베이지안 네트워크를 활용한 치매 의심군 예측요인 연구

Abstract

Objective: This study aims to identify predictive factors for screening suspected dementia among middle-aged and older adults and provide basic data to prevent dementia and prepare early customized intervention plans for middle-aged and older adults.
 Method: A survey of 850 middle-aged and older adults aged 40-65 years living in Japan was conducted to collect data, and a Bayesian network was used to identify predictors and pathways for screening suspected dementia.
 Results: First, to verify the accuracy of the results extracted by the Bayesian network model, the performance was compared with random forest and gradient boosting, and the results showed that the Bayesian network algorithm performed the best, with a predictive power of 90%. Second, among the variables predicting suspected dementia, the most predictive variable was subjective memory decline, followed by early maladaptive behavior, social comfort, adult attachment anxiety, dementia fear, and adult attachment avoidance. Third, the core pathway for suspected dementia was identified as neuroticism to early maladaptive behavior and adult attachment anxiety to suspected dementia, followed by neuroticism to subjective memory decline, social comfort, and adult attachment anxiety to suspected dementia.
 Conclusion: By utilizing Bayesian networks to explore important variables and pathways between variables in suspected dementia, we can provide a framework for early intervention and customized policies for dementia.