On-Premises AI-Tool for Generating Nursing Care Summaries: A Phased-Implementation Study in Japan

Introduction: This study details the phased development and implementation of an Artificial Intelligence (AI)-driven nursing summary system designed to reduce nurses’ documentation workload. Methods: Over a 1-year development period beginning in December 2023, we constructed a system that automatically generates nursing care summaries from patient records to support care continuity. The system utilized large language models with prompt tuning techniques and a subjective, objective, assessment, plan (SOAP) format integration. Results: Through three iterations informed by clinical feedback, the system reduced the median summary creation time from 16 to 10 minutes (p < 0.001), representing an approximate 40% reduction. When validating the AI-generated nursing summaries, total scores for the five evaluation items in Versions 0.2 and 0.3 showed no significant difference. Discussion: This system produces comprehensive summaries that consider both nursing care plans and changes in patient condition, simultaneously improving efficiency and enhancing documentation quality. This study demonstrates the importance of iterative development and expert knowledge integration in implementing AI in clinical practice, offering valuable insights for advancing digital healthcare transformation. Conclusion: This study achieved an approximate 40% reduction in nursing record creation time and produced clinically acceptable summary documents through the phased development and implementation of an AI-driven nursing summary system. Keywords: patient discharge summaries, artificial intelligence, nursing, large language models, electronic health records