Using machine learning algorithms for predicting cognitive impairment and identifying modifiable factors among Chinese elderly people

Shuojia Wang, Weiren Wang, Xiaowen Li, Yafei Liu, Jingming Wei, Jianguang Zheng, Yan Wang, Birong Ye, Ruihui Zhao, Yu Huang, Sixiang Peng, Yefeng Zheng, Yanbing ZengFrontiers in Aging Neuroscience 2022 · DOI

Abstract

Objectives: This study firstly aimed to explore predicting cognitive impairment at an early stage using a large population-based longitudinal survey of elderly Chinese people. The second aim was to identify reversible factors which may help slow the rate of decline in cognitive function over 3 years in the community. Methods: We included 12,280 elderly people from four waves of the Chinese Longitudinal Healthy Longevity Survey (CLHLS), followed from 2002 to 2014. The Chinese version of the Mini-Mental State Examination (MMSE) was used to examine cognitive function. Six machine learning algorithms (including a neural network model) and an ensemble method were trained on data split 2/3 for training and 1/3 testing. Parameters were explored in training data using 3-fold cross-validation and models were evaluated in test data. The model performance was measured by area-under-curve (AUC), sensitivity, and specificity. In addition, due to its better interpretability, logistic regression (LR) was used to assess the association of life behavior and its change with cognitive impairment after 3 years. Results: Support vector machine and multi-layer perceptron were found to be the best performing algorithms with AUC of 0.8267 and 0.8256, respectively. Fusing the results of all six single models further improves the AUC to 0.8269. Playing more Mahjong or cards (OR = 0.49, 95% CI: 0.38-0.64), doing more garden works (OR = 0.54, 95% CI: 0.43-0.68), watching TV or listening to the radio more (OR = 0.67, 95% CI: 0.59-0.77) were associated with decreased risk of cognitive impairment after 3 years. Conclusions: Machine learning algorithms especially the SVM, and the ensemble model can be leveraged to identify the elderly at risk of cognitive impairment. Doing more leisure activities, doing more gardening work, and engaging in more activities combined were associated with decreased risk of cognitive impairment.

Introduction

As the global population ages, cognitive decline and dementia have become critical public health challenges. While early identification is key to effective intervention, many current predictive models rely on expensive or invasive biomarkers, such as brain imaging or genetic testing, which are often inaccessible for large-scale community screening. This study sought to determine if cognitive impairment could be accurately predicted years in advance using low-cost, non-invasive longitudinal survey data and to identify modifiable lifestyle factors—specifically those relevant to Chinese culture—that might slow cognitive decline.

Methods and Framework

The researchers utilized a robust machine learning pipeline to analyze longitudinal data. Feature selection was performed using Sequential Forward Floating Selection (SFFS) to identify the most predictive variables from demographics, activities of daily living (ADL), and lifestyle habits.

Schematic diagram of the prediction framework.

The framework compared six distinct algorithms: Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF), XGBoost (XGB), LightGBM (LGB), and Multi-layer Perceptron (MLP). Additionally, a "stacking" ensemble method was used to fuse these models. While the ML models provided superior predictive power, Logistic Regression was employed to calculate Odds Ratios (OR) for better interpretability of lifestyle factors.

Data and Participants

Data were drawn from 12,280 elderly individuals (totaling 22,920 records) across four waves of the Chinese Longitudinal Healthy Longevity Survey (CLHLS) from 2002 to 2014. The CLHLS covers 23 out of 31 provinces in China, representing approximately 85% of the population. Cognitive function was assessed using the Chinese version of the Mini-Mental State Examination (MMSE), with impairment defined as a score below 18.

The inclusion and exclusion criteria of this study.

Key Results

Predictive Performance

The machine learning models demonstrated high accuracy in predicting cognitive impairment three years in advance. The SVM and MLP were the top individual performers with AUCs of 0.8267 and 0.8256, respectively. The Ensemble model further optimized performance, achieving an AUC of 0.8269. These results suggest that non-linear interactions between risk factors are better captured by ML than traditional statistical methods.

Modifiable Lifestyle Factors

The study identified several protective factors associated with a significantly reduced risk of cognitive impairment:

  • Playing more Mahjong or cards: OR = 0.49 (51% reduced risk).
  • Doing more gardening: OR = 0.54 (46% reduced risk).
  • Watching TV or listening to the radio more: OR = 0.67 (33% reduced risk).

Notably, the change in behavior was critical; individuals who increased their frequency of these activities showed a marked decrease in risk compared to those with static habits. Culturally specific activities like Mahjong, which require social interaction and strategic planning, likely contribute to "cognitive reserve."

Conclusion

Machine learning algorithms, particularly SVM and ensemble models, can be effectively leveraged to identify elderly individuals at risk of cognitive impairment using simple survey data. Engaging in more leisure activities and gardening is strongly associated with a decreased risk of decline. While limited by self-reporting and the lack of biological data, this study provides a scalable framework for community-based cognitive health screening.

Summary updated

Abstract

<b>Objectives:</b> This study firstly aimed to explore predicting cognitive impairment at an early stage using a large population-based longitudinal survey of elderly Chinese people. The second aim was to identify reversible factors which may help slow the rate of decline in cognitive function over 3 years in the community. <b>Methods:</b> We included 12,280 elderly people from four waves of the Chinese Longitudinal Healthy Longevity Survey (CLHLS), followed from 2002 to 2014. The Chinese version of the Mini-Mental State Examination (MMSE) was used to examine cognitive function. Six machine learning algorithms (including a neural network model) and an ensemble method were trained on data split 2/3 for training and 1/3 testing. Parameters were explored in training data using 3-fold cross-validation and models were evaluated in test data. The model performance was measured by area-under-curve (AUC), sensitivity, and specificity. In addition, due to its better interpretability, logistic regression (LR) was used to assess the association of life behavior and its change with cognitive impairment after 3 years. <b>Results:</b> Support vector machine and multi-layer perceptron were found to be the best performing algorithms with AUC of 0.8267 and 0.8256, respectively. Fusing the results of all six single models further improves the AUC to 0.8269. Playing more Mahjong or cards (OR = 0.49,95% CI: 0.38-0.64), doing more garden works (OR = 0.54,95% CI: 0.43-0.68), watching TV or listening to the radio more (OR = 0.67,95% CI: 0.59-0.77) were associated with decreased risk of cognitive impairment after 3 years. <b>Conclusions:</b> Machine learning algorithms especially the SVM, and the ensemble model can be leveraged to identify the elderly at risk of cognitive impairment. Doing more leisure activities, doing more gardening work, and engaging in more activities combined were associated with decreased risk of cognitive impairment.
BibTeX
@inproceedings{wang2022using,
  author = {Shuojia Wang and Weiren Wang and Xiaowen Li and Yafei Liu and Jingming Wei and Jianguang Zheng and Yan Wang and Birong Ye and Ruihui Zhao and Yu Huang and Sixiang Peng and Yefeng Zheng and Yanbing Zeng},
  title = {{Using machine learning algorithms for predicting cognitive impairment and identifying modifiable factors among Chinese elderly people}},
  booktitle = {Frontiers in Aging Neuroscience},
  year = {2022},
  doi = {10.3389/fnagi.2022.977034},
}
Figure