Home Hospital” — Applications, Concrete Results, and Challenges

The analysis explores the feasibility of implementing a Home Hospital (HaH) system with AI conversational assistants in Romania, concluding that there are grounds for a pilot project, but national scaling requires overcoming significant barriers.

19 October 2026 · 5 min

Key points

The analysis explores the feasibility of implementing a Home Hospital (HaH) system with AI conversational assistants in Romania, concluding that there are grounds for a pilot project, but national scaling requires overcoming significant barriers.
Strengths: The political framework is favorable through the National Digitalization Strategy in Health 2026–2030, and the Electronic Health Record (EHR) will be available from summer 2026. The European project TRANSCARE, coordinated by the Heart Institute in Cluj-Napoca, demonstrates the ability to attract funding and validate AI-assisted remote monitoring. Private home care providers are already authorized, and Romania is aligning with Regulation (EU) 2024/1689 (AI Act).
Major barriers: The legal framework for telemedicine is general and undifferentiated, with implementation rules still undeveloped. The ECOG criterion 3 or 4 severely limits patient eligibility. Home care is chronically underfunded (0.20% of medical expenses), and AI governance in health is not systematically addressed. AI literacy among professionals is moderate, and access to services remains uneven, especially in rural areas.

1. Introduction: Why AI in Hospital at Home?

The "Hospital at Home" (HaH) model has already demonstrated solid results regarding mortality, readmissions, patient satisfaction, and costs. However, scaling this model has encountered a fundamental barrier: the manual identification of eligible patients is extremely time-consuming and resource-intensive. In a typical program, medical staff must manually review hundreds of medical records to find a few patients who meet the eligibility criteria. This explains why, despite strong evidence, most HaH programs remain small. Artificial intelligence intervenes precisely at this critical point, providing solutions for automating triage, predicting complications, continuous monitoring, and optimizing logistics. In the following sections, I will analyze each of these applications in depth, with concrete results from real implementations.

Applications of AI in Home Hospitalization 2.1 Automated Identification and Triage of Eligible Patients This is one of the most promising applications of AI in Home Hospitalization (HaH). Several healthcare systems have implemented artificial intelligence-based tools that scan electronic medical records to identify patients eligible for acute home care. University of Utah Health has implemented an AI tool that continuously evaluates vital signs, laboratory values, and clinical stability indicators to flag patients who may be ready for home care. The "Heal at Home" program has expanded from 2 specialties in 2021 to 22 specialties in 2026, serving over 1,500 patients, with satisfaction rates nearing 99% and fewer emergency visits and readmissions compared to traditional discharges. Dr. Kensaku Kawamoto, chief health AI transformation officer at U of U Health, stated: "Instead of manually reviewing over 100 individual patient charts, we can quickly review a smaller set of patients flagged by the tool."

2. Applications of AI in Home Hospitalization

2.1 Identification and Automated Triage of Eligible Patients This is one of the most promising applications of AI in HaH. Several healthcare systems have implemented artificial intelligence-based tools that scan electronic medical records to identify patients eligible for acute home care. University of Utah Health has implemented an AI tool that continuously assesses vital signs, lab values, and clinical stability indicators to flag patients who may be ready for home care. The "Heal at Home" program has expanded from 2 specialties in 2021 to 22 specialties in 2026, serving over 1,500 patients, with satisfaction rates nearing 99% and fewer emergency visits and readmissions compared to traditional discharges. Dr. Kensaku Kawamoto, chief health AI transformation officer at U of U Health, stated: "Instead of manually reviewing over 100 individual patient charts, we can quickly review a smaller set of patients flagged by the tool." Mass General Brigham, which operates one of the largest HaH programs in the U.S., began using an internally developed AI tool in April 2026 that helps emergency department clinicians prioritize patients for referral to home hospital care. The tool scans electronic medical records and classifies eligibility based on diseases, vital signs, and other factors, updating rankings in real-time as new information is recorded. Dr. Shiv Sutaria, assistant chief information officer, explained: "If it sees a new very low blood pressure, as soon as the update appears, it will lower the patient on the ranking list." UMass Memorial Health is developing an AI tool with MIT that will consider approximately 30 different criteria from electronic medical records, including how patients compare to others previously enrolled in the HaH program with similar conditions. Michael Kalfopoulos, associate medical director, stated: "With this tool, our nurses will need to screen less than half of the patients they currently screen to find a single candidate for the program." Jewish General Hospital (JGH) in Montreal has developed, together with Signal 1, an AI tool for patient screening. Dr. Lawrence Rudski, chief of cardiology, explained: "There is a heavy burden on clinicians for screening patients for admission to the program, which is one of the reasons why Hospital@Home has not expanded more rapidly in other jurisdictions. Screening is a time-consuming process, and there is a formidable cognitive load on doctors and nurses." The JGH program has treated approximately 3,400 patients at home since 2022, with a satisfaction rate of 93%, freeing up thousands of bed days and reducing congestion in the Emergency Department. Baptist Health in Jacksonville has partnered with Vega Health to implement a screening AI tool that uses a large language model to scan the records of hospitalized patients according to Baptist Health's eligibility criteria. Alison Bartfield, MD, medical director, stated: "Accurate eligibility screening is fundamental to the program, as it prioritizes patient safety and allows the program to grow." 2.2 Prediction of Readmissions and Complications AI demonstrates remarkable capability in predicting which patients are at risk of readmission or complications, allowing for proactive interventions. A study published in Frontiers in Medicine (2025) utilized machine learning algorithms to predict the readmission of COVID-19 patients transferred to the HaH program. Among the tested algorithms, Support Vector Machine (SVM) showed the best performance: 85% sensitivity, 87% specificity, 86% accuracy, 84% F1 score.