Abstract
The evaluation of social support within hospital at home (HaH) programs has been limited. We performed a secondary analysis of a prospective cohort evaluation of 295 participants receiving HaH care and 212 patients undergoing traditional hospitalization from November of 2014 to August of 2017. We examined the confounding and moderating effects of instrumental and informational social support upon length of stay (LOS) and 30-day re-hospitalization, emergency department (ED) visit, and skilled nursing facility (SNF) admission. Instrumental social support attenuated the effects of HaH upon any ED visit (base model: OR 0.61, p=0.037; controlling for social support: OR 0.71, p=0.15). The association of HaH with other outcomes remained unchanged. Interactions between HaH and informational or instrumental social support for all outcomes were not significant. Lack of high-levels of social support had little effect on the positive outcomes of HaH care, suggesting similar benefits of HaH services for patients with lower levels of social support.
INTRODUCTION
Delivering acute hospital-level care at home has been demonstrated to be a safe, effective, and cost-effective alternative to traditional hospitalization (Caplan et al., 2012; Cryer et al., 2012; Federman et al., 2018). However, the decision to provide or receive hospital at home (HaH) may be influenced by the availability and degree of social support, such as family and other caregivers who may be asked to, or feel they should, assist with care during the HaH care provision. While randomized control trials (RCTs) and intention-to-treat approaches can control for this potential selection, most US-based evaluations of HaH were not RCTs (Caplan et al., 2012; Levine et al., 2020). The paucity of social support measurement and adjustment may confound findings (Liao et al., 2018).
Social support may function synergistically with, and enhance the effects of, health service delivery. During times of crisis, such as an acute illness, patients leverage social support networks to navigate treatment and assist in recovery (Smith et al., 2018; Villain et al., 2017). The lack of social support has been implicated in higher rates of readmissions after traditional hospitalization for common chronic conditions, including heart failure and pneumonia (Calvillo King et al., 2013; Tsuchihashi-Makaya et al., 2009). When delivering acute-level care in the home, matching the high level of instrumental social support (i.e., physical assistance with transfers and other activities of daily living) and informational social support (i.e., navigation of treatment and care decisions) of traditional hospitalization may be important in achieving favorable outcomes. Prior HaH programs have required caregivers within the home to provide this support (Aimonino Ricauda et al., 2008; Tibaldi et al., 2009). More recent programs have not had this requirement(Federman et al., 2018; Levine et al., 2020). Studies have demonstrated that caregiver satisfaction is greater, burden is unchanged, and stress is lower with HaH services (Caplan et al., 2012). However, the degree to which these social support networks are assisting with disease management and recovery may be an important component for the effectiveness of HaH.
This study first aimed to determine whether instrumental or informational social support serve as confounders and alter the measured effectiveness of HaH care. Second, we tested if high levels of instrumental or informational social support are effect modifiers of HaH outcomes.
METHODS
Design and Study Population.
We conducted a secondary analysis of data from a prospective evaluation of a HaH program of a large hospital system in New York, NY (Federman et al., 2018). The evaluation included patients receiving HaH and patients receiving traditional in-hospital care. The HaH program evaluation was established through a Center for Medicare and Medicaid Innovation contract in 2014. From November of 2014 to August 2017, patients, who presented to two emergency departments (EDs) and required admission for their acute illness, were screened for HaH eligibility. Eligibility criteria included adults 18 years of age and older requiring acute hospital admission for one of 19 acute medical conditions representing 65 diagnostic-related groups, and having fee-for-service Medicare or coverage from one private insurer (Medicare Advantage and Managed Medicaid) that contracted with HaH services. Patients were ineligible if deemed clinically unstable, needed cardiac or intensive care monitoring, lived in an unsafe home environment, or lived outside Manhattan. Adequate social support was not required. If patients were able to access the telephone, obtain meals, and use the bathroom, a caregiver within the home was not required. Patients who accepted the program were transported home to receive services. Patients, who declined HaH and agreed to be followed (9.6%) or who presented to the ED with HaH-qualifying diagnoses between 4PM and 8AM when HaH providers were not available to admit new patients (90.4%), served as the control group of patients receiving traditional in-hospital care.
This study was initiated for internal program evaluation and reporting to the Centers for Medicare and Medicaid Services. For this initial group of patients receiving HaH (50.9%), consent was not required, and baseline data, including social support measures, were not collected. For analyses involving the subsequent HaH group and the entire control group, we received approval from the Institutional Review Board (IRB). After obtaining consent, patient data from this group were collected via surveys and abstraction of administrative records at baseline, during study period, and at 30 days after admission. After completion of the Center for Medicare and Medicaid Innovation award, we received approval from IRB to retrospectively abstract and use data of the initial group of HaH patients.
Hospital at Home program.
Hospital-at-Home included a care bundle of home-based acute care and transitional services in the 30-day post-acute period. Hospital-at-Home was provided by a team of physicians, physician assistants, nurse practitioners, registered nurses, social workers, and physical and occupational therapists who delivered hospital services in the patient’s home. These included daily physical examination and assessment, diagnostic studies (e.g., electrocardiograms, plain films, laboratory studies, and ultrasounds), and treatment (e.g., intravenous fluids and medications, oxygen, wound care, and durable medical equipment). Providers were available by phone, or if more urgent, by video or in-person 24-hours per day through community paramedics. When acute illness resolved, patients were discharged to the 30-day post-acute period.
For 30 days after discharge, nurses and social workers provided transitional services, including self-management support and coordination of care. If needed or requested, physicians and nurse practitioners were available to provide home-based follow-up care, and HaH social workers coordinated with a certified home health agency to supply patients with a home health aide or other qualifying services.
Outcome variables.
Primary outcomes were hospital length of stay (LOS), hospital readmission or ED visit within 30-days post-discharge, and referral to a skilled nursing facility (SNF). We determined LOS from the date of admission documented by the admitting physician to the date of post-acute care initiation for HaH patients or in-hospital discharge for control patients. Readmission and referral data were extracted from charts and from surveys during the follow-up period.
Measures of Social Support (SS).
The baseline survey assessed patients’ perceived social support, using the Patient-Reported Outcomes Measurement Informational System (PROMIS) Short Form V2.0 4-item informational and instrumental support measures. Informational support encompasses the ability to obtain advice and information for important things in life, during a crisis, or to problem solve. Instrumental support includes physical assistance when sick, including transportation to the doctor, help with chores, and running errands. Each item was scored on a 5-point scale with total scores ranging from 4 to 20 for each domain and higher scores indicating greater support. We defined high social support if the composite score was ≥1 standard deviation from the mean score, normalized to the general population (PROMIS, 2015).
Using this cutoff, scores of 19 and 20 for informational and scores of 20 for instrumental were defined as high social support.
Covariate Variables.
Data were abstracted from charts for age, sex, insurance, and admission diagnosis. Data on race and ethnicity, education, impairments of activities of daily living, self-reported health status, and presence of a caregiver (defined as a person assisting with care, medical appointments, or medical decisions) were collected by interview.
Statistical approach.
Data were missing for race/ethnicity (12%), education (20%), functional status (23%), self-reported health (16%), caregiver (24%), and informational (40%) and instrumental (38%) social support. Missing data occurred for 2 reasons: patients failed to complete the survey questions or were among the initial group of HaH patients who did not receive the survey. We assumed that these data were missing at random and any systematic differences between the missing and observed values could be explained by the differences in observed data. To handle missing data, we used multiple imputation (SAS Proc MI). We used fully conditional specified linear and logistic regression to impute missing continuous and binary or categorical covariates, respectively. Covariates included in the imputation were age, sex, race and ethnicity, education, self-reported health status, insurance type, admission diagnosis, caregiver status, paid caregiver status, and the four outcomes. We created 30 imputed datasets and modeled propensity scores and outcomes for each dataset, then combined the results to generate overall model estimates and standard errors (Leyrat et al., 2019).
We specified statistical models to identify potential confounding and moderation of informational and instrumental social support (Figure). To investigate confounding, we examined the change in estimate before and after adjustments for continuous measures of social support and used the recommended cutoff of 10% (P. H. Lee, 2014). We specified four models for each of the four outcomes: Model 1, no adjustment for social support; Model 2, adjustment for informational social support; Model 3, adjustment for instrumental social support; and Model 4, adjustment for both informational and instrumental social support. To test for moderation, we added social support domain by site of care interaction terms to Model 4. Consistent with reporting guidelines (VanderWeele & Knol, 2014), we calculated and presented the differences in LOS and ORs of 30-day post-acute outcomes for each dichotomous stratum of social support (high vs. lower) and treatment (HaH vs. control) with lower social support and traditional in-hospital care as the reference group. Finally, we directly tested and reported estimates for interaction on multiplicative scale.
Figure: Analysis Diagram testing confounding and moderating effects of Social Support within Hospital-at-Home.
*Information and Instrumental social support were assessed through baseline survey with four questions and 5 potential responses to each question (1-Never, 2-Rarely, 3-Sometimes, 4-Usually, 5-Always) with highest score of 20. Analysis test if high information and instrumental social support confounded# and moderated^ the association of HaH with lower length of stay (LOS) and lower post-acute ED visit, hospitalization, and SNF referral compared to traditional hospital care.
To limit potential selection bias due to nonrandomized assignment to HaH, all models applied inverse probability of treatment weights (IPTW) (Abadie & Imbens, 2008; Garrido et al., 2014). These IPTWs were calculated separately for the four models. All IPTWs included additional covariates of age, sex, race and ethnicity, education, self-reported health status, insurance type, admission diagnosis, caregiver status, and paid caregiver status. All probabilities demonstrated balance of covariates across in-hospital and HaH groups with mean standardized differences across 30 imputed datasets less than 15% for all covariates (Abadie & Imbens, 2008; Garrido et al., 2014). All analyses were performed in SAS version 9.4 (SAS Institute, Cary, NC). Significance for the primary analysis was set as p< .05. For secondary analyses of moderation, we applied a Bonferroni adjustment for the testing of 8 models (4 outcomes modeled separately with instrumental or informational social support). Significance for these analyses was set at p< .006.
Sensitivity Analyses. Given the disproportionate and substantial missing social support data from the HaH due to differences in baseline data collection, we performed sensitivity analyses restricted to the sample of patients offered the comprehensive baseline interview, which included social support measures.
RESULTS
Population characteristics and missing data.
A total of 460 patients were screened, of whom 295 were eligible and agreed to receive HaH care. Another 479 patients who declined HaH or presented outside HaH admitting hours were approached, of whom 212 consented to data collection and served as the in-hospital control cohort. The study population was predominantly ≥75 years of age (57%), female (79%), had high school or higher education (79%), and had fee-for-service Medicare insurance, with or without Medicaid (61%) (Table 1). Data on informational and instrumental social support was disproportionately missing for HaH patients (N missing [%] for informational social support: HaH, 176 [59.7%] vs. control, 28 [13.2%]; instrumental social support: HaH, 171 [58.0%] vs. control, 19 [9.0%]) (Supplemental Table S1). Complete data were available for 86 HaH (29.2%) patients and 130 controls (61.3%).
Table 1:
Characteristics of patients with high and lower instrumental and informational social support
| All | Informational SS | Instrumental SS | ||||
|---|---|---|---|---|---|---|
|
| ||||||
| Characteristic | N Imputed | Lower Score ≤ 18 | High Score ≥ 19 | Lower Score ≤19 | High Score =20 | |
|
| ||||||
| N = | 507 | 277 | 230 | 335 | 168 | |
|
| ||||||
| Hospital-at-Home (%) | 0 | 58.2 | 62.7 | 52.7 | 60.6 | 53.4 |
|
| ||||||
| Age, mean (Std. Err) | 0 | 74.7 (0.7) | 74.4 (1.0) | 74.9 (1.1) | 74.5 | 75.0 |
| Age Categories (%) | 0 | |||||
| < 65 years | 21.1 | 23.9 | 17.7 | 22.8 | 17.7 | |
| 65–74 years | 22.3 | 19.7 | 25.4 | 21.1 | 24.7 | |
| 75–84 years | 26.8 | 25.6 | 28.4 | 26.2 | 28.1 | |
| 85 years or older | 29.8 | 30.8 | 28.5 | 30.0 | 29.4 | |
|
| ||||||
| Gender, male (%) | 0 | 31.4 | 30.2 | 32.7 | 31.9 | 30.3 |
|
| ||||||
| Race/Ethnicity (%) | 62 | |||||
| White/Other | 42.9 | 42.1 | 43.9 | 44.4 | 40.0 | |
| Black | 25.6 | 25.4 | 25.8 | 24.3 | 28.0 | |
| Hispanic | 31.9 | 32.5 | 31.2 | 32.0 | 31.8 | |
|
| ||||||
| Education (%) | 101 | |||||
| Elementary School | 20.4 | 20.4 | 20.5 | 21.2 | 19.2 | |
| Any High School | 31.2 | 31.5 | 30.8 | 31.4 | 30.8 | |
| Any College | 28.0 | 29.0 | 26.9 | 27.0 | 30.1 | |
| Any post-graduate | 20.1 | 19.0 | 21.4 | 20.3 | 19.8 | |
|
| ||||||
| Insurance (%) | 0 | |||||
| Medicare, FFS | 40.6 | 40.8 | 40.4 | 41.6 | 38.7 | |
| Medicare, Private | 19.5 | 21.0 | 17.7 | 19.5 | 19.6 | |
| Any Medicaid | 39.8 | 38.2 | 41.8 | 38.9 | 41.7 | |
|
| ||||||
| Any ADL impairment (%) | 114 | 65.2 | 62.4 | 68.8 | 62.0 | 71.6 |
|
| ||||||
| Self-Rated Health, Poor (%) | 83 | 63.5 | 64.9 | 61.9 | 64.7 | 61.1 |
|
| ||||||
| Any caregiver (%) | 121 | 76.0 | *68.7 | *84.9 | *67.4 | *93.2 |
|
| ||||||
| Paid Caregiver (%) | 131 | 31.6 | 29.5 | 34.2 | 29.5 | 36.0 |
|
| ||||||
| Social support | ||||||
| Informational, mean (St Err) | 204 | 16.4 (0.3) | *15.2 (0.3) | *18.8 (0.2) | ||
| Informational, High, ≥19 (%) | 0.45 | *28.1 | *78.7 | |||
|
| ||||||
| Instrumental, mean (Std Er) | 190 | 16.1 (0.3) | *14.6(0.4) | *17.9(0.3) | ||
| Instrumental, High, =20 (%) | 0.33 | *13.0 | *58.5 | |||
|
| ||||||
| LOS, days, mean (Std Er) | 0 | 4.4 (0.15) | 4.4 (0.21) | 4.4 (0.24) | 4.3 (0.17) | 4.6 (0.29) |
| 30-day post-acute (%) | ||||||
| Re-hospitalization | 0 | 8.4 | 8.3 | 8.5 | 9.4 | 6.4 |
| ED visit | 0 | 11.6 | 10.4 | 13.0 | 11.5 | 11.8 |
| SNF Referral | 0 | 5.3 | 4.4 | 6.5 | 6.2 | 3.7 |
Social Support (SS); Fee-for-service (FFS); Activities of daily living (ADL); Length of stay (LOS); Emergency Department (ED); Skilled Nursing Facility (SNF)
p<0.05
p<0.001
After imputation of missing data, characteristics among patients with high and lower informational and instrumental social support were largely similar (Table 1). However, patients with high levels of informational and instrumental social support were more likely to have an informal caregiver (Informational social support, 85% vs. 69%, p<0.001; Instrumental social support, 93% vs. 67%, p<0.001).
Confounding by Social Support (SS).
When not adjusting for social support (Model 1), HaH was associated with shorter LOS (−1.89 days; 95% CI −2.48, −1.29; p<0.001) and lower odds of 30-day re-hospitalization (OR 0.63; 95% CI 0.42, 0.96; p=0.03), ED visit (OR 0.61; 95% CI 0.39, 0.97; p=0.37), and referral to SNF (OR 0.14; 95% CI 0.07, 0.28; p<0.0001) (Table 2). Controlling for informational social support (Model 2) did not change the ß-coefficients for the effects of HaH on these outcomes by >10%. Adding instrumental social support to the models did not demonstrate confounding, with the exception of ED visits. When controlling for instrumental social support (Models 3), the association between HaH and ED visits was no longer statistically significant (% change in ß coefficient, 29%; OR 0.71, 95% CI 0.44, 1.14; p=0.15). This was similarly observed in Model 4 when controlling for both informational and instrumental social support (% change in ß coefficient, 38%; OR 0.74, 95% CI 0.46, 1.18; p=0.20).
Table 2:
Effects of the HaH treatment upon length of stay and 30-day post-acute outcomes when adjusting for informational and instrumental social support
| 30-day Post-Acute Period | ||||||||
|---|---|---|---|---|---|---|---|---|
|
|
||||||||
| Length of Stay | Re-hospitalization | ED visit | SNF Referral | |||||
|
| ||||||||
| Days (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | |
|
| ||||||||
| Adjusting for SS | ||||||||
| Model 1: No SS | −1.89 (−2.48, −1.29) | <0.001 | 0.63 (0.42, 0.96) | 0.031 | 0.61 (0.39, 0.97) | 0.037 | 0.14 (0.07, 0.28) | <0.001 |
| Model 2: Info. SS | −1.89 (−2.48, −1.30) | <0.001 | 0.63 (0.41, 0.96) | 0.033 | 0.63 (0.40, 1.00) | 0.049 | 0.13 (0.06, 0.27) | <0.001 |
| Model 3: Instrum. SS | −1.93 (−2.53, −1.32) | <0.001 | 0.63 (0.41, 0.97) | 0.034 | 0.71 (0.44, 1.14) | 0.151 | 0.15 (0.07, 0.32) | <0.001 |
| Model 4: Info. & Instrum. SS | −1.95 (−2.55, −1.34) | <0.001 | 0.62 (0.40, 0.95) | 0.030 | 0.74 (0.46, 1.18) | 0.203 | 0.16 (0.07, 0.33) | <0.001 |
Linear regression used for length of stay and logistic regression used for 30-day post-acute outcomes. Models used applied inverse probability of treatment weights to adjust for potential confounding of social support (SS). Model 1 did not include social support measures in adjustments. Model 2 included Informational (Info.) SS; Model 3, Instrumental (Instrum.) SS; Model 4, Informational and Instrumental SS. All models include all patients (N=507) after multiple imputation for missing values. Social support were included as continuous measures. Hospital-at-Home(HaH); Emergency Department (ED); Skilled Nursing Facility (SNF)
Moderation by Social Support (SS).
We examined outcomes across stratified groups of high and lower social support and HaH and in-hospital care. After adjusting for multiple comparisons (α=0.006), we observed statistically significant associations with reduced LOS and referral to SNF only when comparing groups with and without HaH (Table 3). Statistical interactions between site of care (HaH vs. in-hospital care) and levels of informational or instrumental social support were not statistically significant for any of the four outcomes (Table 3).
Table 3:
Evaluation of the interaction between Social Support (SS) and HaH on outcomes of care
|
|
||||
| Informational SS | Instrumental SS | |||
|
| ||||
| Length of Stay | Days (95% CI) | P value | Days (95% CI) | P value |
| Hospital, SS lower | Referent (0.0) | Referent (0.0) | ||
| Hospital, SS high | 0.51 (−0.37, 1.40) | 0.255 | 1.08 (0.16, 2.01) | 0.022 |
| HaH, SS lower | *−1.39 (−2.24,−0.54) | 0.001 | *−1.40 (−2.13, −0.68) | <0.001 |
| HaH, SS high | *−1.90 (−2.82,−0.97) | <0.001 | *−1.71 (−2.67, −0.76) | <0.001 |
| Interaction | Coef. (95% CI) | Coef. (95% CI) | ||
| SS * HaH | −1.02 (−2.36, 0.32) | 0.137 | −1.39 (−2.77, −0.01) | 0.048 |
|
| ||||
| 30-day post-acute period | ||||
|
| ||||
| Re-hospitalization | OR (95% CI) | P value | OR (95% CI) | P value |
| Hospital, SS lower | Referent (1.0) | Referent (1.0) | ||
| Hospital, SS high | 1.64 (0.87, 3.11) | 0.127 | 1.22 (0.69, 2.14) | 0.492 |
| HaH, SS lower | 0.92 (0.47, 1.77) | 0.796 | 0.78 (0.46, 1.31) | 0.343 |
| HaH, SS high | 0.66 (0.29, 1.53) | 0.331 | 0.43 (0.18, 1.06) | 0.067 |
| Interaction | Coef. (95% CI) | Coef. (95% CI) | ||
| SS * HaH | −0.82 (−1.96, 0.32) | 0.155 | −0.78 (−1.88, 0.31) | 0.16 |
| ED visit | ||||
| Hospital, SS lower | Referent (1.0) | Referent (1.0) | ||
| Hospital, SS high | 0.94 (0.50, 1.77) | 0.857 | 0.92 (0.49, 1.73) | 0.793 |
| HaH, SS lower | 0.51 (0.25, 1.01) | 0.052 | 0.75 (0.43, 1.29) | 0.295 |
| HaH, SS high | 0.75 (0.39, 1.46) | 0.399 | 0.35 (0.14, 0.85) | 0.021 |
| Interaction | Coef. (95% CI) | Coef. (95% CI) | ||
| SS * HaH | 0.45 (−0.55, 1.46) | 0.377 | −0.67 (−1.80, 0.46) | 0.244 |
| Referral to SNF | ||||
| Hospital, SS lower | Referent (1.0) | Referent (1.0) | ||
| Hospital, SS high | 1.90 (1.01, 3.57) | 0.045 | 0.47 (0.23, 0.97) | 0.040 |
| HaH, SS lower | *0.21 (0.07, 0.65) | 0.006 | *0.11 (0.05, 0.28) | <0.001 |
| HaH, SS high | 0.12 (0.03, 0.58) | 0.008 | *0.08 (0.02, 0.36) | 0.001 |
| Interaction | Coef. (95% CI) | Coef. (95% CI) | ||
| SS * HaH | −1.20 (−3.24, 0.85) | 0.250 | 0.36 (−1.57, 2.28) | 0.72 |
Linear regression used for length of stay and logistic regression for binary post-acute outcomes. Models adjusted with inverse probability of treatment weighting. Subgroups report the marginal effects for designated population in reference to patients who underwent traditional hospitalization (hospital) and reported lower informational or instrumental social support (SS). Hospital-at-Home (HaH); Emergency Department (ED); Skilled Nursing Facility (SNF); Coefficient (Coef.)
Statistically significant after Bonferroni adjustment, p<0.006
Sensitivity Analyses.
When we isolated the analyses to the subgroup of patients who received the baseline survey (Supplemental Table S2–S3), HaH continued to be significantly associated with shorter LOS (−2.27 days; 95% CI −2.93, −1.62; p<0.001) and lower odds of 30-day rehospitalization (OR 0.51; 95% CI 0.31, 0.86; p=0.012) and referral to SNF (OR 0.14; 95% CI 0.05, 0.34; p<0.001). The association with any ED visit was not observed (OR 0.88; 95% CI 0.54, 1.45; p=0.624).
Informational and instrumental social support did demonstrate confounding for the association of HaH with LOS and re-hospitalization (Supplemental Table S3). However, adjusting for instrumental social support changed the ß-coefficients for the association of HaH with referral to SNF by 15% (OR 0.71; 95% CI 0.44, 1.14; p=0.15). Adjusting for informational and instrumental social support changed the ß-coefficients for the association of HaH with any ED visit by 15% (OR 0.87; 95% CI 0.52, 1.46, p=0.59) and 261% (OR 1.22, 95% CI 0.73, 2.06; p=0.45), respectively; however, these associations were not statistically significant.
DISCUSSION
This study addressed the questions of whether the positive effects of HaH may be due to selection bias of HaH for patients with higher levels social support. Further, we tested if higher levels of social support enhance, or moderate, the positive effects of HaH. We observed that instrumental social support confounds the association of HaH with ED visit in the 30-day post-acute period; however, neither informational nor instrumental social support accounted for the strong associations of HaH with other clinical outcomes, including lower LOS, readmission, and SNF referral. Further, we observed that different levels of social were not associated with differences in the outcomes.
After imputation, patients participating in the HaH program had lower levels of informational and instrumental social support than patients in the comparison group. Despite prior evidence in studies of traditional hospital care suggesting that lower social support confers worse clinic outcomes (Calvillo King et al., 2013; Northcott et al., 2016), patients participating in HaH experienced improved post-acute outcomes. This suggests that HaH and its 30-day post-acute period of transitional care has the potential to effectively serve patients with lower levels of social support.
Hospital-at-Home included an experienced, multidisciplinary team able to recognize and fill gaps of social support during the acute episode and the 30 days of post-acute, homebased transitional care services. During the acute episode, nurses, providers and social workers with significant home-based care experience were able to fill informational gaps about treatment expectations, assess the home environment, and arrange needed instrumental assistance, such as a home health aide.
The HaH nurses and social workers provided ongoing self-management support and care coordination to all patients, filing gaps of outpatient care and assisting with transition in 30-day post-acute period. These services included patient education and direct communication with primary care providers. In addition, HaH delivered post-acute, in-home primary care and referred patients to certified home nursing agencies to provide added home support, potentially circumventing deficits in social support (Federman et al., 2018). In the post-acute period, patients rely upon social networks to support adherence to treatment and assist with physical recovery. Transitional care services that involve multidisciplinary teams and provide care coordination in the early post-acute period reduce recurrence of readmission, mortality, and costs (K. K. Lee et al., 2016; Wiest et al., 2019). Further, primary care outpatient visits occurring within 7 days of hospital discharge reduce hospital readmissions (K. K. Lee et al., 2016; Wiest et al., 2019). However, in the general Medicare population, few receive these transitional services (Bindman & Cox, 2018). Patients, who lack the level of informational social support to navigate new treatments or instrumental social support to facilitate transportation to the clinic, may be particularly vulnerable.
In addition, HaH’s delivery of transitional care services in the home may be particularly valuable in recognizing and mitigating social support deficits. In a meta-analysis of non-HaH transitional care services, home visits within days of hospital discharge conferred improved clinical outcomes (Verhaegh et al., 2014). Further, a study among Medicaid recipients identified that home visits compared to other nurse-led transitional support was associated with reduction in admissions and total costs, particularly for higher-risk medical patients (Jackson et al., 2016). Our study identified that patients with lower social support in HaH compared to inpatient care experienced lower referral to SNF; however, demonstrated no significant difference in readmission or ED visit in post-acute period. While findings may have been limited by statistical power, we also could not account for transitional services, including home visits, among the control population.
Informational and instrumental social support may function differently for those who undergo hospitalization for acute illness. First, only instrumental social support demonstrated any confounding effects. Further, patients who had inpatient hospitalization and high levels of informational social support were more likely to be referred to SNF in the post-acute period, while patients with high compared to lower instrumental social support were less likely. These associations did not reach statistical significance after adjustment for multiple comparisons and may have been limited by sample size and statistical power. Considering these potentially diverging associations, further investigation into the independent roles of informational and instrumental social support in clinically important outcomes in the post-acute period is warranted.
Limitations.
There were several potential limitations in this analysis. First, there was considerable missingness among self-reported social support measures, disproportionately higher among patients receiving the intervention because a large portion of HaH patients did not receive the social support survey. Multiple imputation aimed to overcome biases of missing data. We further applied sensitivity analyses limited to patients who received the social support survey, producing similar results. Second, due to lack of randomization, factors associated with receiving HaH and with clinical outcomes may have confounded our findings. We applied IPW and controlled for potential confounders; however, endogeneity and unmeasured confounding may persist. Third, measures of social support that we used were self-reported and subject to reporting bias; however, both measures have been validated (Patient-Reported Outcomes Measurement Information System (PROMIS), 2015).
Conclusions and Implications.
Controlling for informational and instrumental social support had little effect on the association of HaH with clinical outcomes, indicating that HaH combined with a 30-day post-acute period of transitional care is effective even for patients lacking high levels of social support.
Supplementary Material
Suggested callouts:
The lack of social support has been implicated in higher rates of readmissions after traditional hospitalization for common chronic conditions, including heart failure and pneumonia
For 30 days after discharge, nurses and social workers provided transitional services, including self-management support and coordination of care.
The study population was predominantly ≥75 years of age (57%), female (79%), had high school or higher education (79%), and had fee-for-service Medicare insurance, with or without Medicaid (61%)
Hospital-at-Home included an experienced, multidisciplinary team able to recognize and fill gaps of social support during the acute episode and the 30 days of post-acute, homebased transitional care services.
Acknowledgements:
We especially thank the patience, advice, and consultation of Hung-Mo Lin, Jimmy Akrivos, Liangyuan Hu, and Lihua Li from the Department of Population Health Science at the Icahn School of Medicine of Mount Sinai. They were all helpful in determining the best approach for imputation and model specification for data analyses.
Conflicts of Interest and Source of Funding. Research from this publication was supported by the US Department of Health and Human Services, Centers for Medicare & Medicaid Services (1C1CMS331334–01-00), the National Institute on Aging, Claude D. Pepper Older Americans Independence Center (TS, 3P30AG028741), and The John A. Hartford Foundation. The HaH clinical project described was supported by grant (1C1CMS331334) from the US Department of Health and Human Services, Centers for Medicare & Medicaid Services. Dr. Augustine was supported by the Empire Clinical Research Investigator Program awarded through the Department of Medicine at the Icahn School of Medicine at Mount Sinai.
Dr. Leff is a consultant for Medically Home and DispatchHealth. Drs. Augustine, DeCherrie, Federman, and Siu are either part- or full-time employees of the Icahn School of Medicine, which has an ownership interest in a joint venture with Contessa Health, a venture that manages acute care services provided to people in their homes through prospective bundled payment arrangements. These persons have no personal financial interest in the joint venture. The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Contributor Information
Matthew R. Augustine, Division of General Internal Medicine, Department of Geriatrics and Palliative Medicine, Icahn School of Medicine at Mount Sinai, New York, NY; Geriatric Research Education and Clinical Center, James J Peters VA Medical Center, Bronx, NY.
Albert L. Siu, Department of Geriatrics and Palliative Medicine, Icahn School of Medicine at Mount Sinai, New York, NY; Geriatric Research Education and Clinical Center, James J Peters VA Medical Center, Bronx, NY.
Kenneth S. Boockvar, Department of Geriatrics and Palliative Medicine, Icahn School of Medicine at Mount Sinai, New York, NY; Geriatric Research Education and Clinical Center, James J Peters VA Medical Center, Bronx, NY.
Linda V. DeCherrie, Division of General Internal Medicine, Department of Medicine, Icahn School of Medicine at Mount Sinai, New York, NY; Department of Geriatrics and Palliative Medicine, Icahn School of Medicine at Mount Sinai, New York, NY.
Bruce A. Leff, Division of Geriatrics, Department of Medicine, School of Medicine, Johns Hopkins University, Baltimore, MD.
Alex D. Federman, Division of General Internal Medicine, Department of Geriatrics and Palliative Medicine, Icahn School of Medicine at Mount Sinai, New York, NY; Department of Geriatrics and Palliative Medicine, Icahn School of Medicine at Mount Sinai, New York, NY.
References
- Abadie A, & Imbens GW (2008). On the Failure of the Bootstrap for Matching Estimators. Econometrica, 76(6), 1537–1557. [Google Scholar]
- Aimonino Ricauda N, Tibaldi V, Leff B, Scarafiotti C, Marinello R, Zanocchi M, & Molaschi M (2008). Substitutive Hospital at Home Versus Inpatient Care for Elderly Patients with Exacerbations of Chronic Obstructive Pulmonary Disease: A Prospective Randomized, Controlled Trial. Journal of the American Geriatrics Society, 56(3), 493–500. [DOI] [PubMed] [Google Scholar]
- Bindman AB, & Cox DF (2018). Changes in Health Care Costs and Mortality Associated With Transitional Care Management Services After a Discharge Among Medicare Beneficiaries. JAMA Internal Medicine, 178(9), 1165–1171. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Calvillo King L, Arnold D, Eubank KJ, Lo M, Yunyongying P, Stieglitz H, & Halm EA (2013). Impact of social factors on risk of readmission or mortality in pneumonia and heart failure: systematic review. Journal of general internal medicine, 28(2), 269–282. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Caplan GA, Sulaiman NS, Mangin DA, Aimonino Ricauda N, Wilson AD, & Barclay L (2012). A meta-analysis of “hospital in the home”. The Medical journal of Australia, 197(9), 512–519. [DOI] [PubMed] [Google Scholar]
- Cryer L, Shannon SB, Van Amsterdam M, & Leff B (2012). Costs for “hospital at home” patients were 19 percent lower, with equal or better outcomes compared to similar inpatients. Health Affairs, 31(6), 1237–1243. [DOI] [PubMed] [Google Scholar]
- Federman AD, Soones T, DeCherrie LV, Leff B, & Siu AL (2018). Association of a Bundled Hospital-at-Home and 30-Day Postacute Transitional Care Program With Clinical Outcomes and Patient Experiences. JAMA Internal Medicine, 178(8), 1033–1040. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Garrido MM, Kelley AS, Paris J, Roza K, Meier DE, Morrison RS, & Aldridge MD (2014). Methods for Constructing and Assessing Propensity Scores. Health services research, 49(5), 1701–1720. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Jackson C, Kasper EW, Williams C, & DuBard CA (2016). Incremental Benefit of a Home Visit Following Discharge for Patients with Multiple Chronic Conditions Receiving Transitional Care. Population health management, 19(3), 163–170. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Lee KK, Yang J, Hernandez AF, Steimle AE, & Go AS (2016). Post-discharge Follow-up Characteristics Associated With 30-Day Readmission After Heart Failure Hospitalization. Medical care, 54(4), 365–372. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Lee PH (2014). Is a cutoff of 10% appropriate for the change-in-estimate criterion of confounder identification? Journal of Epidemiology, 24(2), 161–167. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Levine DM, Ouchi K, Blanchfield B, Saenz A, Burke K, Paz M, Diamond K, et al. (2020). Hospital-Level Care at Home for Acutely Ill Adults: A Randomized Controlled Trial. Annals of Internal Medicine, 172(2), 77–85. [DOI] [PubMed] [Google Scholar]
- Leyrat C, Seaman SR, White IR, Douglas I, Smeeth L, Kim J, Resche-Rigon M, et al. (2019). Propensity score analysis with partially observed covariates: How should multiple imputation be used? Statistical methods in medical research, 28(1), 3–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Liao JM, Navathe A, & Press MJ (2018). Hospital-at-Home Care Programs-Is the Hospital of the Future at Home? JAMA Internal Medicine, 178(8), 1040–1041. [DOI] [PubMed] [Google Scholar]
- Northcott S, Moss B, Harrison K, & Hilari K (2016). A systematic review of the impact of stroke on social support and social networks: Associated factors and patterns of change. Clinical rehabilitation, 30(8), 811–831. [DOI] [PubMed] [Google Scholar]
- Patient-Reported Outcomes Measurement Information System (PROMIS). (2015, July 8). Informational Social Support: A brief guide to the PROMIS Informational Support v2.0 instruments. Retrieved January 22, 2019, from http://www.healthmeasures.net/administrator/components/com_instruments/uploads/15-09-01_15-40-35_PROMISInformationalSupportScoringManual.pdf
- Smith PJ, Snyder LD, Palmer SM, Hoffman BM, Stonerock GL, Ingle KK, Saulino CK, et al. (2018). Depression, social support, and clinical outcomes following lung transplantation: a single-center cohort study. Transplant international: official journal of the European Society for Organ Transplantation, 31(5), 495–502. [DOI] [PubMed] [Google Scholar]
- Tibaldi V, Isaia G, Scarafiotti C, Gariglio F, Zanocchi M, Bo M, Bergerone S, et al. (2009). Hospital at home for elderly patients with acute decompensation of chronic heart failure: a prospective randomized controlled trial. Archives of internal medicine, 169(17), 1569–1575. American Medical Association. [DOI] [PubMed] [Google Scholar]
- Tsuchihashi-Makaya M, Kato N, Chishaki A, Takeshita A, & Tsutsui H (2009). Anxiety and poor social support are independently associated with adverse outcomes in patients with mild heart failure. Circulation journal: official journal of the Japanese Circulation Society, 73(2), 280–287. [DOI] [PubMed] [Google Scholar]
- VanderWeele TJ, & Knol MJ (2014). A Tutorial on Interaction. Epidemiologic Methods, 3(1), 33–72. 10.1515/em-2013-0005 [DOI] [Google Scholar]
- Verhaegh KJ, MacNeil-Vroomen JL, Eslami S, Geerlings SE, de Rooij SE, & Buurman BM (2014). Transitional care interventions prevent hospital readmissions for adults with chronic illnesses. Health Affairs, 33(9), 1531–1539. [DOI] [PubMed] [Google Scholar]
- Villain M, Sibon I, Renou P, Poli M, & Swendsen J (2017). Very early social support following mild stroke is associated with emotional and behavioral outcomes three months later. Clinical rehabilitation, 31(1), 135–141. [DOI] [PubMed] [Google Scholar]
- Wiest D, Yang Q, Wilson C, & Dravid N (2019). Outcomes of a Citywide Campaign to Reduce Medicaid Hospital Readmissions with Connection to Primary Care within 7 Days of Hospital Discharge. JAMA networ en, 2(1), e1. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.