Proposed Topic (Most preferred): :
Clinical Safety and Quality Service III (Projects aiming at quality service to patients and their carers)
Proposed Topic (Second preferred): :
Clinical Safety and Quality Service I (Projects aiming to improve efficiency and effectiveness of care delivery to meet international standards)
Authors (including presenting author) :
HO SKS, CHAN TS, CHAN HL, TAI TL, LI PY, CHAN YK, LI HY, KAN KC, SIU TS, PANG HS, PO MY, WONG CW, KNG PLC
Affiliation :
Division of Geriatrics, Department of Medicine and Geriatrics, RTSKH
Introduction :
The Hospital (Clinical) Command Centre (HCC) is a key strategic initiative under the HA Smart Hospital Strategy. Its essence is to leverage “real-time” data and artificial intelligence (AI) to optimize operational efficiency and patient safety at the bedside. It identifies high risk cases with alerts and empowers staff with smart visualization of patients’ clinical conditions across a ward for early intervention. HCC was introduced to RTSKH in mid-November 2023. To harness the HCC for elders’ quality and safety care, our team developed a data-driven service model for clinical nursing handover, predicting discharge and unplanned readmission.
Objectives :
To explore a data-driven service model using the HCC in geriatrics wards by Testing its use in clinical nursing handover through staff feedback Proactively supporting high risk unplanned readmission cases with post-discharge geriatric community services Correlating AI predictions for discharge and unplanned readmissions with actual outcomes
Methodology :
Ward nurses of 2 acute and 2 convalescent geriatrics wards were trained to use the HCC for nursing handover. Protocol guide and access right was granted to ward managers and advance practice nurses. Implementation of the HCC guided ‘A shift to P shift handover’ during 4 Dec to 10 Dec 2023 used real-time information - Extreme vitals, Critical lab results, MDRO, Fall risk, Pressure Injury (PI) risk, AI predicted IP discharge and AI unplanned readmission. Staff feedback on the usefulness of HCC in patient care was collected using an on-line survey. Accuracy of AI predicted IP discharge and AI unplanned readmission was tested against actual discharge date and the 7-day readmission data respectively using 2-week data during 4 Dec to 15 Dec 2023. Our data-driven model will pilot the use of AI predicted unplanned readmission risk to trigger geriatric post-discharge community support.
Result & Outcome :
Result: 15 respondents completed the on-line survey with 9 questions using a 6-point scale (1 being strongly disagree and 6 being strongly agree) and one free text question. Overall average score was 3.89 (range 3.67 – 4.27), that reflected nursing staff were slightly satisfied with the usefulness of HCC in daily clinical handover. Higher ratings for usefulness, ranging from 4.27 to 3.93 were given to HCC information panels for MDRO, AI unplanned readmission, Extreme vitals, Critical lab result and PI risk. Overall, most respondents (10, 66.7%) agreed to encourage colleagues to use the HCC as a clinical handover tool. However, negative feedback related to delay of laboratory results update, and missing data on unplanned readmission. AI models could identify causal variables used in its algorithms leading to targeted interventions. Accuracy of AI predicted discharge was high, 34 (69.4%) of 49 patients were discharged on the date or within 3 days as predicted. Its sensitivity is 32.69% (95% CI 23.8% – 42.5%) and specificity is 97.64% (95% CI 96.14% - 98.67%). For 70(9.5%) patients tagged with AI unplanned readmission risk, 19 (27.1%) lived at home and 51 (72.9%) lived at residential care homes for the elderly (RCHE), while 24 (47.1%) are covered by HKECGAT. As at 5/1/2024, 13(14.3%) cases were readmitted in 7 days after discharged, 10 lived at residential home and 3 patients lived at home. Presently, HCC application in nursing handover has limited utility and could be strengthened if more timely updates and integration with the Clinical Dash Board to allow “One Stop Information Point”. The good accuracy of AI predicted discharge can facilitate efficient ward turnover. AI unplanned readmission tags in our data-driven service model automatically trigger early attention for community services such as tele clinic and nurse follow up for better support post-discharge. Initial results show favorable results in reducing unplanned readmission.
