Authors (including presenting author) :
Chung KY(1), Chan CY(1), Wan S(1), Mak MY(1)
Affiliation :
(1)Physiotherapy Department, Tuen Mun Hospital, Hospital Authority
Keyword 1: :
AI-assisted Generation
Keyword 2: :
Medical reports writing
Keyword 3: :
Smart Hospital
Introduction :
According to HA’s strategic plan on developing smart hospitals, integration of advanced technology to streamline clinical workflows is a key priority. For decades, traditional method of drafting medical reports in the Physiotherapy OPD of TMH relied on manual synthesis of fragmented notes - a time-intensive process that competed with direct patient care. Addressing this need proactively, we pioneered a “Secure AI-Assisted Workflow” for generation of physiotherapy medical reports tailored to our requirements with no financial implication. Through iterative clinician-led testing, we successfully engineered our own specialized AI prompt at the HA GenAI platform. This initiative demonstrated how frontline professionals could independently drive digital transformation to enhance work efficiency.
Objectives :
(1)To evaluate the efficiency gain of the AI-assisted workflow compared to the traditional workflow (2)To assess user confidence in the accuracy and quality of the AI-generated report
Methodology :
We self-developed a robust AI prompt capable of structuring anonymized notes into professional case summaries at the HA GenAI platform with secure internal access. AI-assisted workflow involved: 1) physiotherapists extracted clinical notes from the Clinical Management System (CMS), 2) input data into AI prompt at the HA GenAI platform to generate a structured report draft, 3) reviewed and finalized the report. In Jan 2026, 10 physiotherapists were assigned into two groups: Control Group (n=5) using the traditional workflow, and Study Group (n=5) using the AI-assisted workflow. For comparison, we measured the average processing time to generate a report for the same case in both groups. Additionally, physiotherapists in the Study Group rated their confidence in the accuracy and quality of the AI-generated report on a 10-point scale (1 = not at all confident, 10 = extremely confident).
Result & Outcome :
Results: Significant reduction in the processing time (p< 0.001) was found in the Study Group (average 12.72 mins per report) compared to the Control Group (average 33.29 mins per report). Moreover, Study Group reported high confidence in the accuracy and quality of the AI-generated report (average score = 7.8/10). Conclusion: We demonstrated that the “Secure AI-Assisted Workflow” was a viable and efficient alternative to traditional method for generation of physiotherapy medical reports. We effectively integrated smart technology into daily practice through independent innovation to enhance our work efficiency.
