Background: Thirty-day unplanned hospital readmission remains a significantly challenging metric to improve [1]. Evidence suggests successful readmission reduction strategies have both inpatient and outpatient components [2,3]. A Hospital-at-Home (HaH) Early Supported Discharge model may act as a bridge between these components; however, HaH implementation is heterogeneous across health systems (i.e. admission avoidance (AA) versus ESD models) [4]. Despite the rapid rise of HaH programs during the pandemic, there is a lack of data on ESD models supporting high risk of readmission discharges, but it remains an attractive off label intervention for health systems [5]. Here we evaluate the hypothesis that patients receiving HaH ESD would experience decreased odds of composite hospital readmission and mortality at both 7-and 30-days compared to routine discharge (RD).
Methods: We conducted a retrospective cohort study at five hospitals (quaternary academic and regional hospital centers) using electronic health record data. Adults (>18 years old) admitted to Hospital Medicine, General Internal Medicine, or Cardiology services between March 2024 and May 2025 were included if discharged to Home or Home Health Care. Encounters not meeting CMS 30-day unplanned readmission criteria were excluded. In our HaH ESD model, patients are referred to HaH at time of discharge and subsequently seen in person by a HaH trained medical services team and virtually by a hospitalist. We compared cohort populations (HaH vs. RD) using descriptive statistics. Combined readmission and mortality rate were compared at 7-and 30-days from discharge using a multivariate logistic regression. Covariates included age, Charlson Comorbidity Index (CCI), length of stay, ICU stay, encounter diagnosis related group (DRG), and annual care utilization (inpatient visits and primary care visits).
Results: We analyzed 30,791 inpatient encounters (HaH n=4,909; RD n=25,882). Demographics including area deprivation index and prevalence of social determinants of health between HaH and RD were similar between cohorts. The overall cohort was 49.1% female, with a mean age of 62.5 years, and predominantly white (71.4%). Regarding the HaH cohort, median time to first visit was 2 days (mean 2.31 days), and the median number of visits was 2 (mean 2.51). The HaH cohort had higher comorbidity burden compared to RD (mean CCI 5.74 vs. 4.68, p-value < 0.001). Interestingly, Epic System’s 30-day unplanned readmission model (Version 1) underestimated risk of readmission compared to actual rates for both HaH and RD when risk was >20% and < 50% (e.g. of the HaH patients with readmission risk between 31-50%, 51.3% experienced hospital readmission) [6]. HaH patients had higher unadjusted composite of 7 and 30-day unplanned readmission and mortality (7.8% and 21.2%) compared to standard discharge (5.6% and 16.4%). After covariate adjustment, odds of 7 and 30-day composite were higher in HaH (1.27 CI 1.03, 1.57; p-value < 0.027 and 1.12 CI 1.02, 1.23; p-value < 0.014).
Conclusions: A HaH ESD model was not associated with decreased odds of composite readmission and mortality at 7- or 30-days. Hospital readmission is a complex problem, and though a prescription for HaH ESD to bridge the transition of care is attractive, patient selection (risk and impact) and outcome metrics should be considered [7]. Our study is limited by its retrospective nature, generalizability of HaH ESD model and single health system experience.