PD28-12 USING REMOTE MONITORING USING COMMERCIALLY AVAILABLE DEVICES TO PREDICT RE-ADMISSIONS FOR PATIENTS DISCHARGED FOLLOWING RADICAL CYSTECTOMY

Abstract

INTRODUCTION AND OBJECTIVE: Radical cystectomy (RC) has 30-day and 90-day re-admission rates of over 15% and 20% respectively. The majority of complications are infection and bowelrelated. Modern Bluetooth-enabled commercial wearable devices such as smart watches can measure and transmit a patient's health data directly to a mobile phone. Data is either stored on the phone for future retrieval or uploaded to a web server in real-time to facilitate remote patient monitoring. In this study, we test the feasibility of remote monitoring for 30 days in patients discharged from hospital following RC. METHODS: As part of the DREAMPath (ISRCTN62293620) study, patients were provided with an iPhone, Apple Watch Series 4 and three Bluetooth peripherals (sphygmomanometer, pulse oximeter, thermometer). They were instructed to wear the Apple Watch at all times, and measure their observations twice a day. Additionally, patients were asked to complete a quality of recovery 15 questionnaire daily. At the end of 30 days, re-admission data was collected. A backend data pipeline has also been developed to process collected data, allowing us to develop clinically-useful analytics. The primary focus is on developing anomaly detection and machine learning models that predict premature hospital re-admission. RESULTS: 20 patients have been recruited into the study. At the time of our analysis, 5,687 data points have been collected per patient during the month of monitoring. Four patients were re-admitted: reasons for admission are pulmonary embolism (1), infection (2) and stent problems (1) respectively. Prior to each re-admission, the model constructed around these physiological measures was able to predict re-admission > 48 hours prior to patient-initiated visit to hospital accident and emergency department, except for the patient with stent problems. The patient with stent symptoms was identified due to reduction in their daily questionnaire score. CONCLUSIONS: Remote monitoring is feasible, and patients can engage with it with minimal effort all while generating a large amount of data that is relevant to their recovery. In this small study, we show that it is possible to detect physiological disturbances before patients feel unwell enough to be re-admitted to hospital. It would be informative to design an intervention to triage these patients earlier in an effort to reduce total re-admission days.