A simple algorithm to predict disability in community-dwelling older Japanese adults.

BACKGROUND

With a worldwide aging population, the prevention of disability in older adults has become an important issue. Therefore, the purpose of this study was to develop a model for predicting disability risk in older adults based on multiple factors, using a decision tree analysis. This model may be used with a mobile application when it is difficult to interview older adults, and to obtain individualized information for prioritizing interventions.

METHODS

We examined the data from a cohort study conducted by the National Center for Geriatrics and Gerontology-Study of Geriatric Syndromes. We included 12,000 older adults without a disability and performed a decision tree analysis using the Chi-square automatic interaction detection (CHAID) algorithm.

RESULTS

Among the 12,000 participants without a disability, 11,503 and 497 participants remained disability-free and developed disability, respectively. The CHAID analysis identified 24 end nodes with five levels of partition and 16 partitioning variables for 34 questionnaire variables, with incident disability probabilities ranging from 0.0% to 96.7%. The classification accuracy and area under the curve of the CHAID model were 73.4% and 0.76, respectively. We found that maintaining mental health was important for older adults in their 80s and older, and that lifestyles and geriatric syndromes were important factors for those in their 70s.

CONCLUSIONS

The magnitude of the influences on the risk of developing a disability differ by age group. The results of this study may provide useful information for the development of mobile applications that predict the risk of developing disability and create tailor-made interventions.