Dementia risk analysis using temporal event modeling on a large real-world dataset.

Area of Science:

• Medical Informatics
• Neurology
• Psychiatry

Background:

• Dementia diagnosis relies on identifying risk factors.
• Electronic Health Records (EHR) offer vast real-world data for analysis.
• Discovering novel temporal risk factors can improve early dementia detection.

Purpose of the Study:

• To identify healthcare events and temporal trajectories preceding dementia diagnosis.
• To uncover novel and known risk factors for dementia using a data-driven approach.

Main Methods:

• Utilized a large real-world dataset of electronic health records (EHR).
• Employed a data-driven approach to identify temporally ordered healthcare code pairs and trajectories.
• Applied trajectory and clustering analysis to uncover patterns.

Main Results:

• Identified known dementia risk factors (e.g., Down syndrome, Parkinson's disease).
• Discovered novel risk factors including Brief psychotic disorder and Toxic effect of metals.
• Found strongest associations with mental health (Brief psychotic disorder, Dissociative disorders) and neurological conditions (Dystonia, Lumbar Puncture).

Conclusions:

• The study provides valuable insights into patient progression towards dementia.
• Identified temporal risk factors can improve the recognition of individuals at risk for dementia.
• This data-driven approach reveals previously unobvious risk factors and trajectories.