A systematic exposure‐wide framework leveraging machine learning to identify multidomain exposure factors and their joint influence on cognitive function: Evidence from a neurological cohort

Figures

• 

  FIGURE 1. Overview of analytic design.

• 

  FIGURE 2. Contribution rank of variables and domain components. (A) SHAP summary dot plot illustrates the impact on model output (SHAP values) for the whole population during the baseline period (N = 1142). The MoCA scores increase with the SHAP values of the features. Each dot represents a participant's SHAP value for a specific variable, with colors indicating the actual variable values: higher values are shown in purple, and lower values in orange. Dots are stacked vertically to reflect density. The figure displays SHAP values for concise model. (B) Explained variance of predicted MoCA scores was determined by XGBoost algorithms after fivefold cross‐validation in the left‐out group. These were trained exclusively on variables from each domain component (see Table S3 for variables included in each group) or the concise model (all selected 36 variables). (C) The longitudinal association of protective cognitive exposure scores with global cognitive function and memory function was analyzed using a linear mixed‐effects model. ADLs, activities of daily living; DM, demographic factors; MoCA, Montreal Cognitive Assessment; SES, socioeconomic status; SHAP, SHapley Additive exPlanation; XGBoost,…

• 

  FIGURE 3. The characteristic distribution of domain components by phenotype for the longitudinal cohort. (A) The radar chart shows domain components by phenotype. Each panel corresponds to a phenotype, with each bar showing the median value of a domain factor for all participants in that phenotype. The concentric circles represent the minimum, 25th, 50th, and 75th percentiles, and maximum in the whole sample, with the median shown in a darker color. The height and color of the bars represent the median level of each risk factor, positioned relative to the distribution in the whole population, ensuring a common scale across all phenotypes. (B) The comparison of variables between phenotypes. In all panels, the variables are standardized so that all means are scaled to 0 and SDs to 1. A value of 1 for the standardized variable (x‐axis) signifies that the mean value for the phenotype was 1 SD higher than the mean value for both phenotypes shown in the graph as a whole. SD, standard deviation.

• 

  FIGURE 4. Monte Carlo simulation of MoCA score changes with varying phenotypic frequencies in the total population. In this cohort, the actual distribution of the data in the given phenotypes and the legend for harms, benefits, and neutral effects using Monte Carlo simulation analysis are presented in A. Each simulation was conducted with 10,000 iterations using sampling with replacement. The MoCA scores analyzed by changing the phenotype distributions of the data are presented in B (Phenotype 1), C (Phenotype 2), D (Phenotype 3), and E (Phenotype 4). MoCA, Montreal Cognitive Assessment.