April 2020to March 2021

[Secretariat]
Kazuko Yuma (Chief Fellow, Institute for International Socio-Economic Studies)

   We conducted research on AI to support the creation of nursing care plans from FY2017 to FY2019 (Phase 1) as part of projects for promoting health and wellness for the elderly (Elderly health and welfare promotion subsidies) and summarized the results and issues identified in our studies. Leveraging the results of Phase 1, from FY2020, we have been continuing our research activities towards social implementation for the next three years under Phase 2 of the research.

   For FY2020, which is the first year of Phase 2, we conducted the following empirical studies with the cooperation of two local governments in addition to the residential care support facilities.

   In Study A, we continued initiatives began last year on the visualization of the care manager's thoughts (care-manager-oriented visualization) and enhanced the single-line model designed for users (cardiac patients) into a multi-line model.

   In Study B, we worked on refining the structured and organized labels by checking them with standardized items for appropriate care management.

   In Study C, we held an online training session jointly with local governments for care managers to learn methods for standardizing care management (cardiac disease). In addition to the classroom lectures, as part of the training, participants reviewed the Care Plan 2 they created, and the insights from the review were converted into data. By clarifying what care items are frequently seen in care plans and what items are often lacking, we will also obtain evidence for rule-making, such as in prioritization of care items in the AI for care planning support.

   In Study D, we conducted a literature review to compare the assessment items (company-wide collaboration system) with other objective evaluation criteria in order to reexamine how the changes in the user's condition (outcome index) are grasped in the AI for care planning support.

   In the AI analysis of Study E, we conducted an analysis to grasp the changes in conditions by prioritizing the subjective evaluation of the monitoring sheet. Also, we changed the method for data cleansing of input data from the method used in Phase 1 to increase interpretability.

   Through these experiments, we will continue to examine how we can increase the accuracy of the algorithms and improve the feasibility of the AI for care planning support.