.1) | 8.9 (8.1–9.8) | 12.5 (11.7–13.3) | 3.1 (2.7–3.5) | |
| No charge | 3,285 (0.3) | 1.2 (0.4–2.1) | 72.5 (68.9–76.0) | 9.4 (7.1–11.7) | 13.1 (10.5–15.7) | — | |
| Otherc | 19,235 (1.9) | 1.9 (1.5–2.4) | 56.1 (54.4–57.8) | 12.2 (11.1–13.2) | 24.3 (22.8–25.7) | 1.5 (1.1–1.9) | |
| Race, ethnicity, and primary insurance | <0.001 | ||||||
| White/private | 116,970 (12.0) | 1.7 (1.5–1.8) | 62.5 (61.6–63.3) | 9.0 (8.6–9.5) | 22.8 (22.1–23.4) | 0.96 (0.83–1.09) | |
| Black/private | 34,070 (3.5) | 1.3 (1.04–1.6) | 54.6 (53.3–55.9) | 12.0 (11.1–12.9) | 28.0 (26.9–29.2) | 1.2 (0.97–1.5) | |
| Hispanic/private | 16,085 (1.6) | 1.06 (0.71–1.4) | 63.0 (61.3–64.7) | 9.5 (8.5–10.6) | 22.3 (20.8–23.8) | 1.2 (0.85–1.6) | |
| AAPI/private | 6,640 (0.68) | 2.1 (1.4–2.8) | 56.2 (53.5–59.0) | 11.8 (10.0–13.6) | 26.3 (23.9–28.7) | 0.60 (0.19–1.01) | |
| NatAm/other/private | 5,855 (0.60) | 2.3 (1.5–3.1) | 61.6 (58.8–64.3) | 8.8 (7.1–10.5) | 23.4 (21.0–25.8) | 1.3 (0.64–1.9) | |
| White/Mcare/aid | 505,775 (51.8) | 3.1 (2.9–3.2) | 36.1 (35.7–36.5) | 16.6 (16.3–16.9) | 40.9 (40.5–41.3) | 0.95 (0.88–1.01) | |
| Black/Mcare/aid | 121,335 (12.4) | 2.2 (2.1–2.4) | 34.3 (33.6–35.1) | 17.9 (17.3–18.5) | 41.7 (41.0–42.4) | 1.5 (1.3–1.6) | |
| Hispanic/Mcare/aid | 66,755 (6.8) | 2.4 (2.1–2.6) | 41.9 (40.8–43.0) | 18.2 (17.4–19.1) | 33.3 (32.3–34.3) | 1.7 (1.4–2.0) | |
| AAPI/Mcare/aid | 21,980 (2.2) | 3.7 (3.1–4.3) | 35.7 (34.2–37.3) | 19.4 (18.1–20.7) | 38.1 (36.7–39.6) | 0.73 (0.47–0.99) | |
| NatAm/other/Mcare/aid | 20,080 (2.0) | 3.0 (2.5–3.6) | 37.3 (35.6–39.0) | 18.8 (17.4–20.2) | 36.3 (34.5–38.0) | 2.0 (1.6–2.5) | |
| White/self-pay/other | 30,975 (3.2) | 2.3 (1.9–2.7) | 64.6 (63.3–65.9) | 9.3 (8.6–10.1) | 17.5 (16.5–18.5) | 3.0 (2.5–3.4) | |
| Black/self-pay/other | 16,250 (1.7) | 1.8 (1.3–2.2) | 65.4 (63.7–67.2) | 12.4 (10.9–13.8) | 15.7 (14.4–17.1) | 2.3 (1.8–2.8) | |
| Hispanic/self-pay/other | 9,970 (1.0) | 1.3 (0.77–1.7) | 74.8 (72.7–77.0) | 8.0 (6.6–9.4) | 12.1 (10.4–13.7) | 1.6 (0.96–2.1) | |
| AAPI/self-pay/other | 1,675 (0.17) | 2.1 (0.56–3.6) | 62.7 (57.6–67.8) | 12.5 (9.0–16.1) | 17.0 (13.1–20.9) | 3.0 (1.2–4.8) | |
| NatAm/other/self-pay/other | 2,755 (0.28) | 2.4 (1.10–3.6) | 67.0 (62.8–71.2) | 8.0 (5.6–10.3) | 16.0 (12.4–19.5) | 2.7 (1.4–4.0) | |
| Median household income for patient zip codes | <0.001 | ||||||
| First quartile | 293,925 (29.8) | 2.6 (2.5–2.8) | 42.2 (41.6–42.7) | 35.7 (35.2–36.2) | 15.6 (15.2–16.0) | 1.5 (1.4–1.6) | |
| Second quartile | 256,660 (26.1) | 2.5 (2.4–2.7) | 42.8 (42.3–43.3) | 35.8 (35.3–36.3) | 14.9 (14.5–15.3) | 1.2 (1.1–1.3) | |
| Third quartile | 237,315 (24.1) | 2.6 (2.4–2.7) | 43.1 (42.5–43.6) | 35.8 (35.2–36.3) | 15.0(14.6–15.4) | 1.0 (0.9–1.1) | |
| Fourth quartile | 197,010 (20.0) | 2.7 (2.5–2.8) | 42.5 (41.9–43.1) | 35.7 (35.1–36.3) | 15.6 (15.1–16.1) | 0.9 (0.8–1.0) |
Abbreviations: AAPI = Asian American/Pacific Islander; HHC = home health care; Left AMA = left against medical advice, referring to patients who declined recommendations to stay in the hospital; Mcare/aid = Medicare/Medicaid; NatAm/other = Native American or other races.
Data represented estimated weighted admissions.
Home health care included organized home health service and home hospice.
Facility discharge included discharge to inpatient rehabilitation, skilled nursing facility, hospice facility, long-term care facility, and psychiatric hospital. Transfer to short-term hospital or discharged alive with destination unknown was not included.
Other insurances included no charge, insurance unknown, the Civilian Health and Medical Program of the Department of Veterans Affairs, workers' compensation, and other government programs.
Regional Disparities in Poststroke Discharge Disposition
The numeric rates of the 3 main discharge destinations in each region are presented as heat maps in Figure 2 and as bar graphs in Figure 3. For ischemic stroke, the proportion of patients who were discharged to home without HHC was the lowest in New England census division (32.3%) and highest in West North Central (46.1%), with other regions all above 40%. HHC discharge showed a reversed pattern with the highest in New England (20.2%) and lowest in West North Central (10.3%). The rates of facility discharge ranged from the highest in East North Central division (39.2%), followed by Mountain area (38.6%), West North Central (38.1%), New England (37.6%), Middle Atlantic (36.7%), East South Central (36.5%), West South Central (34.8%), and South Atlantic (34.4%), to the lowest in Pacific (31.2%). Similar variations were observed after ICH.
Figure 2. Heat Maps of Post-Acute Service Utilization by Census Division.
Home with HHC included organized home health service and home hospice. Facility discharge included discharge to inpatient rehabilitation, skilled nursing facility, hospice facility, long-term care facility, and psychiatric hospital. Beginning in 2012, hospital state is no longer available in the National Inpatient Sample (NIS), and the NIS is stratified by census divisions rather than census regions. HHC = home health care; ICH = intracerebral hemorrhage.
Figure 3. Discharge Destination by Census Division.
Home with HHC included organized home health service and home hospice. Facility discharge included discharge to inpatient rehabilitation, skilled nursing facility, hospice facility, long-term care facility, and psychiatric hospital. States in each census division: New England (ME, VT, MA, RI, CT); Mid-Atlantic (NY, PA, NJ); East North Central (WI, MI, IL, IN, OH); West North Central (MO, ND, SD, NE, KA, MN, IA); South Atlantic (DE, MD, DC, VA, WV, NC, SC, GA, FL); East South Central (KY, TN, MI); West South Central (OK, TX, AR, LA); Mountain (MT, WY, NV, UT, CO, AZ, NM); Pacific (AK, WA, OR, CA, HI). Data from the State of AL, ID, and NH were not included in the 2017 and 2018 National Inpatient Sample (NIS). However, the NIS sampling and weighting strategy were adjusted accordingly. HHC = home health care; Left AMA = left against medical advice, referring to patients who declined the recommendations to stay in the hospital.
Racial, Ethnic, and Regional Disparities in Discharge to One Pathway Over Others After Ischemic Stroke
Further analyses were performed to examine the associations of sex, race, ethnicity, insurance, and geographic locations with the odds of discharge to one pathway over others using multinomial generalized logistic regression models, which can simultaneously study multiple outcomes (Table 2). Compared with White men, White women and Black women and men were more likely to be discharged to a facility over home (adjusted odds ratio [aOR] 1.07, 1.16, and 1.34, respectively), as well as HHC over home without HHC (aOR, 1.44, 1.77, and 1.44, respectively). Hispanic women had lower odds of facility over home discharge (aOR, 0.78; 95% CI 0.73–0.83) but higher odds of HHC over home without HHC discharge (aOR, 1.35; 95% CI 1.25–1.47). Overall, compared with those with private insurance, uninsured patients had lower odds to be discharged to a facility over home (aOR, 0.44, 95% CI 0.40–0.48) and HHC over home without HHC (aOR, 0.79, 95% CI 0.71–0.88). Racial and ethnic disparities in post-acute service utilization differed by insurance. Relative to White patients, the odds of being discharged to a facility over home were lower among Hispanic patients with Medicare/Medicaid insurance (aOR, 0.80; 95% CI 0.76–0.84) and uninsured/self-pay (aOR, 0.75; 95% CI 0.63–0.93), but not private insurance (aOR, 1.08; 95% CI 0.98–1.20). Relative to White patients, the odds of HHC discharge over home without HHC were lower among uninsured Hispanic patients (aOR, 0.74; 95% CI 0.57–0.97), but not insured. In addition, the odds of facility over home discharge were higher among Black patients with Medicare/Medicaid or private insurance, but not uninsured. The odds of HHC discharge over home without HHC were higher among Black patients within any payment strata, as well as Asian American/Pacific Islanders with Medicare/Medicaid or private insurance.
Table 2.
Factors Associated With Discharge to One Pathway vs Others After Ischemic Stroke
| Factor | Facilitya over any home | HHCb over home without HHC | ||
| Adjusted OR (95% CI) | p Value | Adjusted OR (95% CI) | p Value | |
| Age, y (reference: ≥80) | ||||
| 18–49 | 0.23 (0.22–0.24) | <0.001 | 0.19 (0.18–0.21) | <0.001 |
| 50–64 | 0.38 (0.36–0.39) | <0.001 | 0.30 (0.29–0.32) | <0.001 |
| 65–79 | 0.55 (0.53–0.56) | <0.001 | 0.44 (0.42–0.46) | <0.001 |
| Race, ethnicity, and sex, (reference: White male) | ||||
| White female | 1.07 (1.04–1.10) | <0.001 | 1.44 (1.38–1.49) | <0.001 |
| Black female | 1.16 (1.11–1.21) | <0.001 | 1.77 (1.66–1.88) | <0.001 |
| Black male | 1.34 (1.28–1.40) | <0.001 | 1.44 (1.35–1.54) | <0.001 |
| Hispanic female | 0.78 (0.73–0.83) | <0.001 | 1.35 (1.25–1.47) | <0.001 |
| Hispanic male | 0.97 (0.91–1.03) | 0.36 | 1.15 (1.06–1.24) | <0.001 |
| AAPI/NatAm/other female | 0.93 (0.87–1.00) | 0.04 | 1.67 (1.52–1.83) | <0.001 |
| AAPI/NatAm/other male | 1.06 (0.99–1.14) | 0.11 | 1.28 (1.16–1.41) | <0.001 |
| Primary insurance (reference: private insurance) | ||||
| Medicare/Medicaid | 1.35 (1.31–1.40) | <0.001 | 1.62 (1.54–1.70) | <0.001 |
| Self-pay/NC/other | 0.44 (0.40–0.48) | <0.001 | 0.79 (0.71–0.88) | <0.001 |
| Race, ethnicity, and insurance interactionc (reference: White) | ||||
| Medicare/Medicaid | ||||
| AAPI | 0.94 (0.87–1.01) | 0.08 | 1.23 (1.11–1.37) | <0.001 |
| Black | 1.18 (1.14–1.23) | <0.001 | 1.29 (1.22–1.37) | <0.001 |
| Hispanic | 0.80 (0.76–0.84) | <0.001 | 1.05 (0.98–1.12) | 0.21 |
| Private insurance | ||||
| AAPI | 1.35 (1.18–1.55) | <0.001 | 1.50 (1.24–1.82) | <0.001 |
| Black | 1.33 (1.24–1.42) | <0.001 | 1.42 (1.29–1.56) | <0.001 |
| Hispanic | 1.08 (0.98–1.20) | 0.12 | 1.10 (0.96–1.27) | 0.17 |
| Self-pay | ||||
| AAPI | 0.90 (0.58–1.40) | 0.64 | 1.32 (0.80–2.17) | 0.28 |
| Black | 0.91 (0.76–1.08) | 0.28 | 1.42 (1.17–1.73) | <0.001 |
| Hispanic | 0.75 (0.63–0.93) | 0.01 | 0.74 (0.57–0.97) | 0.03 |
| Median household income for patient zip codes (reference: 4th quartile) | ||||
| First quartile | 1.02 (0.98–1.06) | 0.32 | 1.17 (1.10–1.23) | <0.001 |
| Second quartile | 1.04 (1.00–1.08) | 0.06 | 1.09 (1.04–1.15) | <0.001 |
| Third quartile | 1.01 (0.98–1.05) | 0.45 | 1.03 (0.98–1.08) | 0.28 |
| Patient residence (reference: large metro) | ||||
| Small metro | 1.02 (0.99–1.05) | 0.21 | 0.94 (0.89–0.99) | 0.01 |
| Micropolitan | 1.02 (0.96–1.08) | 0.50 | 0.85 (0.78–0.93) | <0.001 |
| Rural | 1.05 (0.99–1.11) | 0.12 | 0.88 (0.80–0.97) | 0.01 |
| Hospital characteristics | ||||
| Hospital bedsized (reference: large) | ||||
| Small | 0.96 (0.91–1.01) | 0.10 | 1.06 (0.97–1.14) | 0.18 |
| Medium | 0.99 (0.95–1.03) | 0.64 | 1.01 (0.95–1.08) | 0.65 |
| Annual stroke volume (reference: >500) | ||||
| 5–50 | 1.09 (1.00–1.20) | 0.06 | 0.94 (0.82–1.07) | 0.35 |
| 51–100 | 1.01 (0.93–1.08) | 0.87 | 0.93 (0.83–1.05) | 0.23 |
| 101–300 | 0.98 (0.94–1.03) | 0.53 | 0.94 (0.87–1.02) | 0.12 |
| 301–500 | 1.00 (0.96–1.04) | 1.00 | 0.99 (0.92–1.06) | 0.67 |
| Hospital location/teaching status (reference: rural) | ||||
| Urban, nonteaching | 1.00 (0.93–1.08) | 0.94 | 0.98 (0.88–1.10) | 0.78 |
| Urban, teaching | 1.01 (0.93–1.09) | 0.85 | 0.94 (0.83–1.06) | 0.29 |
| Hospital ownership (reference: government, nonfederal) | ||||
| Private, nonprofit | 0.99 (0.94–1.04) | 0.69 | 1.13 (1.03–1.24) | 0.01 |
| Private, investor-owned | 0.91 (0.85–0.96) | 0.001 | 1.08 (0.97–1.20) | 0.16 |
| Hospital census divisione (reference: New England) | ||||
| Middle Atlantic | 0.91 (0.84–0.98) | 0.02 | 0.59 (0.53–0.66) | <0.001 |
| East North Central | 0.90 (0.83–0.98) | 0.01 | 0.41 (0.37–0.45) | <0.001 |
| West North Central | 0.90 (0.82–0.98) | 0.02 | 0.33 (0.29–0.38) | <0.001 |
| South Atlantic | 0.77 (0.72–0.84) | <0.001 | 0.61 (0.55–0.67) | <0.001 |
| East South Central | 0.83 (0.76–0.91) | <0.001 | 0.44 (0.38–0.50) | <0.001 |
| West South Central | 0.85 (0.78–0.93) | 0.001 | 0.48 (0.43–0.53) | <0.001 |
| Mountain | 0.94 (0.86–1.03) | 0.20 | 0.45 (0.39–0.51) | <0.001 |
| Pacific | 0.66 (0.60–0.71) | <0.001 | 0.54 (0.49–0.60) | <0.001 |
| Admission variables | ||||
| Weekend admission | 1.03 (1.00–1.05) | 0.03 | 0.99 (0.96–1.02) | 0.58 |
| Hemiplegia/paralysis | 3.22 (3.15–3.30) | <0.001 | 1.68 (1.63–1.74) | <0.001 |
| Aphasia | 1.82 (1.70–1.95) | <0.001 | 1.74 (1.58–1.91) | <0.001 |
| Endovascular thrombectomy | 1.69 (1.56–1.82) | <0.001 | 1.17 (1.03–1.32) | 0.02 |
| Intravenous thrombolysis | 0.85 (0.82–0.89) | <0.001 | 0.93 (0.88–0.98) | 0.01 |
| Urinary tract infection | 1.85 (1.79–1.92) | <0.001 | 1.72 (1.62–1.82) | <0.001 |
| Pneumonia | 2.41 (2.26–2.57) | <0.001 | 2.03 (1.83–2.26) | <0.001 |
| DVT | 1.72 (1.28–2.31) | <0.001 | 2.05 (1.36–3.09) | <0.001 |
| Intubation or mechanic ventilation | 3.42 (3.09–3.78) | <0.001 | 1.88 (1.58–2.23) | <0.001 |
Abbreviations: AAPI = Asian American/Pacific Islander; CI = confidence interval; HHC = home health care; NatAm/other = Native American or other races; NC = no charge; OR = odds ratio.
Home health care included organized home health service and home hospice.
Facility discharge included discharge to inpatient rehabilitation, skilled nursing facility, hospice facility, long-term care facility, and psychiatric hospital.
Adjusted odds ratios reflected least square mean estimates of the main and interaction effects together.
Hospital bedsize (total number of beds) is adjusted for region, location, and teaching status by HCUP. Please see hcup-us.ahrq.gov/db/vars/hosp_bedsize/nisnote.jsp.
States in each census division: New England (ME, VT, MA, RI, CT); Mid-Atlantic (NY, PA, NJ); East North Central (WI, MI, IL, IN, OH); West North Central (MO, ND, SD, NE, KA, MN, IA); South Atlantic (DE, MD, DC, VA, WV, NC, SC, GA, FL); East South Central (KY, TN, MI); West South Central (OK, TX, AR, LA); Mountain (MT, WY, NV, UT, CO, AZ, NM); Pacific (AK, WA, OR, CA, HI).
The adjusted analyses demonstrated significant regional disparities in discharge to one pathway over others. Compared with residences in New England, patients in other census divisions had lower odds of facility over any home discharge, with the lowest odds in the Pacific division (aOR, 0.66; 95% CI 0.60–0.71). Other regions also had lower odds of HHC discharge over home without HHC, with the lowest in West North Central (aOR, 0.33; 95% CI 0.29–0.38). The associations of neighborhood income, rural vs metro, hospital bedsize, and stroke volumes with post-acute service utilization were very modest.
Proxies of stroke severities that were reported in the administrative database, including intubation or mechanic ventilation, hemiplegia or hemiparesis, and aphasia, were significantly associated with higher odds of facility over home discharge (aOR 3.42, 3.22, and 1.82, respectively). In-hospital complications including pneumonia, urinary tract infection, and deep vein thrombosis were also associated with significantly higher adjusted odds of facility over any home discharge (2.41, 1.85, and 1.72, respectively), as well as HHC discharge over home without HHC (2.03, 1.72, and 2.05, respectively). Comorbidities were associated with discharge to one pathway over others at various degrees (eTable 8).
Racial, Ethnic, and Regional Disparities in Discharge to One Pathway Over Others After Intracerebral Hemorrhage
The associations of sex, race, ethnicity, insurance, and region with discharge to one pathway over others after ICH are shown in Table 3. Compared with White men, the odds of facility over any home discharge after ICH were lower among Hispanic women (0.70), higher among White women (1.12), and comparable among other sex-specific racial and ethnic groups. Similar to ischemic stroke, compared with those with private insurance, uninsured patients had the lowest odds to be discharged to a facility over home (aOR, 0.31; 95% CI 0.25–0.41) and HHC over home without HHC (aOR, 0.70; 95% CI 0.50–0.96). Disparities were again evident among uninsured patients and Medicare/Medicaid beneficiaries with Hispanic patients, relative to White patients, having lower odds of facility discharge over home (aOR, 0.44 for uninsured and 0.74 for Medicare/Medicaid; 95% CI 0.27–0.73 and 0.63–0.88, respectively). Relative to White patients, Hispanic patients also had lower odds of HHC discharge over home without HHC in all insurance strata but the odds were the lowest among uninsured Hispanic patients (aOR 0.33 for uninsured, 0.81 for private insurance, and 0.82 for Medicare/Medicaid patients; 95% CI 0.30–0.36, 0.78–0.84, and 0.79–0.93).
Table 3.
Factors Associated With Discharge to One Pathway vs Others After Intracerebral Hemorrhage
| Factor | Facilitya over any home | HHCb over home without HHC | ||
| Adjusted OR (95% CI) | p Value | Adjusted OR (95% CI) | p Value | |
| Age, (reference: ≥80 y) | ||||
| 18–49 y | 0.28 (0.24–0.33) | <0.001 | 0.14 (0.11–0.19) | <0.001 |
| 50–64 y | 0.43 (0.38–0.50) | <0.001 | 0.25 (0.20–0.32) | <0.001 |
| 65–79 y | 0.65 (0.58–0.72) | <0.001 | 0.41 (0.34–0.50) | <0.001 |
| Race, ethnicity, and sex (reference: White male) | ||||
| White female | 1.12 (1.00–1.25) | 0.04 | 1.23 (1.02–1.47) | 0.03 |
| Black female | 1.02 (0.87–1.21) | 0.79 | 1.65 (1.25–2.17) | <0.001 |
| Black male | 1.03 (0.87–1.22) | 0.72 | 1.00 (0.75–1.33) | 0.99 |
| Hispanic female | 0.70 (0.56–0.86) | <0.001 | 1.18 (0.87–1.61) | 0.29 |
| Hispanic male | 0.90 (0.74–1.08) | 0.26 | 1.12 (0.84–1.47) | 0.44 |
| AAPI/NatAm/other female | 0.83 (0.67–1.04) | 0.10 | 1.80 (1.33–2.43) | <0.001 |
| AAPI/NatAm/other male | 1.13 (0.91–1.40) | 0.27 | 0.79 (0.57–1.08) | 0.14 |
| Primary insurance (reference: private insurance) | ||||
| Medicare/Medicaid | 1.11 (0.99–1.24) | 0.08 | 1.33 (1.10–1.62) | 0.004 |
| Self-pay/NC/other | 0.32 (0.25–0.41) | <0.001 | 0.70 (0.50–0.96) | 0.03 |
| Race, ethnicity, and insurance interactionc (reference: White) | ||||
| Medicare/Medicaid | ||||
| AAPI | 0.91 (0.72–1.14) | 0.42 | 1.09 (0.83–1.41) | 0.54 |
| Black | 0.96 (0.83–1.12) | 0.62 | 1.02 (0.80–1.31) | 0.86 |
| Hispanic | 0.74 (0.63–0.88) | <0.001 | 0.82 (0.79–0.93) | <0.001 |
| Private insurance | ||||
| AAPI | 1.23 (0.86–1.74) | 0.26 | 1.35 (0.86–2.11) | 0.19 |
| Black | 1.13 (0.90–1.44) | 0.26 | 1.71 (1.18–2.66) | 0.005 |
| Hispanic | 0.98 (0.73–1.30) | 0.87 | 0.81 (0.78–0.84) | <0.001 |
| Self-pay | ||||
| AAPI | 0.45 (0.17–1.20) | 0.11 | 0.86 (0.38–1.93) | 0.71 |
| Black | 0.54 (0.33–0.87) | 0.01 | 1.19 (0.60–2.34) | 0.62 |
| Hispanic | 0.44 (0.27–0.73) | 0.002 | 0.33 (0.30–0.36) | <0.001 |
| Median household income for patient zip codes (reference: 4th quartile) | ||||
| First quartile | 1.00 (0.87–1.14) | 1.00 | 0.89 (0.72–1.12) | 0.32 |
| Second quartile | 1.01 (0.89–1.15) | 0.89 | 0.98 (0.79–1.21) | 0.86 |
| Third quartile | 0.98 (0.86–1.10) | 0.71 | 0.95 (0.78–1.16) | 0.64 |
| Patient residence (reference: large metro) | ||||
| Small metro | 1.05 (0.95–1.17) | 0.33 | 0.98 (0.82–1.16) | 0.79 |
| Micropolitan | 0.86 (0.70–1.06) | 0.15 | 1.02 (0.71–1.47) | 0.92 |
| Rural | 0.96 (0.76–1.20) | 0.74 | 0.91 (0.63–1.30) | 0.59 |
| Hospital characteristics | ||||
| Hospital bedsized (reference: large) | ||||
| Small | 0.92 (0.78–1.09) | 0.33 | 1.41 (1.07–1.86) | 0.01 |
| Medium | 0.94 (0.84–1.05) | 0.27 | 1.09 (0.91–1.31) | 0.33 |
| Annual volume of intracerebral hemorrhage (reference: 51–100) | ||||
| 5–50 | 0.89 (0.80–1.00) | 0.05 | 0.95 (0.78–1.14) | 0.57 |
| 101–300 | 1.03 (0.92–1.16) | 0.60 | 0.83 (0.68–1.01) | 0.06 |
| 301–500 | 0.96 (0.74–1.24) | 0.76 | 1.29 (0.77–2.15) | 0.34 |
| Hospital location (reference: rural) | ||||
| Urban, nonteaching | 0.71 (0.51–0.98) | 0.04 | 0.88 (0.49–1.59) | 0.67 |
| Urban, teaching | 0.76 (0.55–1.04) | 0.09 | 0.85 (0.47–1.53) | 0.59 |
| Hospital ownership (reference: government, nonfederal) | ||||
| Private, nonprofit | 1.08 (0.94–1.25) | 0.27 | 1.41 (1.09–1.83) | 0.009 |
| Private, investor-owned | 1.10 (0.91–1.32) | 0.32 | 1.31 (0.95–1.81) | 0.10 |
| Hospital census divisione (reference: New England) | ||||
| Middle Atlantic | 0.83 (0.65–1.06) | 0.13 | 0.76 (0.51–1.12) | 0.17 |
| East North Central | 0.90 (0.70–1.15) | 0.39 | 0.48 (0.32–0.72) | <0.001 |
| West North Central | 0.82 (0.62–1.09) | 0.17 | 0.53 (0.32–0.87) | 0.01 |
| South Atlantic | 0.69 (0.54–0.86) | 0.001 | 0.86 (0.59–1.25) | 0.43 |
| East South Central | 0.84 (0.64–1.09) | 0.18 | 0.60 (0.38–0.94) | 0.03 |
| West South Central | 0.63 (0.49–0.82) | <0.001 | 0.67 (0.45–1.01) | 0.06 |
| Mountain | 0.67 (0.51–0.89) | 0.005 | 0.76 (0.47–1.21) | 0.24 |
| Pacific | 0.53 (0.42–0.68) | <0.001 | 0.76 (0.52–1.11) | 0.15 |
| Admission variables | ||||
| Weekend admission | 1.10 (1.00–1.21) | 0.04 | 0.97 (0.83–1.13) | 0.67 |
| Hemiplegia/paralysis | 3.85 (3.51–4.23) | <0.001 | 1.84 (1.57–2.15) | <0.001 |
| Aphasia | 2.04 (1.26–3.29) | 0.004 | 1.09 (0.51–2.31) | 0.82 |
| Urinary tract infection | 1.65 (1.43–1.90) | <0.001 | 1.37 (1.08–1.74) | 0.009 |
| Pneumonia | 1.87 (1.56–2.25) | <0.001 | 1.43 (1.01–2.03) | 0.04 |
| Intubation or mechanic ventilation | 3.41 (2.92–3.99) | <0.001 | 1.84 (1.37–2.45) | <0.001 |
Abbreviations: AAPI = Asian American/Pacific Islander; CI = confidence interval; HHC = home health care; NatAm/other = Native American or other races; NC = no charge; OR = odds ratio.
Home health care included organized home health service and home hospice.
Facility discharge included discharge to inpatient rehabilitation, skilled nursing facility, hospice facility, long-term care facility, and psychiatric hospital.
Adjusted odds ratios reflected least square mean estimates of the main and interaction effects together.
Hospital bedsize (total number of beds) is adjusted for region, location, and teaching status by HCUP. Please see hcup-us.ahrq.gov/db/vars/hosp_bedsize/nisnote.jsp.
States in each census division: New England (ME, VT, MA, RI, CT); Mid-Atlantic (NY, PA, NJ); East North Central (WI, MI, IL, IN, OH); West North Central (MO, ND, SD, NE, KA, MN, IA); South Atlantic (DE, MD, DC, VA, WV, NC, SC, GA, FL); East South Central (KY, TN, MI); West South Central (OK, TX, AR, LA); Mountain (MT, WY, NV, UT, CO, AZ, NM); Pacific (AK, WA, OR, CA, HI).
Consistent with ischemic stroke, large regional variations in post-acute service utilization were also evident after ICH. Compared with New England, hospitals in other census divisions had lower odds of discharging patients to a facility over any home with the lowest in Pacific (aOR, 0.53; 98% CI, 0.42–0.68), and similarly for HHC over home without HHC discharge with the lowest in East North Central (aOR, 0.48; 95% CI 0.32–0.72).
Proxies of stroke severities in the administrative database, including intubation or mechanic ventilation, hemiplegia or hemiparesis, and aphasia, were significantly associated with higher odds of facility over home discharge (aOR 3.41, 3.85, and 2.04, respectively). In-hospital complications including pneumonia and urinary tract infection were associated with significantly higher adjusted odds of discharge to a facility over home (1.87 and 1.65, respectively), and HHC over home without HHC (1.43, and 1.37, respectively). Comorbidities were associated with discharge to one pathway over others at various degrees (eTable 9).
Discussion
This study examined current disparities by sex, race, ethnicity, insurance, and region in post-acute service utilization after ischemic and hemorrhagic stroke in the United States while accounting for other patient and hospital factors including disease severity. The results demonstrated significant insurance-dependent racial and ethnic disparities as well as large regional variations in post-acute service utilization. Uninsured patients were the least likely to be discharged to a facility or HHC, with uninsured Hispanic patients being the lowest. Although patients with Medicare/Medicaid had the highest post-acute service utilization, the odds of discharge to a facility over home were the lowest among Hispanic Medicare/Medicaid beneficiaries. Hispanic women were less likely to be discharged to a facility over home. Hospitals in the Pacific region had the lowest odds to discharge patients to a facility over home after both ischemic stroke and ICH. Of interest, New England had the highest odds of both facility over any home discharge and HHC over home without HHC discharge. The results are in line with previous studies but add new data about current racial, ethnic, and regional disparities of post-acute service utilization in the United States for both ischemic and hemorrhagic stroke.7-9,40 These findings indicate that targeted interventions to improve post-acute care access for specific regions and patients without insurance, especially Hispanic patients, are needed to ensure optimal functional recovery and successful community transition for all patients in the United States.
Racial disparities of poststroke care have been previously noted, but the specific populations at risk were undefined and remained unclear after series of major policy changes.7-9,12-14,16 Some earlier studies reported that, compared with White patients, Black and Hispanic patients were more likely to be discharged to HHC and less likely to SNFs, but other studies using state data showed that Black patients were more likely to be discharged to facility care than White patients.7-9 Our study using the current US national database demonstrated that compared with White men, only Hispanic women were less likely to be discharged to a facility over home, but not other race/ethnicity and sex combined subgroups. The lower formal health care utilization among Hispanic patients were further evident after stratifying the patients by insurance. Relative to White patients, the odds of facility over home discharge were 25% lower in Hispanic patients without insurance and 20% lower in Hispanic patients with Medicare or Medicaid. For Black patients, our study results were consistent with some previous studies,9,41 but different from others.7,8 Our results demonstrated no overall racial disparities in post-acute service utilization, and among Medicare/Medicaid beneficiaries and those with private insurance, Black patients had more post-acute care utilization than White patients.
This study demonstrated that the racial and ethnic disparities of poststroke care depended on the insurance status. The results demonstrate that Hispanic patients without insurance had the lowest formal health care utilization while Hispanic Medicare/Medicaid beneficiaries also had lower odds of facility over home discharge than White Medicare/Medicaid beneficiaries. The lack of formal post-acute care may be a major contributor of the worse functional outcomes among Hispanic stroke survivors.42,43 A previous survey suggested that lack of insurance, language barrier, suboptimal patient-provider interactions, and cultural preference underlie the low health care utilization among Hispanic population.44,45 Studies have suggested that older Hispanic patients have a greater tendency to rely on family support for their care needs instead of using a formal/professional caregiver or facility.44,46 Strong family and social support have been associated with better mental health outcomes among Hispanic families.47 However, when family members are not available or not capable, the needs of the patients may go unattended resulting in adverse safety or functional outcomes.44,48 Further health equity intervention should consider providing health insurance and culturally tailored post-acute care to all Americans and not missing uninsured Hispanic population.
Our study provided disease-specific data about the regional disparities in post-acute service utilization after ischemic and hemorrhagic stroke. We found significant regional disparities in each discharge destination as well as discharge to one pathway over others. Hospitals in New England (ME, VT, MA, RI, CT) had the highest odds of facility over home discharge, as well as HHC over home without HHC discharge. The same patient would have 2-fold the odds to be discharged to a facility over home if they would have been hospitalized in New England vs Pacific region (AK, WA, OR, CA, HI). A patient would have 3-fold the odds to be discharged to a HHC over home without HHC if they would have been hospitalized in New England vs East North Central region (IL, IN, MI, OH, WI). Resource-utilization mismatch existed because the regional variations were not consistent with the number of HHC agencies or facility beds in each region.6,49 States in New England had fewer SNF beds and HHC agencies than the South and Midwest.6,49 West North Central and South regions (South Atlantic, East South Central, and West South Central divisions) had more HHC and post-acute care facilities but lower facility and HHC discharge than New England.6,49 Further studies of the variations and outcomes at state level coupled with policy analysis may guide regulatory changes to reduce regional disparities of post-acute care.
This study has several limitations. First, IRF and SNF were combined as one group because they could not be discerned in NIS. Second, this study used the ICD-10 codes that reflected claims billing, which could be affected by the incentives of maximizing reimbursement.50 Third, the study did not adjust for NIHSS because the reporting rate of NIHSS was low and a previous study showed that NIHSS reporting and missingness in NIS were not random.34 To mitigate this issue, we included several proxies of stroke severity including hemiplegia/paralysis, aphasia, in-hospital infections, and intubation or mechanic ventilation. Fourth, despite that we adjusted for several socioeconomic factors, comorbidities, and index stroke hospitalization information, there may be residual confounders and other social factors including the presence of other caregivers at home or marital/partnership status.
The study demonstrated a distinct pattern of disparities in post-acute service utilization after both ischemic stroke and ICH in the United States by sex, race/ethnicity, insurance, and region. Uninsured patients had lower odds to be discharged to post-acute facilities or HHC, with uninsured Hispanic patients being the lowest. Among Medicare beneficiaries, relative to White patients, Hispanic patients also had lower odds of facility over home discharge. There were large regional disparities and resource-utilization mismatch across the United States, with the Pacific census division having significantly lower post-acute service utilization and New England having the highest utilization. Targeted efforts are needed to improve post-acute care access for uninsured patients especially uninsured Hispanic patients and people in certain regions.
Acknowledgment
The authors thank the supported of the Center for Populations Health Research (CPHR) Collaboration Awards at Cleveland Clinic on this research.
Appendix. Authors
| Name | Location | Contribution |
| Shumei Man, MD, PhD | Cerebrovascular Center, Department of Neurology, Cleveland Clinic | Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data |
| David Bruckman, MS | Center for Population Health Research, Department of Quantitative Health Sciences, Cleveland Clinic | Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data |
| Ken Uchino, MD | Cerebrovascular Center, Neurological Institute, Cleveland Clinic | Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data |
| Jesse D. Schold, PhD | Center for Population Health Research, Department of Quantitative Health Sciences, Cleveland Clinic | Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data |
| Jarrod Dalton, PhD | Center for Population Health Research, Department of Quantitative Health Sciences, Cleveland Clinic | Drafting/revision of the manuscript for content, including medical writing for content |
Study Funding
The authors report no targeted funding.
Disclosure
K. Uchino reported being on data safety monitoring board for clinical trials sponsored by Genentech, Inc. and Evaheart, Inc. Consultant for Abbott Laboratories, Inc. The other authors reported no disclosures. Full disclosure form information provided by the authors is available with the full text of this article at Neurology.org/cp.
References
- 1.Virani SS, Alonso A, Benjamin EJ, et al.; American Heart Association Council on Epidemiology and Prevention Statistics Committee and Stroke Statistics Subcommittee. Heart disease and stroke statistics—2020 update: a report from the American Heart Association. Circulation. 2020;141(9):e139-e596. doi: 10.1161/CIR.0000000000000757 [DOI] [PubMed] [Google Scholar]
- 2.Winstein CJ, Stein J, Arena R, et al.; American Heart Association Stroke Council, Council on Cardiovascular and Stroke Nursing, Council on Clinical Cardiology, and Council on Quality of Care and Outcomes Research. Guidelines for adult stroke rehabilitation and recovery: a guideline for healthcare professionals from the American Heart Association/American Stroke Association. Stroke. 2016;47(6):e98-e169. doi: 10.1161/STR.0000000000000098 [DOI] [PubMed] [Google Scholar]
- 3.Chan L, Wang H, Terdiman J, et al. Disparities in outpatient and home health service utilization following stroke: results of a 9-year cohort study in Northern California. PM R. 2009;1(11):997-1003. doi: 10.1016/j.pmrj.2009.09.019 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Gage B. Impact of the BBA on post-acute utilization. Health Care Financ Rev. 1999;20(4):103-126. [PMC free article] [PubMed] [Google Scholar]
- 5.Bogasky S. Examining Relationships in an Integrated Hospital System | ASPE. Accessed June 3, 2022. aspe.hhs.gov/sites/default/files/migrated_legacy_files//42751/report.pdf [Google Scholar]
- 6.Wang Y, Zhang Q, Spatz ES, et al. Persistent geographic variations in availability and quality of nursing home care in the United States: 1996 to 2016. BMC Geriatr. 2019;19(1):103. doi: 10.1186/s12877-019-1117-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Morgenstern LB, Sais E, Fuentes M, et al. Mexican Americans receive less Intensive stroke rehabilitation than non-hispanic whites. Stroke. 2017;48(6):1685-1687. doi: 10.1161/STROKEAHA.117.016931 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Kind AJH, Smith MA, Liou JI, Pandhi N, Frytak JR, Finch MD. Discharge destination's effect on bounce-back risk in Black, White, and Hispanic acute ischemic stroke patients. Arch Phys Med Rehabil. 2010;91(2):189-195. doi: 10.1016/j.apmr.2009.10.015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Cho JS, Hu Z, Fell N, Heath GW, Qayyum R, Sartipi M. Hospital discharge disposition of stroke patients in Tennessee. South Med J. 2017;110(9):594-600. doi: 10.14423/SMJ.0000000000000694 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Onukwugha E, Mullins CD. Racial differences in hospital discharge disposition among stroke patients in Maryland. Med Decis Mak. 2007;27(3):233-242. doi: 10.1177/0272989X07302130 [DOI] [PubMed] [Google Scholar]
- 11.Xian Y, Thomas L, Liang L, et al. Unexplained variation for hospitals' use of inpatient rehabilitation and skilled nursing facilities after an acute ischemic stroke. Stroke. 2017;48(10):2836-2842. doi: 10.1161/STROKEAHA.117.016904 [DOI] [PubMed] [Google Scholar]
- 12.McWilliams JM, Gilstrap LG, Stevenson DG, Chernew ME, Huskamp HA, Grabowski DC. Changes in postacute care in the medicare shared savings program. JAMA Intern Med. 2017;177(4):518-526. doi: 10.1001/jamainternmed.2016.9115 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Dummit LA, Kahvecioglu D, Marrufo G, et al. Association between hospital participation in a medicare bundled payment initiative and payments and quality outcomes for lower extremity joint replacement episodes. JAMA. 2016;316(12):1267-1278. doi: 10.1001/jama.2016.12717 [DOI] [PubMed] [Google Scholar]
- 14.Bundled Payments for Care Improvement (BPCI) Initiative: General Information | CMS Innovation Center. Accessed May 4, 2021. innovation.cms.gov/innovation-models/bundled-payments [Google Scholar]
- 15.Olmsted MG, Powell R, Murphy J, et al. Methodology U.S. News & World Report 2020-21 Best Hospitals: Specialty Rankings. Accessed January 14, 2021. health.usnews.com/media/best-hospitals/BH_Methodology_2020-21 [Google Scholar]
- 16.Agarwal R, Liao JM, Gupta A, Navathe AS. The impact of bundled payment on health care spending, utilization, and quality: a systematic review. Health Aff. 2020;39(1):50-57. doi: 10.1377/hlthaff.2019.00784 [DOI] [PubMed] [Google Scholar]
- 17.The Lewin Group. CMS Bundled Payments for Care Improvement Initiative Models 2-4: Year 5 Evaluation & Monitoring Annual Report. Centers Medicare Medicaid Serv. Accessed December 14, 2022. downloads.cms.gov/files/cmmi/bpci-models2-4-yr5evalrpt.pdf [Google Scholar]
- 18.Adil MM, Vidal GA, Beslow LA. Clinical outcomes among transferred children with ischemic and hemorrhagic strokes in the nationwide inpatient sample. J Stroke Cerebrovasc Dis. 2016;25(11):2594-2602. doi: 10.1016/j.jstrokecerebrovasdis.2016.06.043 [DOI] [PubMed] [Google Scholar]
- 19.Reimer AP, Schiltz N, Koroukian SM, Madigan EA. National incidence of medical transfer: patient characteristics and regional variation. J Health Hum Serv Adm. 2016;38(4):509-528. doi: 10.1177/107937391603800404 [DOI] [PubMed] [Google Scholar]
- 20.Singh T, Peters SR, Tirschwell DL, Creutzfeldt CJ. Palliative care for hospitalized patients with stroke: results from the 2010 to 2012 national inpatient sample. Stroke. 2017;48(9):2534-2540. doi: 10.1161/STROKEAHA.117.016893 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Otite FO, Liaw N, Khandelwal P, et al. Increasing prevalence of vascular risk factors in patients with stroke: a call to action. Neurology. 2017;89(19):1985-1994. doi: 10.1212/WNL.0000000000004617 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Otite FO, Khandelwal P, Malik AM, Chaturvedi S. National patterns of carotid revascularization before and after the carotid revascularization endarterectomy vs stenting trial (CREST). JAMA Neurol. 2018;75(1):51-57. doi: 10.1001/jamaneurol.2017.3496 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Rockville M. NIS Database Documentation. Accessed April 3, 2022. hcup-us.ahrq.gov/db/nation/nis/nisdbdocumentation.jsp [Google Scholar]
- 24.Introduction to the HCUP National Inpatient Sample (NIS), 2018. Accessed December 1, 2023. hcup-us.ahrq.gov/db/nation/nis/NIS_Introduction_2018.jsp#howto [Google Scholar]
- 25.Producing National HCUP Estimates - Accessible Version. Accessed December 2, 2023. hcup-us.ahrq.gov/tech_assist/nationalestimates/508_course/508course_2018.jsp#weighting [Google Scholar]
- 26.Van Walraven C, Austin PC, Jennings A, Quan H, Forster AJ. A modification of the Elixhauser comorbidity measures into a point system for hospital death using administrative data. Med Care. 2009;47(6):626-633. doi: 10.1097/MLR.0b013e31819432e5 [DOI] [PubMed] [Google Scholar]
- 27.Elixhauser Comorbidity Software Refined for ICD-10-CM. Accessed April 3, 2022. hcup-us.ahrq.gov/toolssoftware/comorbidityicd10/comorbidity_icd10.jsp [Google Scholar]
- 28.Race and Ethnicity Data Improvement Toolkit. Accessed July 13, 2021. hcup-us.ahrq.gov/datainnovations/raceethnicitytoolkit/case_for_re.jsp [Google Scholar]
- 29.Howland RE, Tsao TY. Evaluating race and ethnicity reported in hospital discharge data and its impact on the assessment of health disparities. Med Care. 2020;58(3):280-284. doi: 10.1097/MLR.0000000000001259 [DOI] [PubMed] [Google Scholar]
- 30.Greene WH. Econometric Analysis (Seventh ed.). Pearson Education; 2012:803-806. [Google Scholar]
- 31.Lichtman JH, Leifheit-Limson EC, Jones SB, Wang Y, Goldstein LB. 30-day risk-standardized mortality and readmission rates after ischemic stroke in critical access hospitals. Stroke. 2012;43(10):2741-2747. doi: 10.1161/STROKEAHA.112.665646 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Man S, Schold JDD, Uchino K. Impact of stroke center certification on mortality after ischemic stroke: the medicare cohort from 2009 to 2013. Stroke. 2017;48(9):2527-2533. doi: 10.1161/STROKEAHA.116.016473 [DOI] [PubMed] [Google Scholar]
- 33.Man S, Bruckman D, Tang AS, Uchino K, Schold JD. The association of socioeconomic status and discharge destination with 30-day readmission after ischemic stroke. J Stroke Cerebrovasc Dis. 2021;30(12):106146. doi: 10.1016/j.jstrokecerebrovasdis.2021.106146 [DOI] [PubMed] [Google Scholar]
- 34.Saber H, Saver JL. Distributional validity and prognostic power of the National INSTITUTES of Health Stroke Scale in US Administrative Claims Data. JAMA Neurol. 2020;77(5):606-612. doi: 10.1001/JAMANEUROL.2019.5061 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Simpson AN, Wilmskoetter J, Hong I, et al. Stroke administrative severity index: using administrative data for 30-day poststroke outcomes prediction. J Comp Eff Res. 2018;7(4):293-304. doi: 10.2217/CER-2017-0058 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Sung SF, Hsieh CY, Kao Yang YH, et al. Developing a stroke severity index based on administrative data was feasible using data mining techniques. J Clin Epidemiol. 2015;68(11):1292-1300. doi: 10.1016/j.jclinepi.2015.01.009 [DOI] [PubMed] [Google Scholar]
- 37.Pedersen AB, Mikkelsen EM, Cronin-Fenton D, et al. Missing data and multiple imputation in clinical epidemiological research. Clin Epidemiol. 2017;9:157-166. doi: 10.2147/CLEP.S129785 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Jakobsen JC, Gluud C, Wetterslev J, Winkel P. When and how should multiple imputation be used for handling missing data in randomised clinical trials—a practical guide with flowcharts. BMC Med. Res Methodol 2017;17(1):162. doi: 10.1186/s12874-017-0442-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.HCUP Methods Series Calculating National Inpatient Sample (NIS) Variances for Data Years 2012 and Later. Accessed December 3, 2023. hcup-us.ahrq.gov/reports/methods/2015_09.jsp [Google Scholar]
- 40.Goodwin JS, Howrey B, Zhang DD, Kuo YF. Risk of continued institutionalization after hospitalization in older adults. J Gerontol A Biol Sci Med Sci. 2011;66(12):1321-1327. doi: 10.1093/gerona/glr171 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Skolarus LE, Feng C, Burke JF. No racial difference in rehabilitation therapy across all post-acute care settings in the year following a stroke. Stroke. 2017;48(12):3329-3335. doi: 10.1161/STROKEAHA.117.017290 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Burks JD, Chen SH, Luther EM, et al. Effect of Hispanic status in mechanical thrombectomy outcomes after ischemic stroke: insights from STAR. Stroke. 2021;52(11):E715-E719. doi: 10.1161/STROKEAHA.120.033326 [DOI] [PubMed] [Google Scholar]
- 43.Lisabeth LD, Sánchez BN, Baek J, et al. Neurological, functional, and cognitive stroke outcomes in Mexican Americans. Stroke. 2014;45(4):1096-1101. doi: 10.1161/STROKEAHA.113.003912 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Min JW, Barrio C. Cultural values and caregiver preference for Mexican-American and non-Latino White elders. J Cross Cult Gerontol. 2009;24(3):225-239. doi: 10.1007/S10823-008-9088-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Amirehsani KA, Hu J, Wallace DC, et al. U.S. Healthcare experiences of hispanic patients with diabetes and family members: a qualitative analysis. J Community Health Nurs. 2017;34(3):126-135. doi: 10.1080/07370016.2017.1340556 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Katz JN. Patient preferences and health disparities. JAMA. 2001;286(12):1506-1509. doi: 10.1001/JAMA.286.12.1506 [DOI] [PubMed] [Google Scholar]
- 47.Finch BK, Vega WA. Acculturation stress, social support, and self-rated health among Latinos in California. J Immigr Health. 2003;5(3):109-117. doi: 10.1023/A:1023987717921 [DOI] [PubMed] [Google Scholar]
- 48.Dietz TL. The Mexican American elderly and supplemental security income: reasons and characteristics associated with nonuse. J Appl Gerontol. 2001;20(3):292-306. doi: 10.1177/073346480102000303 [DOI] [Google Scholar]
- 49.Center for Health Statistics N. Vital and Health Statistics, Series 3, Number 47. Accessed June 8, 2022. cdc.gov/nchs/data/series/sr_03/sr03-047.pdf [Google Scholar]
- 50.Motheral BR, Fairman KA. The use of claims databases for outcomes research: Rationale, challenges, and strategies. Clin Ther. 1997;19(2):346-366. doi: 10.1016/S0149-2918(97)80122-1 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The data used in this study will not be shared by the authors because of our Data Use Agreement with the HCUP, which prohibits data redistribution. The HCUP Central Distributor is the entity that accepts, processes, and fulfills applications for the databases and manages data use agreements for all data users. Investigators may request access to anonymized patient data at distributor.hcup-us.ahrq.gov/.