Prediction tool for disability progression and mortality in older adults eligible for Japanese long-term care insurance: Koriyama study.

Abstract

This study aims to create a simple model for predicting disability progression and death among older adults with Japanese long-term care insurance certification. This retrospective study analyzed the anonymized data provided by Koriyama City. The participants were 7706 older adults who were initially certified to be support levels 1 and 2 or care levels 1 and 2 for the purpose of obtaining Japanese long-term care insurance. The results of the certification questionnaire at the initial survey stage were used to create decision tree models intended to predict whether disability progression and death would occur within 1 year. In support levels 1 and 2, among those who scored both "daily decision making" item as other than "possible" and the "taking drugs" item as other than "independent," 64.7% had an adverse outcome. In care levels 1 and 2, among those who scored both the "shopping" item as "totally dependent" and the "defecation" item as other than "independent," 58.6% had an adverse outcome. The accuracy of classification of the decision trees were 61.1% in support levels 1and 2, and 61.7% in care levels 1 and 2. The overall accuracy of the decision tree is low, making it impractical to use it for all subjects. Nevertheless, based on the results of the 2 assessments in this study, the process of identifying a particular group of older adults at a high risk of an increased need for long-term care or possible death within a year is very simple and useful.