Development of Prediction Models for Needs and Resources Service Using Big Data Related to Elderly Care

Abstract

Objectives: In this study, we aimed to verify the possibility of utilizing the integrated case management big data in Korea to identify the complex care needs of the elderly in the community and link them to appropriate resources through a machine learning-based predictive model. Methods: Three merged datasets were made by combining different combinations of the data from integrated case management, elderly care service, and Community Integrated Care Leading Project conducted in Korea. Due to the issue of multiple selection of the response variables, the analysis considering multi-label classification method was conducted. The performance of the predictive model was evaluated based on the F1 score. Results: (1) ‘Maintaining daily life’ and ‘Health’ labels in the needs, (2) ‘Maintaining physical health’ and ‘food, clothing and shelter’ labels in the problem domain, (3) ‘Physical health’, ‘daily life’, and ‘housing’ labels in the major classification of resources, and (4) ‘meal food support’, ‘disease prevention and health care’, and ‘living environment improvement’ labels in the major classification of resources showed F1 ≥ 0.8. Conclusions: In this study, we verified that the elderly need-resource linkage prediction based on big data for integrated case management was at an appropriate level, and the possibility of supporting integrated case management using this was confirmed.