Development of a nomogram model for predicting dementia risk in the older adult population of Weifang, Shandong Province, China: based on the biopsychosocial model

TL;DR

This study creates a dementia risk prediction model for older adults in Weifang, China, using a biopsychosocial approach and validated with a large sample.

Contribution

A novel dementia risk nomogram model tailored for older adults in Weifang, integrating biopsychosocial factors and validated with high predictive accuracy.

Findings

The model achieved an AUC of 0.751 in training and 0.794 in testing, indicating strong predictive performance.

Key predictors included napping, lack of concentration, education level, and social interaction.

Calibration curves showed good alignment between predicted and observed dementia cases.

Abstract

Dementia has emerged as a predominant health challenge. However, there is a notable research gap in the collective screening of dementia risks. Hence, there is a pressing need to formulate a dementia prediction tool tailored to the older adult demographic, enabling the identification of high-risk individuals for dementia.

From May to October 2023, a multi-stage sampling method was utilized to survey older adult individuals aged 60 and above in Weifang. This study employed the Brief Community Screening Instrument for Dementia (BCSI-D) for the identification of individuals with dementia. We integrated the biopsychosocial model to construct a comprehensive pool of factors influencing dementia. Employing the least absolute shrinkage and selection operator and multivariate logistic regression analyses, independent influencing factors were identified to construct a nomogram prediction model.

Six hundred and sixty valid questionnaires were included in the final analysis, with a validity rate of 95.23%. We identified 178 cases of dementia using the BCSI-D. Napping, lack of concentration, self-assessed health status, education level, residence, social interaction and medical insurance were independent influencing factors for dementia. The efficiency analysis of the prediction model, constructed using these factors, demonstrated area under the receiver operating characteristic of 0.751 for the training set and 0.794 for the test set. The decision curve analysis threshold probabilities for the training and test sets were 5–60% and 1–60%, respectively. The calibration curves of both datasets exhibited a high degree of fitting with the predicted curve.

We developed a dementia risk identification model with noteworthy predictive performance. The proposed model offers theoretical and data support for collective dementia screening.

Figures