Artificial Intelligence Approaches to Social Determinants of Cognitive Impairment and Its Associated Conditions.

Abstract

Background and purpose: This study uses an artificial-intelligence model (recurrent neural network) for evaluating the following hypothesis: social determinants of disease association in a middle-aged or old population are different across gender and age groups. Here, the disease association indicates an association among cerebrovascular disease, hearing loss and cognitive impairment.

Methods: Data came from the Korean Longitudinal Study of Ageing (2014-2016), with 6,060 participants aged 53 years or more, that is, 2,556 men, 3,504 women, 3,640 aged 70 years or less (70-), 2,420 aged 71 years or more (71+). The disease association was divided into 8 categories: 1 category for having no disease, 3 categories for having 1, 3 categories for having 2, and 1 category for having 3. Variable importance, the effect of a variable on model performance, was used for finding important social determinants of the disease association in a particular gender/age group, and evaluating the hypothesis above.

Results: Based on variable importance from the recurrent neural network, important social determinants of the disease association were different across gender and age groups: 1) leisure activity for men; 2) parents alive, income and economic activity for women; 3) children alive, education and family activity for 70-; and 4) brothers/sisters cohabiting, religious activity and leisure activity for 70+.

Conclusions: The findings of this study support the hypothesis, suggesting the development of new guidelines reflecting different social determinants of the disease association across gender and age groups.