Background: Hospital-at-Home (HaH) is a safe and effective alternative to high-cost, traditional hospitalization in patients with a range of medical conditions, such as non-severe community-acquired pneumonia (CAP). The COVID-19 pandemic created a unique situation with favorable policy changes, hospital resource constraints, and a clearly defined patient population that helped to increase HaH program deployment and mitigate known challenges to HaH utilization. However, once hospital demand associated with COVID-19 diminishes, established barriers to HaH implementation are likely to limit sustainability at the current scale. We hypothesize that Clinical Decision Support (CDS) that integrates data from disparate sources and risk scoring can improve point-of-care decisions regarding patient eligibility for HaH, thus overcoming historically low participation rates. We present preliminary findings from an electronic health record (EHR) algorithm to identify potential HaH-appropriate patients with a high probability of CAP as an initial step towards development of an interactive CDS application to improve HaH utilization.

Methods: We conducted a pre-deployment, prospective validation study to examine appropriateness of candidates identified for potential HaH enrollment between September 29 to October 29, 2021. Each day, our EHR algorithm identified adults (≥18 years) with suspected CAP who were hospitalized via the emergency department (ED) at any of 5 hospitals within a large healthcare system in the Southeastern U.S. Eligible patients had (1) antibiotics ordered in ED, (2) chest X-ray or CT performed in ED, and (3) low (71-90) to moderate (91-130) risk of severe CAP, as determined by Pneumonia Severity Index (PSI) score. We excluded patients with abnormal clinical or lab findings associated with conditions that may preclude receipt of HaH care (i.e., creatine ≥2, troponin ≥30, sodium < 128 or >150, hemoglobin ≤9, bilirubin ≥3, lactate ≥2.5, heart rate >115, systolic blood pressure < 90 or ≥180, altered mental status, oxygen support >4L/min, or stroke protocol initiated with head CT). Patients who lived outside of the state or those admitted from a skilled nursing facility were also excluded to align with eligibility criteria for our HaH program. All sociodemographic, clinical, and outcomes data were captured from the EHR. The primary outcomes were hospital mortality and requirement for intensive care unit (ICU) admission within 72 hours. We also evaluated discharge diagnoses using ICD 10 codes, including pneumonia (J12-J18) and COVID-19 (U07.1). We conducted descriptive analyses to assess patient characteristics and clinical outcomes.

Results: There were 198 eligible patients, including 100 (50%) female, 142 (72%) white, and with a median age of 74 years (Table). Based on the PSI score at eligibility determination, 93 (47%) had low risk of severe CAP and 105 (53%) had moderate risk. The median length of hospital stay was 4.2 days. Overall, 2 (1%) patients died in the hospital, and 6 (3%) patients required ICU admission within 72 hours. There were 148 (75%) patients who had a discharge diagnosis related to any type of infection, including 49 (25%) patients who had either pneumonia or COVID-19 diagnoses.

Conclusions: Few HaH eligible patients experienced adverse outcomes, supporting the feasibility of our EHR-based approach to select appropriate low-risk candidates. Additional opportunities exist to increase specificity for CAP identification.

IMAGE 1: The characteristics of the CAP cohort