Home Health Care after Skilled Nursing Facility Discharge Following Heart Failure Hospitalization

Abstract

Background/Objective

Heart failure (HF) readmission rates have plateaued despite scrutiny of hospital discharge practices. Many HF patients are discharged to skilled nursing facility (SNF) after hospitalization before returning home. Home health care (HHC) services received during the additional transition from SNF to home may affect readmission risk. Here, we examined whether receipt of HHC affects readmission risk during the transition from SNF to home following HF hospitalization.

Design

Retrospective cohort study.

Setting

Fee-for-service Medicare data, 2012–2015.

Participants

Beneficiaries 65 and older hospitalized with HF who were subsequently discharged to SNF then discharged home.

Measurements

The primary outcome was unplanned readmission within 30 days of discharge to home from SNF. We compared time to readmission between those with and without HHC services using a Cox model.

Results

Of 67,585 HF hospitalizations discharged to SNF and subsequently discharged home; 13,257 (19.6%) were discharged with HHC, 54,328 (80.4%) without. Patients discharged home from SNF with HHC had lower 30-day readmission rates than patients discharged without HHC (22.8% vs 24.5%, p<0.0001) and a longer time to readmission. In adjusted model, hazard for readmission was 0.91 (0.86 – 0.95) with receipt of HHC.

Conclusions

Recipients of HHC were less likely to be readmitted within 30 days versus those discharged home without HHC. This is unexpected, as patients discharged with HHC likely have more functional impairments. Since patients requiring a SNF stay after hospital discharge may have additional needs, they may particularly benefit from restorative therapy through HHC; however only about 20% received such services.

Introduction

Interventions to improve post-hospital outcomes in patients with heart failure (HF) have traditionally focused on hospital discharge.¹ Though skilled nursing facilities (SNFs) are often used to transition vulnerable patients from hospital to home, it is increasingly recognized that the additional transitions from hospital to SNF² and then from SNF to home^3–5 can affect readmission risk. As a result, in October 2018, the Centers for Medicare and Medicaid Services (CMS) rolled out the SNF Value-Based Purchasing (VBP) Program which offers incentive payments and penalties to SNFs paid under the SNF Prospective Payment System based on performance on 30-day readmission measures.⁶

Patients with HF may particularly benefit from increased attention to post-acute care transitions as HF readmission rates remain high⁷ despite many being discharged to SNF prior to returning home.⁸ Home health care (HHC) involves the provision of skilled services to patients for both acute and chronic care management and can be used as support both during the transition from hospital to home and from SNF to home. Early high intensity home care after HF hospitalization has previously been found to be associated with lower 30-day readmissions,^9,10 while prior work has demonstrated that HHC is associated with reduced rates of readmission during the SNF to home transition in a safety net cohort.¹¹

However, little is known about the utility of HHC in the transition from SNF to home after a HF hospitalization. Therefore, we examined the association between receipt of HHC after SNF discharge and readmission risk among a national cohort of Medicare patients discharged from SNF to home following HF hospitalization from 2012–2015.

Methods

Study sample and Data Sources

This was a retrospective cohort study of all Medicare fee-for-service beneficiaries 65 and older discharged from July 2012 to June 2015 with a principal discharge diagnosis of HF, identified by methods adopted by the Centers for Medicare and Medicaid Services (CMS).^12,13 Medicare Standard Analytic Files were used to identify admissions to acute care hospitals and skilled nursing facilities. These were merged with Medicare Denominator Files containing information on patient-level characteristics including date of birth, sex, and race.¹⁴ We excluded hospitalizations linked with SNF discharge dates more than 30 days after hospital discharge as the SNF VBP Program’s SNF 30-Day All-Cause Readmission Measure (SNFRM), measures the rate of all-cause, unplanned hospital readmissions for SNF residents within 30 days of discharge from a prior hospital stay.⁶ We also excluded admissions with less than one day in SNF and hospital admissions originating from SNF, as the latter represents long-term care patients.

Discharge with HHC may be confounded by quality of SNF care. To address this, we incorporated SNF-level characteristics via CMS’s SNF Quality Reporting Program from Nursing Home Compare. Nursing Home Compare is a website administered by CMS that allows comparison of quality of care at every Medicare-certified nursing home. The inaugural release of this data was published in October 2018 and consists of data from 2016–2018.

Primary Independent Variable of Interest:

The primary independent variable of interest was discharge from SNF to home with or without HHC.

Main Outcome Measure:

The primary outcome was days until any unplanned readmission following SNF to home discharge, censoring at 30 days; unplanned readmissions were defined using CMS’s HF readmission measure methodology.¹⁵

Covariates:

We used the SNFRM risk adjustment,¹⁶ including age, sex, whether the index hospitalization included intensive care unit (ICU) days, index hospital length of stay, and the number of acute care hospitalization days in the year prior to the index hospitalization. However, we did not use the SNFRM comorbidity variables as they are not condition specific; instead we used the 35 co-morbidities included in the Medicare HF readmission measure. Our model also accounted for additional patient-level variables: length of stay in SNF, Medicaid dual eligibility, race, and whether the patient had a SNF stay in the year prior to the index hospitalization.

We also accounted for SNF-level variables in our model: SNF ownership type, bed size of SNF, SNF overall rating, SNF quality measure rating, SNF staffing rating, physical therapist staffing hours per resident per day, adjusted total nurse staffing hours per resident per day, urban vs. rural setting, and geographic region. The SNF overall star rating incorporates health inspections, quality of resident care measures, and staffing data.¹⁷ The SNF quality measure rating incorporates information on physical and clinical measures for nursing home residents. The SNF staffing rating is based on registered nurse (RN) hours per resident per day and total staffing hours (RN, licensed practical nurse, nurse aide) per resident per day.¹⁸ Ratings range from 1–5, with 5 representing highest quality.

Statistical Analysis

We first compared those discharged home from SNF with and without HHC with descriptive statistics, testing for differences using χ² and t-tests. Table 1 also includes a comparison of Elixhauser co-morbidity scores between these groups. Next, we plotted time to unplanned readmission with Kaplan-Meier curves and compared these groups with a log-rank test. Finally, we compared time to unplanned readmission using a multivariable Cox proportional hazards model that adjusted for all covariates. Variance inflation factor (VIF) was calculated to estimate multicollinearity for each predictor. A predictor of VIF > 2.5 was considered as an indicative of serious collinearity. This was observed with the “SNF RN staffing rating” and the “adjusted total nurse staffing hours per resident per day” variables; thus “SNF RN staffing rating” was excluded from the model. Planned readmissions as defined by Horwitz et al.^12,19 and death were considered competing risks in this model and thus were censored. Patients were also censored at 30 days. This model included a random effect to account for correlation of patient outcome by SNF.²⁰ We found that the variance of this effect for outcome by hospital was very small (0.01) and therefore we did not account for hospital effects in our model. A sensitivity analysis in which indicator for missing data in the Cox proportional hazards model was also conducted. Analyses were conducted using SAS (SAS Institute), version 9.4. The need to obtain informed consent was waived by the Yale Human Investigation Committee and the NYU School of Medicine Institutional Review Board; both provided approval for the overall study.

Table 1:

Characteristics of Those Discharged Home from SNF with and without Home Health Care

Characteristic	No Home Health Care n=54328	Home Health Care n=13257	p-value
	Median (IQRa) or Mean ± SDb; or N (%)
Age	83.3 ± 7.9	83.3 ± 7.7	0.29
Male	21374 (39.3%)	5012 (37.8%)	0.001
Race
Other	1697 (3.1%)	337 (2.5%)	<.0001
White	48358 (89.0%)	11726 (88.5%)
Black	4273 (7.9%)	1194 (9.0%)
Length of stay in SNFc	18.0 (11.0 – 23.0)	15.0 (10.0 – 21.0)	<.0001
Elixhauser comorbidity score	22.0 (12.0 – 32.0)	22.0 (12.0 – 32.0)	0.32
Dual eligible	7497 (13.8%)	1854 (14.0%)	0.58
Hospitalization in the past year	36863 (67.9%)	8814 (66.5%)	0.002
SNF utilization in the past year	21530 (39.6%)	4857 (36.6%)	<.0001
ICUd stay during index hospitalization	20360 (37.5%)	4883 (36.8%)	0.17
Length of stay of index hospitalization	7.50 (3.76)	7.61 (4.46)	0.01
SNF ownership
Missing	1413 (2.6%)	1562 (11.8%)	<.0001
For profit	36474 (67.1%)	6439 (48.6%)
Government	1621 (3.0%)	517 (3.9%)
Non profit	14820 (27.3%)	4739 (35.7%)
SNF beds
Missing	1413 (2.6%)	1562 (11.8%)	<.0001
1–99	18221 (33.5%)	5797 (43.7%)
100–199	29265 (53.9%)	5010 (37.8%)
200–299	3988 (7.3%)	646 (4.9%)
300–399	959 (1.8%)	206 (1.6%)
400–499	217 (0.4%)	18 (0.1%)
500+	265 (0.5%)	18 (0.1%)
SNF Overall rating
Missing	1413 (2.6%)	1562 (11.8%)	<.0001
1	3770 (6.9%)	591 (4.5%)
2	9594 (17.7%)	1420 (10.7%)
3	7971 (14.7%)	1600 (12.1%)
4	12809 (23.6%)	2698 (20.4%)
5	18771 (34.6%)	5386 (40.6%)
Location
Missing	1509 (2.8%)	1628 (12.3%)	<.0001
Urban	47214 (86.9%)	10311 (77.8%)
Rural	5605 (10.3%)	1318 (9.9%)
Region
Missing	1416 (2.6%)	1562 (11.8%)	<.0001
New england	6085 (11.2%)	1045 (7.9%)
Middle atlantic	9492 (17.5%)	1482 (11.2%)
East north central	10693 (19.7%)	1844 (13.9%)
West north central	3549 (6.5%)	734 (5.5%)
South atlantic	10155 (18.7%)	2687 (20.3%)
East south central	2746 (5.1%)	1403 (10.6%)
West south central	3184 (5.9%)	1056 (8.0%)
Mountain	1937 (3.6%)	541 (4.1%)
Pacific	5071 (9.3%)	903 (6.8%)

^aIQR =Interquartile Range

^bStandard Deviation

^cSNF: Skilled Nursing Facility

^*Additional patient- and SNF-level characteristics listed in Supplementary Table S1

Results

Of 155,440 hospitalizations associated with SNF discharge dates documented in Medicare Part A, 67,585 HF hospitalizations were discharged to SNF and subsequently discharged home; 13,257 (19.6%) were discharged with HHC, 54,328 (80.4%) without. Figure 1 delineates the cohort selection. Patients discharged with HHC were more likely to be female, black and have a shorter SNF length of stay compared to those discharged without HHC. Several co-morbidities were different between groups as well, including vascular/circulatory disease, renal failure, acute coronary syndrome, major psychiatric disorders, liver/biliary disease, stroke, dementia, other psychiatric disorders, fibrosis of lung, and depression (Supplementary Table S1). Bivariate analysis also demonstrated that those discharged with HHC were more likely to be discharged from a SNF that was a non-profit, with a higher overall rating (Table 1), higher SNF staffing rating, higher SNF nurse staffing rating, but lower SNF quality measure rating. Patients discharged with HHC were also more likely to be discharged from a SNF with more physical therapist staffing hours per resident per day and higher adjusted total nurse staffing hours per resident per day (Supplementary Table S1).

Figure 1:

Flow chart for cohort selection

Among the overall study cohort, 16,333 (24.2%) were readmitted within 30 days of SNF discharge. Patients discharged home from SNF with HHC had lower crude 30-day readmission rates than patients discharged without HHC (22.8% [95% CI; 22.1–23.5%;] vs 24.5% [95% CI; 24.1–24.9%;], p<0.0001, Figure 2). Supplementary Figure S3 illustrates unadjusted time to readmission between these groups, and shows that patients discharged home from SNF with HHC have a longer time to readmission (p<0.0001). Median time to readmission for those readmitted within 30 days after discharge home from SNF with HHC was 11 days (IQR; 5–19) and 9 days (IQR; 3–18) for those discharged home without HHC (p<0.0001). Mortality without readmission within 30 days was also lower in those discharged with HHC compared to those discharged without HHC (3.1% [95% CI; 2.83–3.43] vs 4.1% [95% CI; 3.93–4.27], p<0.0001).

Figure 2:

Patients discharged home from SNF with HHC had lower crude 30-day readmission rates than patients discharged without HHC. FFS, fee-for-Service; SNF, skilled nursing facility.

After risk-adjustment, patients discharged home with HHC still had a lower hazard of 30-day readmission (HR 0.91 [95% CI; 0.86–0.95]. Table 2 outlines the hazard ratios for the SNF-level variables in the model. Supplementary Table S2 lists the hazard ratios for all the variables in the model. This finding that patients discharged home with HHC had a lower hazard of 30-day readmission was consistent in our sensitivity analysis, which included an indicator for missing data in the Cox proportional hazards model (HR 0.90 [95% CI; 0.87–0.94]).

Table 2:

Multivariable analysis examining effect of home care on readmission after discharge from SNF to home following HF hospitalization: selected SNF-level predictors

Characteristic	Hazard Ratio (95% CIa)	p-value
Home Health Care	0.91 (0.86 – 0.95)	<.0001
SNFb ownership
For profit	reference
Government	0.95 (0.86 – 1.05)	0.003
Non profit	0.93 (0.89 – 0.97)
SNF beds
1–99	reference
100–199	0.98 (0.94 – 1.02)	0.48
200–299	0.999 (0.93 – 1.08)
300–399	1.02 (0.89 – 1.17)
400–499	1.13 (0.86 – 1.50)
500+	1.22 (0.95 – 1.56)
SNF Overall Rating
1	reference
2	0.95 (0.87 – 1.03)	0.08
3	0.92 (0.84 – 0.99)
4	0.91 (0.84 – 0.99)
5	0.89 (0.82 – 0.98)
Location
Urban	reference
Rural	0.97 (0.92 – 1.03)	0.31
Region
New England	reference
Middle Atlantic	1.02 (0.95 – 1.10)	<.0001
East North Central	1.10 (1.03 – 1.18)
West North Central	1.06 (0.97 – 1.16)
South Atlantic	1.15 (1.07 – 1.23)
East South Central	1.12 (1.02 – 1.22)
West South Central	1.05 (0.96 – 1.15)
Mountain	0.90 (0.81 – 1.01)
Pacific	1.03 (0.95 – 1.12)

^aCI: Confidence Interval

^bSNF: Skilled Nursing Facility

^*Model also controlled for age, sex, race, length of stay in SNF, Medicaid dual-eligibility, number of acute care hospitalization days in year prior to index hospitalization, SNF utilization in the year prior to index hospitalization, whether the index hospitalization included intensive care unit days, index hospital length of stay, the 35 comorbidities included in the Medicare HF readmission measure, SNF size (number of beds), SNF Quality Measure rating, SNF Staffing rating, physical therapist staffing hours per resident per day, andadjusted total nurse staffing hours per resident per day. See Supplementary Table S2 for full model.

Discussion

Our study of all Medicare fee-for-service beneficiaries hospitalized for HF discharged to SNF and subsequently discharged home found that patients who received HHC on discharge were less likely to be readmitted within 30 days than those discharged home without HHC. This is unexpected, as patients discharged with HHC are likely to have more functional impairments²¹ and had shorter SNF stays. The latter is a particularly salient finding as patients with shorter SNF stays theoretically may be at higher risk for readmission²² because they are fewer days out from hospital discharge and consequently may be less recovered from their acute illness than those with longer SNF stays.

Under the Medicare SNF VBP program, nearly 11,000 SNFs will be penalized in Fiscal Year 2019 based on performance on 30-day readmission measures.^23,24 Despite our finding that patients discharged with HHC experienced decreased risk of readmission, only 20% were discharged with these services. This may represent “low hanging fruit” for SNFs searching for strategies to reduce readmissions and improve their performance in this program.

It is unclear why so few patients were discharged from SNF with HHC. Our bivariate analysis demonstrated that SNF-level characteristics appeared to be different in those discharged with HHC versus without (Table 1). However, given the disparate level of missing data (11.8–12.6% in the with HHC group compared to 2.6%−3.2% in the without HHC group), we are cautious in drawing generalizations. We speculate that SNFs in integrated health systems or with financial relationships with local HHC agencies may be more likely to discharge their patients with HHC. SNFs location in areas of high HHC agency density, such as those in urban settings, may also play a role.

In Medicare data, HHC classification includes skilled services from nurses, physical therapists, occupational therapists and/or speech-language pathologists. Regardless, we find benefit of HHC services in general after SNF discharge for HF patients which appears consistent with aforementioned work examining HHC during the SNF-to-home transition in other patient populations¹¹ and other research evaluating the utility of home based transitional care after HF hospital discharge,²⁵ such as home nursing visits within randomized controlled trial settings,⁹ and early intensive HHC combined with physician follow up within 7 days.¹⁰ On the other hand, another study demonstrated that patients discharged from HF hospitalization with HHC referral had a higher risk of 30-day readmission; however the authors acknowledge that this finding is likely explained by selection bias in that those receiving HHC are more likely to have more advanced illness.²⁶ Patients discharged to SNF after hospitalization not only have functional impairments but are likely to be medically complex⁸ and have social complexity requiring additional caregiver support--- all of which may benefit from HHC.

Prior work has indicated that SNF stays may also be protective against readmission.²⁷ Avoiding “post-hospital syndrome”²⁸ in patients with HF may require SNF care and then HHC after discharge home as they are vulnerable to swift decline if recommended treatment is not followed. The additional skilled support afforded by HHC could, for instance, increase adherence to medications,²⁹ dietary recommendations³⁰ and other discharge instructions. However, given the broad range of what HHC can entail, the specific mechanisms through which HHC may lower readmission risk need to be elucidated.²⁵ Further work should prospectively evaluate patients enrolled in HHC to examine patient level-factors such as disease acuity, cognitive impairment, physical functioning, and symptomology measured through structured assessments such as those found in MDS and OASIS data,³¹ in order to help providers identify which patients may most benefit from HHC and tailor the type of HHC required.

Strengths and Limitations

Our study is the first to characterize use of HHC in the transition from SNF to home after HF hospitalization using a national dataset. Most HF patients are over 65,³² making this group a high priority for Medicare. However, our study is limited by its observational design, which precludes causal inferences. Additionally, we did not have MDS or OASIS assessments and thus could not account for disease severity or patient function. As a result patients who select for HHC may differ systematically from those who do not. However, this is likely to be a conservative bias since HHC referral should be more common for those with greater functional needs and readmission risk. We also did not include Medicare Advantage patients in this study; however while the SNF VBP program applies to all Medicare fee-for-service patients, this is optional for other payors, including Medicare Advantage. Finally, while we accounted for SNF quality of care in our study, this data is from 2016–2018 (the only data available), and thus may not exactly reflect SNF quality of care from 2012–2015.

Conclusion

Use of SNF care has been increasing for the past decade,³³ placing an increased number of patients at risk of readmission after discharge to home from SNF. Although we found that following a HF hospitalization, patients discharged from SNF to home with HHC are at decreased risk of readmission, only about 20% of discharges received such services. Increasing use of HHC services after SNF discharge may help reduce readmission risk. Since patients requiring a SNF stay after hospital discharge are likely to be functionally impaired with caregiving needs, this patient population overall may especially need restorative therapy through HHC.

Supplementary Material

Supplementary Figure S3

Supplementary Figure S3: Kaplan-Meier curve depicting time to readmission after discharge from SNF to home following HF hospitalization.

Abbreviations used:

HF

heart failure

SNF

skilled nursing facility

VBP

Value-Based Purchasing

SNFRM

Skilled Nursing Facility Readmission Measure

HHC

home health care

RN

registered nurse

CMS

Centers for Medicare and Medicaid Services

VIF

Variance inflation factor