Authors (including presenting author) :
Lui CT (1), Tong WH (2), Hui Coe (2), Kong Karol (2), Kwok WL (2), Chan HL (2), To CH (2), Lam MS (3), Leung Simon (4), Chan Peter (5), Lo Edwin (5), Kwok Andy (6), Hau LM (6), Tang Karson (7), Wong Xanthus (7), Lo Jason (8), HA IT Innovation Lab (8), Ho TY (9), Leung FT (9), Yuen CY (9), Fong KH (9)
Affiliation :
(1) Department of Accident and Emergency, NTWC
(2) Department of Medicine and Geriatrics, Tin Shui Wai Hospital
(3) Nursing Services Department, NTWC
(4) Biomedical Engineering Services Section, NTWC / HAHO
(5) Information Technology Department, NTWC
(6) Quality and Safety Division, NTWC
(7) Administrative Service Division, NTWC
(8) IT Innovation Lab, HOIT&HI, Hospital Authority
(9) IoT Sensing and AI Technologies (IOTSAI), Hong Kong Applied Science and Technology Research Institute (ASTRI)
Introduction :
Unexpected patient deterioration, fall, and missing patients are persistent threats of in-patient ward operations worldwide. Traditional vital monitoring equipment, including telemetries, are bulky with lots of cables, which would be applied to high-risk patient only. In addition, there is no simultaneous geofencing and fall detection features to tackle different observation pain-points in wards.
Objectives :
To develop and pilot an all-in-one chest patch wearable device in wards to expand the scope of continuous observations of ward patients.
Methodology :
An Innovation Technology Funded project was led by Tin-Shui-Wai Hospital and ASTRI. A 59-gram all-in-one chest wearable, named CHIMP patch, which includes sensors for electrocardiographic signals, sweat, motion and locating engines were developed. CHIMP would continuously monitor the patients’ ECG, respiratory and heart rate, location and motion. The wearable is integrated to a total-solution including IoT-network, backend server with AI analytics, and closed-loop notification to ward staffs by native apps on working mobile and dashboard when there is patient alert. The alerts include 33 arrhythmia patterns, cardiac arrest rhythms including pulseless electrical activity, tachy/bradycardia, tachy/bradypnea, fall motion, diaphoresis and geofencing alerts. The AI models were developed with ResNet neural network, LSTM and rule-based algorithm on various public datasets (PTB-XL, KFall), and externally validated with local data during the pilot.
Result & Outcome :
CHIMP was piloted in two wards in TSWH since Jul 2024. Patients would be introduced on the CHIMP patch on admission and provided leaflet. In first 4 months of pilot, there are totally 1,142 patient-days observation by CHIMP. There are 3 cardiac arrest events all picked up by the CHIMP patch. 4 patients with complete-heart-block pattern and 10 patients of supraventricular tachycardia revealed. There are 131 alert episodes on tachycardia and bradycardia true alerts which assisted early recognition and treatment. The positive predictive value of arrhythmia and heart-rate alerts are on average 60-80%, while the precision of left-ward geofencing alert was 90%. The prototype power of CHIMP can last 2.5 days and then require change-over. A patient satisfaction survey revealed more than 80% patients agreed on enhanced observation and improved sense of safety, and well tolerated on wearing during hospital stay.
There is tremendous potential of AI-IoT technology in enhancing patient observation both in in-patient, A&E and community setting in ambulatory service model and telemedicine. A productization and scale-up is underway.
