Machine learning to detect Alzheimer's disease with data on drugs and diagnoses

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Neurodegenerative Disease Prediction

Background:

• Integrating machine learning (ML) with electronic health records (EHRs) offers a promising avenue for the early detection of Alzheimer's disease (AD).
• Timely interventions for AD can be facilitated by accurate predictive models, improving patient outcomes.
• The study leverages ML to analyze clinical data for enhanced diagnostic sensitivity and specificity.

Purpose of the Study:

• To evaluate the effectiveness of ML in constructing a predictive model for AD up to three years prior to diagnosis.
• To identify crucial factors within ML models that serve as significant predictors of AD.
• To enhance diagnostic procedures' sensitivity and specificity using clinical data.

Main Methods:

• Stochastic Gradient Boosting, an ML technique, was utilized to identify AD-predictive diagnoses from primary healthcare data.
• The study analyzed clinical records from Region Stockholm, Sweden, between 2010 and 2022.
• Analyses were stratified by sex and age groups (41-69 years and over 69 years), excluding patients diagnosed with AD between 2010-2012.

Main Results:

• The ML model achieved robust performance, with Area Under the Curve (AUC) values ranging from 0.748 to 0.816 across different demographic groups.
• Sensitivity and specificity for AD prediction ranged from 0.73-0.79 and 0.66-0.79, respectively.
• Key predictors identified included medical observations, cognitive symptoms, antidepressant use, visit frequency, and vitamin B12/folic acid treatment, confirming known factors and revealing novel ones.

Conclusions:

• ML models applied to clinical data demonstrate significant potential for predicting AD with strong performance across diverse populations.
• The study confirmed established AD risk factors and identified novel predictors, offering valuable insights for future research.
• This ML-driven approach can enhance early AD detection and risk stratification, leading to timely interventions and improved patient care.