Fig. 1: Workflow of the study.

This workflow diagram illustrates the pre-processing and analysis steps of the study. Green box 1 details the variable selection step using natural language processing models. The variables initially filtered from the UK Biobank data dictionary are further selected using natural language processing models based on cosine similarity scores. Green Box 2 outlines the causal network analysis steps - Mixed Graphical Models and Fast Causal Inference are used to construct networks of dementia from each imputed dataset. The results are then pooled into a single comprehensive network of dementia.