Real-World Evaluation of a Remote Patient Monitoring Service for Early Detection of Clinical Deterioration in Skilled Nursing Facilities

OBJECTIVES: To evaluate a remote patient monitoring (RPM) service that relies on contactless cardiorespiratory monitoring in skilled nursing facilities (SNFs) for early detection of patient deterioration. DESIGN: Retrospective observational study using real-world clinical service data. SETTING AND PARTICIPANTS: The study was conducted in 184 SNFs in the United States. Participants included all admitted patients enrolled in RPM during the 14-month evaluation period. METHODS: Patients were monitored using a contactless cardiorespiratory bedside monitor. Device and health record data were reviewed by a remote clinical team, and alerts were issued to on-site staff if changes in health status were detected. Sensitivity of RPM alerts to hospital transfers was assessed. Positive predictive value (PPV) of RPM alerts was assessed by considering all alerts associated with deterioration, including hospital transfers and deaths (administrative events), and deterioration successfully treated on site (identified through a large language model classifier). Overall alert burden was determined. The number of hospital transfers potentially prevented was estimated across all facilities as well as in a subgroup of high-adoption facilities. RESULTS: Of the 7242 hospital transfers that were unplanned and potentially preventable, 23.5% were directly preceded by an RPM alert (sensitivity), with an average lead time of 63 hours. Overall PPV across all facilities was 35.4%. RPM alerted on average for 23.9 unique patients per 100 beds per month. An estimated 2.6 hospital transfers were potentially prevented per 100 beds per month across all facilities, whereas 3.5 transfers were potentially prevented per 100 beds per month in high-adoption facilities. CONCLUSIONS AND IMPLICATIONS: RPM using continuous, contactless cardiorespiratory monitoring enabled early identification of certain patient deterioration events with clinically meaningful sensitivity and lead time, and manageable alert burden, complementing standard care. These findings may guide future design and prospective validation of RPM programs using continuous monitoring technology.