To alleviate the burden of manual tasks on medical personnel, Taichung Veterans General Hospital has collaborated with Yang Ming Chiao Tung University to develop "Healthcare GPT." This large-scale language model tool, developed using de-identified medical data from the hospital, can be used in medical record writing, consultation systems, email responses, and accessing meeting data, among other areas. It saves the time of medical staff, optimizes processes, and facilitates more precise clinical decisions, with an estimated monthly reduction of 60,000 minutes in processing time.
Taichung Veterans General Hospital, in collaboration with the team led by Yang Ming Chiao Tung University's distinguished professor Henry Horng-Shing Lu , invested approximately one year in developing Healthcare GPT using de-identified medical data from the hospital. Through deep learning and natural language processing techniques, it can generate fluent, natural text outputs and answer complex questions. Not only is the output content more suitable for professional medical fields, but it also alleviates concerns about patient data leakage and privacy infringement.
Huang Sheng-Yang, Deputy Executive Director of the Smart Healthcare Committee at Taichung Veterans General Hospital, pointed out that Healthcare GPT significantly reduces the time costs for staff, allowing colleagues to focus more on their essential work. Currently applied in the Medical Record Agent, it automatically completes medical record summaries within 15 seconds using advanced natural language processing techniques. The Call Center AI consultation system provides one-to-many synchronous responses to public queries, integrates and accurately analyzes public inquiries, and then forwards them to professional nurses for further consultation. Email Composer understands email content, automatically collects relevant problem-solving methods, and drafts specific responses for staff, generating precise response drafts. The KM Agent can quickly retrieve accurate summaries of internal meeting records, technical documents, and guidance decision-making matters in 15 seconds.
Taking the example of public complaints in the social work studio, the smart reply function of the Email Composer only needs to interpret the content of the email, then retrieve information from the database and compile draft documents. This, combined with the internal process review and modification, shortens the previous process of obtaining answers by coordinating with other units. The KM Agent only needs 15 seconds to obtain accurate summaries of internal meeting records, technical documents, and guidance decision-making matters, providing more intuitive data presentation during meetings.
A nurse named Chen responsible for clinical operations stated that the automatic completion of medical record summaries liberates medical personnel from tedious text work, enabling more effective use of time to improve medical communication and care quality. This is undoubtedly a timely boost to the current shortage of medical personnel. Another nurse responsible for the 24-hour remote care center mentioned that the system can immediately generate accurate response text by recognizing the voice of the caller, helping him avoid spending time searching for information to respond to callers, greatly reducing wait times for callers.
Resource (Mandarin):
