Post-acute care transitions during COVID-19: racial, ethnic, and socioeconomic differences in older adults with Alzheimer’s disease and related dementia

Abstract

Background.

Differences in the post-acute care (PAC) destinations among racial, ethnic, and socioeconomic groups have been documented before the COVID-19 pandemic. Yet, the pandemic’s impact on these differences remains unknown. We examined the impact of the COVID-19 pandemic on PAC destinations and its variation by individual race, ethnicity, and socioeconomic status among community-dwelling older adults with Alzheimer’s disease and related dementia (ADRD).

Methods.

We linked 2019-2021 national data (Medicare claims, Minimum Data Set, Master Beneficiary Summary File) and several publicly available datasets, including Provider of Services File, Area Deprivation Index, Area Health Resource File, and COVID-19 infection data. PAC discharge destinations included skilled nursing facilities (SNF), home health agencies (HHA), and homes without services. Key variables of interest included individual race, ethnicity, and Medicare-Medicaid dual status. The analytic cohort included 830,656 community-dwelling Medicare fee-for-service beneficiaries with ADRD who were hospitalized between 2019 and 2021. Regression models with hospital random effects and state fixed effects were estimated, stratified by the time periods, and adjusted for the individual, hospital, and county-level covariates.

Results.

SNF discharges decreased while home and HHA discharges increased during the pandemic. The trend was more prominent among racial and ethnic minoritized groups and even more so among dual-eligible beneficiaries. For instance, the reduction in the probabilities of SNF admissions between the pre-pandemic period and the 2^nd year of COVID was 4.6 (White non-duals), 18.5 (White duals), 8.7 (Black non-duals), and 20.1 (Black duals) percentage-point, respectively. We also found that non-duals were more likely to replace SNF with HHA services, while duals were more likely to be discharged home without HHA.

Conclusions.

The COVID-19 pandemic significantly impacted PAC destinations for individuals with ADRD, especially among socioeconomically disadvantaged and racial and ethnic minoritized populations. Future research is needed to understand if and how these transitions may have affected health outcomes.

INTRODUCTION

The COVID-19 pandemic has profoundly impacted various sectors of the U.S. healthcare system, including post-acute care (PAC). PAC provides continuing skilled services for Medicare beneficiaries after an acute hospitalization. ¹ The utilization of PAC services among Medicare beneficiaries has rapidly increased in the past few decades, with over 40% now utilizing these services after acute hospitalizations. ^{2, 3} Skilled nursing facilities (SNFs) and home health agencies (HHAs) are the two most common settings and account for the majority (90%) of PAC services. ² However, after the pandemic outbreak, PAC discharge destination following a hospital stay has changed significantly. For instance, Werner et al. reported that in the first year of the COVID-19 outbreak, the proportion of PAC SNF discharges decreased, while the proportion of discharges to home with self-care and/or with HHA increased. ⁴ These shifts indicate a fundamental transformation in how PAC services were utilized.

Among hospitalized patients, those with Alzheimer’s disease and related dementia (ADRD) are a particularly vulnerable group due to their unique healthcare needs and potential access barriers to care. ^5-7 Prior to the pandemic, patients with ADRD tended to use SNF more frequently than other PAC services. ^{5, 8-11} Furthermore, PAC trajectories among Medicare beneficiaries with ADRD have been shown to vary by race, ethnicity, and dual status. ¹² Studies have documented racial and ethnic differences in access to nursing homes (NHs), including SNF. ^{13, 14 15} For instance, older adults with ADRD who were identified as Black, Latinx, and non-duals were less likely to be discharged to SNF than their counterparts. Moreover, dual eligibility-related differences were more pronounced than racial and ethnic differences in PAC destinations. ¹² However, despite these findings, it is unknown how the COVID-19 pandemic affected PAC transitions among community-dwelling older adults with ADRD, who heavily relied on SNF discharges before the pandemic, and how such impact varied by individual race, ethnicity, and socioeconomic status. Considering the disproportionate impact of COVID-19 on NHs, these differences in PAC destinations may have been exacerbated. ¹⁶ Additionally, it is unclear if the changes in PAC trajectories varied during the course of the COVID-19 pandemic.

To address these knowledge gaps, we examined the impact of the COVID-19 pandemic on the distribution of PAC transitions among community-dwelling older adults with ADRD who were hospitalized for non-COVID-related conditions by using 2019 and 2021 national data. Additionally, we explored differences in PAC destinations by race, ethnicity, and socioeconomic status, further contributing to an understanding of the pandemic’s impact on the distribution of PAC utilization among these vulnerable populations.

METHODS

Data

This study used multiple national data between 2019 and 2021, including Medicare claims data, Minimum Data Set (MDS), Master Beneficiary Summary File (MBSF), Provider of Services File (POS), Area Deprivation Index (ADI), Area Health Resource File (AHRF), and the USAFacts.org public website. ^17-21 Medicare claims (i.e., inpatient, SNF, and home health files) and the MDS data were linked by unique Centers for Medicare and Medicaid Services (CMS) beneficiary ID (Bene ID) and used to determine acute inpatient events and discharge destinations. We used the MBSF to identify beneficiaries who were fee-for-service (FFS) and to obtain their demographics and chronic disease information. Medicare inpatient file was used to identify acute hospitalizations among Medicare FFS beneficiaries and to determine the characteristics of these hospitalizations, such as the admission and discharge date, the primary diagnosis, and the Diagnosis-Related Group (DRG) codes. We also obtained the DRG weights, which reflect the complexity of patients’ conditions. ¹⁸

To gain a better understanding of the characteristics of hospitals where patients received care, we merged the individual-level datasets with the POS file by the hospital provider number. ¹⁷ The POS file provided hospital characteristics, such as the number of beds, teaching status, and ownership. We then linked our datasets with the 2019 ADI, the 2019-2021 AHRF, and the data from USAFacts.org public website (https://usafacts.org/visualizations/coronavirus-covid-19-spread-map/).^19-21 ADI was linked with the Medicare data using the 5-digit ZIP codes from the MBSF, while the AHRF and USAFacts data were linked using the Social Security Administration (SSA) state and county codes. The ADI captures the socioeconomic status of the community in which an individual resides. The AHRF contains factors reflecting county-level PAC supply (e.g., SNF beds and number of HHA per 1000 populations aged 65-year-old and older). The data from USAFacts.org provide information on monthly county-level newly reported COVID-19 cases. By leveraging multiple data sources, we aimed to capture a comprehensive set of factors that might influence patients’ PAC destinations.

Cohort

We included community-dwelling Medicare beneficiaries age 65 and older with a diagnosis of ADRD and with at least one hospitalization between January 1, 2019, and November 30, 2021. The diagnosis of ADRD was based on the MBSF chronic condition file. The community-dwelling population was defined as those who did not have any nursing home (NH) stay 90 days prior to the index hospitalization (defined below). We then restricted the cohort to those who were continuously enrolled in the Medicare fee-for-service (FFS) six months before and one month after the index hospitalization. We included non-Latinx White, non-Latinx Black, or Latinx Medicare beneficiaries based on the validated RTI classification, as these groups represent the major racial and ethnic groups in the U.S. ²²

We used a set of criteria to define eligible index hospitalizations. Among the included beneficiaries, we first identified 2,552,721 hospitalizations that did not end in in-hospital death and those whose principal diagnosis was not related to COVID-19 (i.e., with the ICD-10 codes U071, J1282, B9729, Z8616). ²³ We did not include COVID-related hospitalizations in the main analysis to maintain comparability with the pre-pandemic data. In addition, we focused on medical inpatient events (N=2,086,377) rather than surgical events because 1) medical conditions accounted for the majority of the hospitalizations (81.7%) in our dataset, and 2) the post-acute care needs of patients who were hospitalized for medical conditions could be very different from those who were hospitalized after a surgical event. We then excluded those hospitalizations with fewer than 3-day stays (N=503,012) as they would not be eligible for SNF coverage prior to the pandemic. We further excluded hospital stays ending with transfers to another acute hospital or to hospice (N=111,659). Among the remaining 1,471,706 eligible hospitalizations, for beneficiaries with multiple hospitalizations, we randomly selected one hospitalization per beneficiary per time period (i.e., Jan 2019 to Feb 2020 as the pre-pandemic period, Mar 2020 to Dec 2020 as the 1^st year of COVID, and Jan 2021 to Nov 2021as the 2^nd year of COVID). Our final analytical cohort consisted of 989,374 hospitalizations among 830,656 community-dwelling Medicare beneficiaries with ADRD.

Variables

The outcome of interest was PAC discharge destinations, defined as a categorical variable, with discharges to SNF, HHA, or homes (i.e., without HHA services) within three days of a hospital discharge.

The key independent variables included race and ethnicity (non-Latinx White, non-Latinx Black, and Latinx individuals), Medicare-Medicaid dual status (as a proxy for socioeconomic status), and the time periods. To examine whether racial or ethnic differences in PAC destinations would vary by dual status, we also included interaction terms between racial/ ethnicity and dual status. Time periods were defined as pre-pandemic, 1^st year of COVID, and 2^nd year of COVID. The Medicare-Medicaid dual status (i.e., whether an individual is fully dual-eligible or not) was determined by MBSF at the time of hospital admission. ²⁴

We included individual-, hospital-, and county-level covariates to control for potential confounding factors that may affect patients' PAC destinations. Individual-level characteristics included age, sex, chronic conditions, prior 90-day healthcare utilization (e.g., HHA), and the socioeconomic conditions (i.e., ADI) of the community where an individual resided. ²⁵ We considered a community with 85 or greater ADI scores as a poorer community. ²⁵ We obtained the characteristics of the index hospitalizations, including length of stay (LOS), intensive care unit (ICU) utilization, and DRG weights. Hospital characteristics were obtained from the POS file and included rural/urban locations (based on the core-based statistical area [CBSA] designation where the hospital is located), number of beds, ownership, medical school affiliation, and the percentages of racially or ethnically minoritized patients. For the county-level PAC supply variables, we included the number of HHAs and SNF beds per 1000 populations aged 65 or older per county and monthly cases of newly reported COVID-19.

Statistical analysis

We first calculated the unadjusted rates for each type of PAC destination to describe the trends in PAC discharges before and during the pandemic. To further explore if these changes varied by race, ethnicity, and dual status, we conducted multinomial logistic regressions with hospital random effects and state fixed effects, stratified by the time periods (i.e., pre-pandemic, 1^st year, and 2^nd year of COVID) and adjusting for the covariates as described above. The state fixed effects accounted for the impact of time-invariant factors and between-state heterogeneity. Stratified analyses allowed for variations in the relationships between all covariates and PAC destinations across the three periods. We also computed adjusted probabilities by race, ethnicity, and dual status for the ease of interpretation of the results.

While our analysis primarily centered on non-COVID-19-related hospitalizations, it's essential to acknowledge that COVID-19-related hospitalizations might exhibit distinct PAC trajectories and may have varying effects on different racial, ethnic, and socioeconomic groups. To ensure the robustness of our current findings, we conducted a sensitivity analysis by including COVID-19-related hospitalizations. The study has been approved by the University of Rochester Institutional Review Board. All statistical analyses were performed in SAS 9.4 (SAS Institute Inc, Cary, NC) and Stata 17.0 (StataCorp LLC, College Station, TX).

RESULTS

In Table 1 we present the descriptive characteristics of the study cohort, stratified by the time periods. Of the 989,374 hospitalizations, 80.9% were among non-Latinx White, 12.1% were non-Latinx Black, and 6.9% were Latinx older adults. Beneficiaries with full dual status made up 26.1% of the cohort. The majority (76.2%) were aged 75 years or older and male (62.1%).

Table 1.

Comparison of selected individual characteristic by the time periods

	All (N=989,374)	Pre-pandemic (N=483,408)	1st year of COVID (N=265,361)	2nd year of COVID (N=240,605)
Dependent variable
Home	335,335 (33.9%)	155,286 (32.1%)	94,191 (35.5%)	85,858 (35.7%)
SNF	426,672 (43.1%)	231,938 (48.0%)	100,646 (37.9%)	94,088 (39.1%)
HHA	227,367 (23.0%)	96,184 (19.9%)	70,524 (26.6%)	60,659 (25.2%)
Independent variable
Non-Latinx White older adults	800,739 (80.9%)	391,792 (81.0%)	213,731 (80.5%)	195,216 (81.1%)
Non-Latinx Black older adults	120,156 (12.2%)	58,335 (12.1%)	33,634 (12.7%)	28,187 (11.7%)
Latinx older adults	68,479 (6.9%)	33,281 (6.8%)	17,996 (6.8%)	17,202 (7.2%)
Dual-eligible	258,428 (26.1%)	132,339 (27.4%)	66,839 (25.2%)	59,069 (24.6%)
Selected individual covariates
Age (65-74)	235,823 (23.8%)	119,564 (24.7%)	62,760 (23.7%)	53,499 (22.2%)
75-84	329,083 (33.3%)	161,170 (33.3%)	88,108 (33.2%)	79,805 (33.2%)
>=85	424,468 (42.9%)	202,674 (42.0%)	114,493 (43.2%)	107,301 (44.6%)
Female	374,815 (37.9%)	180,003 (37.2%)	103,212 (38.9%)	91,600 (38.1%)

Pre-pandemic: 2019.01.01-2020.02.29; 1^st year of COVID: 2020.03.01-2020.12.31; 2^nd year of COVID: 2021.01.01-2021.11.30; SNF: Skilled nursing facility; HHA: Home health agency.

Although overall SNF was the most common PAC discharge destination (43.1%), its prevalence decreased from 48.0% prior to the pandemic (2019.01-2020.02) to 37.9% in the first year of the COVID-19 pandemic (2020.03-2020.12) and then to 39.1% in the second year of COVID-19 pandemic (2021). On the other hand, there was an increasing trend in home discharge with or without HHA. Home discharges without services increased from 32.1% before the pandemic to 35.5% and 35.7% in the first and second year of COVID-19, respectively, and discharges to HHA increased from 19.9% to 26.6% and 25.2%, respectively. These results suggest that the impact of COVID-19 on the PAC trajectories persisted in the second year of COVID-19, and more patients were discharged directly back home or to HHA.

Relationships between dual status, race and ethnicity, and PAC discharge destination

Table 2 presents the main results from the multinomial logit model, stratified by the three time periods (full results were presented in Appendix Table S2). Overall, older Black and Latinx adults were less likely to be discharged to SNF than their White counterparts. Additionally, duals had a higher likelihood of being discharged to SNF following hospitalizations. Moreover, racial or ethnic differences varied by dual status, as suggested by the interaction terms between racial/ ethnicity and dual status.

Table 2.

Results of Multinomial regression on PAC destinations, stratified by time periods

	Pre-pandemic		1st year of COVID		2nd year of COVID
VARIABLES	SNF vs. Home	HHA vs. Home	SNF vs. Home	HHA vs. Home	SNF vs. Home	HHA vs. Home
Independent variables
Race and Ethnicity (Ref: Non-Latinx White older adults)
Non-Latinx Black older adults	0.01 (0.02)	0.03 (0.02)	−0.21*** (0.03)	−0.03 (0.03)	−0.17*** (0.03)	0.01 (0.03)
Latinx older adults	−0.23*** (0.03)	−0.05* (0.03)	−0.35*** (0.04)	−0.03 (0.03)	−0.32*** (0.04)	0.02 (0.04)
Dual-eligible	0.91*** (0.02)	0.08*** (0.01)	0.70*** (0.02)	−0.11*** (0.02)	0.24*** (0.02)	−0.11*** (0.02)
Non-Latinx Black x Dual	−0.21*** (0.03)	0.00 (0.03)	−0.02 (0.04)	0.12*** (0.04)	−0.06 (0.04)	0.07* (0.04)
Latinx x Dual	−0.63*** (0.04)	0.02 (0.04)	−0.45*** (0.05)	0.05 (0.04)	−0.28*** (0.05)	0.03 (0.05)
Hospital random effects	Y	Y	Y	Y	Y	Y
State fixed effects	Y	Y	Y	Y	Y	Y
Observations	483,408	483,408	265,361	265,361	240,605	240,605
Number of Hospitals	3,109	3,109	3,029	3,029	3,003	3,003

The numbers were the estimated coefficients derived from the multinomial logistic regression; Robust standard errors were presented in parentheses; *** p<0.01, ** p<0.05, * p<0.1; The models has adjusted for individual-level, hospital-level, and county-level covariates; Pre-pandemic: 2019.01.01-2020.02.29; 1^st year of COVID: 2020.03.01-2020.12.31; 2^nd year of COVID: 2021.01.01-2021.11.30; SNF: Skilled nursing facility; HHA: Home Health Agency

To ease the interpretation of the regression results, we calculated the adjusted probability of discharge to home, HHA, and SNF for each race and ethnicity, by dual status and time periods, as shown in Figure 1. Consistent with the results presented in Table 2, White individuals were slightly more likely to be discharged to SNFs, while Black and Latinx individuals were more likely to be discharged to home with or without HHA services. However, the differences in PAC discharges were greater by dual status than race and ethnicity. For instance, before the pandemic, the dual-related differences in the probabilities of SNF discharge were 20.7 percentage-point among White individuals (i.e., 0.456 for non-duals and 0.663 for duals, P<0.01), 15.9 percentage-point among Black individuals (i.e., 0.456 and 0.615 for non-duals and duals, P<0.01), and 5.8 percentage-point among Latinx individuals (i.e., 0.407 and 0.465 for non-duals and duals, P<0.01). On the other hand, before the pandemic, no significant racial differences in SNF discharges were observed for non-duals (i.e., the adjusted probability of SNF discharge was 0.456 versus 0.456). While a 4.8 percentage-point racial difference was found for the duals (i.e., the adjusted probability of SNF discharge was 0.663 and 0.615, respectively), the racial differences were smaller than the dual-related differences.

Figure 1.

Adjusted probabilities of PAC destination for the racial/ethnic groups with different dual status

The impact of the pandemic on PAC discharge destination

Throughout the pandemic, there was a decrease in discharges to SNFs and a corresponding increase in discharges to home or HHA across all groups. However, the magnitude of these changes varied depending on individuals’ dual status, race, and ethnicity. In particular, among older adults with ADRD who were dual, and among those who were Black, there was a more pronounced decrease in SNF discharges and a greater increase in discharges to home or HHA compared to the pre-pandemic period, as illustrated in Figure 1.

We calculated the differences in the probabilities of discharge destinations for each subgroup between pre-pandemic and 2^nd year of COVID, as outlined in Table 3. We did not include 1^st year of COVID in the calculation because of the significant healthcare delivery system disruptions at the early stage of the pandemic. Overall, the changes in PAC discharge destinations during the pandemic were greater among duals than non-duals across all racial and ethnic subgroups. For instance, among non-duals, the adjusted probabilities of SNF discharges declined from 0.456 in the pre-pandemic period to 0.410 in the 2^nd year of COVID (i.e., 4.6 percentage-points reduction) for White individuals, from 0.456 to 0.369 (i.e., 8.7 percentage-points reductions) for Black individuals, and from 0.407 to 0.335 (i.e., 7.2 percentage-points reduction) for Latinx individuals. On the other hand, among the duals during the same period, the adjusted probability of SNF discharge decreased from 0.663 to 0.478 (i.e., 18.5 percentage-points reductions) for White individuals, from 0.615 to 0.414 (i.e., 20.1 percentage-points reduction) for Black individuals, and from 0.465 to 0.332 (i.e., 13.3 percentage-points reductions) for Latinx individuals. Therefore, duals experienced a greater reduction in SNF discharges, ranging from 13.3-20.1 percentage points across the three racial and ethnic subgroups, while non-duals had a reduction of approximately 4.6-8.7 percentage points in the likelihood of SNF discharges. Additionally, within each dual-eligible subgroup, Black individuals had a higher reduction in the likelihood of SNF discharges between the pre-pandemic period and 2^nd year of COVID, as compared to White individuals.

Table 3.

Adjusted probabilities of discharge destination for the racial and ethnic groups with different dual status

		Non-Latinx White			Non-Latinx Black			Latinx
Time periods		Pre	2nd	Diff	Pre	2nd	Diff	Pre	2nd	Diff
SNF	Non-duals	0.456	0.410	−0.046*	0.456	0.369	−0.087*	0.407	0.335	−0.072*
	Duals	0.663	0.478	−0.185*	0.615	0.414	−0.201*	0.465	0.332	−0.133*
HHA	Non-duals	0.206	0.25	0.044*	0.210	0.269	0.059*	0.218	0.285	0.067*
	Duals	0.134	0.208	0.074*	0.156	0.244	0.088*	0.209	0.272	0.063*
Home	Non-duals	0.338	0.340	0.002	0.334	0.362	0.028*	0.376	0.380	0.004
	Duals	0.203	0.315	0.112*	0.229	0.342	0.113*	0.326	0.395	0.069*

Pre: pre-pandemic period; 2^nd: 2^nd year of COVID; SNF: Skilled nursing facility; HHA: Home Health Agency; Diff: differences; Diff represented the differences in adjusted probabilities changed between pre-pandemic to 2^nd year of COVID for each PAC destination; *: p < 0.05.

The shift from SNF discharge to home discharge with and without HHA

To examine the PAC destinations alternative to SNF during the pandemic, we compared the changes in the probability of being discharged to home with or without HHA by race, ethnicity, and dual status. We found that the decrease in SNF discharges among non-duals was primarily replaced by an increase in HHA use between the pre-pandemic period to 2^nd year of COVID, while duals were more likely to be discharged to home without HHA services (Table 3). For example, as presented in Table 3, among non-duals, approximately 96% (equivalent to a 4.4 out of 4.6 percentage-point change), 68% (5.9 out of 8.7 percentage-point change), and 93% (6.7 out of 7.2 percentage-point change) of the reduction of SNF discharges were substituted by discharged to HHA among White, Black, and Latinx individuals, respectively. However, among duals, a higher proportion of the decline in SNF discharge was substituted by discharges to homes without HHA services: 61% (11.2 out of 18.5 percentage-points), 56% (11.3 out of 20.1 percentage-points), and 52% (6.9 out of 13.3 percentage-points) among White, Black, and Latinx individuals, respectively.

Sensitivity analysis

The sensitivity analysis that included COVID-19-related hospitalizations encompassed a total of 1,032,836 hospitalizations, of which 39,475 (3.8%) were COVID-19-related. During the first and second years of the COVID-19 pandemic, a higher proportion of Black and Latinx older adults were observed among the COVID-19-related hospitalizations (23.5%) than non-COVID-19-related hospitalizations (19.1%). Also, COVID-19-related hospitalizations exhibited a greater tendency toward skilled nursing facility (SNF) discharges (47.5% vs. 38.4%) compared to non-COVID-19-related hospitalizations. Nevertheless, as detailed in Table S3, the sensitivity analysis consistently reaffirmed our main findings.

DISCUSSIONS

This study examined the impact of the pandemic on PAC transitions among Medicare beneficiaries with ADRD. Our findings showed that these patients experienced a significant reduction in PAC discharge to SNF and an increase in home and HHA discharges during the pandemic, with the trend persisting into the second year and particularly among Black and Latinx older adults and those who were dually eligible. Furthermore, our findings suggested that the impact of the pandemic on dual-related differences was greater than racial or ethnic-related differences. Our results also highlighted the importance of considering dual status when addressing differences in PAC utilization among Medicare beneficiaries with ADRD during the COVID-19 pandemic.

Consistent with the prior studies, we found a reduction in SNF transitions among community-dwelling Medicare beneficiaries with ADRD during the pandemic. ^{4, 26} Such reduction was probably due to the concern about COVID-19 infection in nursing homes and expanded eligibility of “homebound” for HHA services. ²⁷ In addition, there was a significant decrease in nursing home beds during the pandemic, which may have created access barriers to SNF services and also contributed to fewer SNF admissions. ²⁸ The extent of reductions in SNF admissions varied by individual race and ethnicity, and it was particularly large among duals, raising the concern of whether these residents received needed services after being discharged to home.

HHA may serve as an alternative to skilled nursing facilities (SNF) in providing post-acute support that aligns with the goals of patient services. Previous research has demonstrated that outcomes for individuals with ADRD receiving care from either SNFs or HHAs were comparable. ²⁹ However, despite the greater reduction in SNF use among racial and ethnic minoritized groups and duals, the substitution of SNF with HHA services was lower among these patients, particularly duals. Before the pandemic, duals with ADRD were more likely to utilize SNFs than non-duals. ¹² It is important to note that duals were also eligible to receive Medicaid-paid home- and community-based services (HCBS), offering a variety of medical and social services. During the pandemic, states received temporary federal funding to bolster their HCBS programs to assist older adults in meeting their care needs. ³⁰ Hence, one possible explanation for the observed decrease in substitution from Medicare SNF to HHA among duals could be that they might have availed themselves of services from HCBS programs and did not necessitate SNF or HHA services upon hospital discharge. On the other hand, it is also plausible that the available HHA services were unable to meet the demand among dual-eligible individuals. In fact, we found that the smaller substitution effect among duals was primarily due to the larger reduction in SNF utilization. Many duals have complex care needs and require extensive post-acute support. However, while the demand for HHAs has increased during the pandemic, the workforce shortage has been exacerbated, making it more challenging for duals to access appropriate services. ³¹ Furthermore, there may be disparities in timely access to HHA that particularly affect the dually eligible. Prior research has reported that a higher proportion of dual-eligible Medicare beneficiaries were found in the groups with incomplete HHA referrals within seven days of hospital discharge compared to those with complete HHA referrals (19.07% vs. 12.50%), suggesting duals might have delayed or were unable to access services from HHAs. ³² The lack of access to needed PAC services at home may lead to poorer health outcomes. Further research is needed to investigate the reasons behind the smaller substitution effect of SNF services with HHA services among these individuals with ADRD, as findings may provide important policy implications.

Access to high-quality HHAs is also important for facilitating smooth transitions after SNF discharge and improving outcomes. ^{33, 34} Research has already documented disparities in access to high-quality HHA services among racial and ethnic minoritized populations, as well as socioeconomically disadvantaged populations prior to the pandemic. ^{35, 36} Such disparities, if they persisted during the pandemic, could further lead to poor health outcomes among individuals with ADRD. Further studies may be needed to examine disparities in access to high-quality HHAs, the reasons behind the disparities, and their impact on health outcomes.

Limitations

Several limitations of our study should be noted. One is the potential misclassification of PAC destinations due to the possibility of some patients receiving PAC services beyond the 3-day window used in our analysis. Second, it is possible that we were not able to capture all factors that may affect the PAC destination, such as the severity of ADRD, the availability of paid and unpaid caregivers, and patients' or their families’ preferences. Nonetheless, we have controlled for a long list of individual health conditions, including comorbidities, prior healthcare utilization, and characteristics of their inpatient stay. Third, this study only focused on the FFS population, and the findings may not be generalized to Medicare Advantage enrollees. Lastly, this study only included non-Latinx White, non-Latinx Black and Latinx individuals. It will be important to understand the patterns of PAC discharge destinations among other race and ethnicity groups in future research.

CONCLUSION

This study showed that the impact of COVID-19 on PAC trajectories among Medicare beneficiaries with ADRD has persisted into 2021. SNF utilization remained lower than the pre-pandemic level, while HHA and home discharges increased. We observed changes in PAC destinations among different racial, ethnic, and socioeconomic groups, with more racial and ethnic minoritized groups and duals discharged to home or HHA during the pandemic. Moreover, non-duals were more likely to replace their SNF admission with HHA services, while duals were more likely to be discharged home without HHA. These findings highlight the effects of a public health emergency on PAC destinations and suggest the need for future studies to examine and understand the reasons for differences in HHA access and PAC outcomes. Understanding the implications of these shifts in PAC destinations for patient outcomes is also vital for improving the quality and equity of care among Medicare beneficiaries with ADRD in the post-pandemic era.

Key points

1. The pandemic led to a shift in PAC discharge destinations among community-dwelling older adults with Alzheimer’s disease and related dementia (ADRD), reducing skilled nursing facility (SNF) discharges and increasing home and home health agency (HHA) discharges.

2. The decrease in SNF discharges was more prominent among racial and ethnic minoritized groups and Medicare-Medicaid dual-eligible older adults with ADRD.

3. The reduction in SNF discharges among non-duals was mainly replaced by an increase in HHA use, but duals were more likely to be discharged to home without HHA, a trend consistent across all racial and ethnic groups in the ADRD population.

Why does this matter

Barriers to care access among diverse racial and ethnic groups, along with socioeconomically disadvantaged populations, have been a long-standing concern. Understanding the effects of a public health emergency on differences in PAC destinations among the ADRD populations identifies areas for future research and informs intervention efforts aimed at ensuring healthcare access and improving outcomes, regardless of individual race, ethnicity, and socioeconomic status.