A machine learning approach to predict reversion from mild cognitive impairment to normal cognition: A population‐based cohort study. (31st December 2021)

• Record Type: Journal Article
• Authors: Sha, Feng; Bi, Jiefeng; Wei, Chang; Zhao, Ziyi; Li, Bingyu
• Is Part Of: Alzheimer's & dementia. Volume 17(2021)Supplement 10
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-12-31
• Subjects: Alzheimer's disease -- Periodicals; Alzheimer Disease -- Periodicals; Dementia -- Periodicals; Démence; Maladie d'Alzheimer; Périodique électronique (Descripteur de forme); Ressource Internet (Descripteur de forme); 616.83
• DOI: 10.1002/alz.057527
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: Predicting the probability of the reversion from mild cognitive impairment (MCI) to cognitively normal (CN) status can inform preventive treatments at individual, institutional and social level. This study aims to build a prediction model using a machine learning approach for reversion from MCI to CN status. Method: The study included 7, 422 participants above 65 years old with MCI from Chinese Longitudinal Health Longevity Survey (Figure 1). LightGBM was used to build a prediction model of 154 variables, including individual and household socioeconomic status, dietary/lifestyle, cardiometabolic, psychological factors and history of diseases. SHAP values were used to interpret the impact of the top 40 variables that contributed to the LightGBM model. Multivariable Cox regression with elastic net penalty was also conducted for comparison. Model performance was assessed according to a range of learning metrics including area under the receiver operating characteristic curve (AUC). Results: 1, 604 (21.6%) participants reversed from MCI to CN with a median follow‐up of 2.8 years. The top 40 features were presented in the figure 2. The concordance index of the LightGBM model was 0.71, which is much higher than the traditional multivariable Cox regression with elastic net penalty (0.66). Conclusion: The machine learning approach could develop a more accurate early identification of recovery of MCI patients. The predicting can guide the recovery process of MCI patients.