Paochien Hospital, a hospital in Pingtung, Taiwan, with more than 30 years of history, built a hospital-wide AI nursing handover system on Amazon Bedrock with AWS partner Advanced Medical Information, according to a May 2026 sponsored article on INSIDE produced for AWS. Nurses' documentation and handover time fell from 105 minutes to 49 minutes, a 53% reduction.
The hospital made nursing handover and medical record translation its first targets: doctors write records in English, and nurses had been translating them, sometimes by photographing them and using online translation. The volume is large: an intensive care patient staying 21 days can generate more than 300 nursing records of around 200 characters each. Most AI tools the IT team evaluated were too costly before it turned to Amazon Bedrock. AWS staff built a prototype in two days to show accuracy, model speed and user experience, and the system was tested and put into use in less than a month.
To protect patient data, Advanced Medical Information designed an architecture in which data is sent through APIs with access control and de-identification applied before it is uploaded, so personal information is not exposed to the large language model. The nursing team compiled clinical abbreviations and Chinese-English terms into a dedicated medical dictionary and worked with the IT team to tune prompts so the model handles clinical terminology. The vendor connected the eight systems that hold nursing records, and the interface includes a keyword traceback function that lets nurses jump from a summary to the original record to guard against AI errors. Red, yellow and green alerts highlight abnormal vital signs and lab values.
The hospital plans to extend AI to physicians' daily progress notes and discharge summaries and to an internal assistant for administrative staff.
Details
- City
- Pingtung
- Organization
- Paochien Hospital
- Continent
- Asia
- Country
- Taiwan