Related Experiment Video
Updated: Jul 29, 2026
06:35
Basics of Multivariate Analysis in Neuroimaging Data
Published on: July 24, 2010
17.3K
Red-flagging multimorbidity clusters for Alzheimer's disease risk using explainable machine learning: Evidence from a
Tursun Alkam1, Ebrahim Tarshizi1, Andrew H Van Benschoten1
1Master's Program of Applied Artificial Intelligence, University of San Diego, San Diego, CA, USA.
Journal of Alzheimer'S Disease Reports
|November 3, 2025
Summary
Emergency department visits can identify Alzheimer's disease risk early. Specific conditions like urinary tract infections and depression, combined with age, significantly increase Alzheimer's risk.
Area of Science:
- Gerontology
- Neurology
- Data Science
Background:
- Emergency department (ED) visits offer valuable diagnostic data for early identification of Alzheimer's disease (AD) risk.
- This data can potentially flag at-risk individuals long before overt cognitive symptoms manifest.
Purpose of the Study:
- To investigate the interaction between age and multimorbidity in predicting AD risk.
- To evaluate the effectiveness of explainable machine learning in enhancing risk stratification using national ED data.
Main Methods:
- Analysis of 554,985 ED visits (2010-2014 National Emergency Department Sample) for adults aged 60 and older.
- Utilized ICD-9-CM codes to identify AD and 17 chronic conditions.
- Employed logistic regression, decision trees, random forest, and XGBoost for predictive modeling, with SHAP for interpretability.
Main Results:
- Independent predictors of increased AD odds included urinary-tract infection (UTI), depression, hypothyroidism, and anemia.
- A dose-response relationship was observed: each "red-flag" condition increased AD risk by 74%, with all four tripling the risk.
- Age demonstrated a steep gradient, with significantly higher AD odds in older age bands.
Conclusions:
- Routine ED diagnostic codes reveal an age-dependent association between multimorbidity clusters and AD.
- An interpretable XGBoost model accurately identifies high-risk patients, enabling practical real-time cognitive risk alerts in acute care.
