We are thrilled to announce the publication of a significant research paper, Predicting Incident Dementia in Community-Dwelling Older Adults Using Primary and Secondary Care Data from Electronic Health Records, co-authored by our team at Red Star and an incredible multidisciplinary group of researchers.
This study evaluates the diagnostic capabilities of Machine Learning models in predicting dementia diagnoses up to 13 years before the appearance of memory-related symptoms. Utilizing data from over 144,000 older adults registered in NHS Lothian, who were dementia-free prior to 2009, the study harnesses Health Data Research UK (HDR UK) phenotype coding to predict dementia diagnoses at 5, 10, and 13-year intervals.
Key Highlights:
- An XGBoost ensemble model and a clinically-supervised alternative were trained using 171 health variables and 22 clinically-curated variables
- The clinically-supervised model achieved specificity of 0.88 and sensitivity of 0.55 at 13 years, comparable to the data-driven model.
- The models identified one in three individuals at risk of dementia within 13 years, at the highest risk decile, significantly higher than the baseline population risk of one in 13.
- Top predictors included age, deprivation, eFI score (excluding memory deficits), smoking, BMI, hypertension, hearing loss, and stroke.
The study suggests that while the current diagnostic quality is moderate, these models could be pivotal in guiding further testing and risk factor surveillance for those at high risk, potentially streamlining brain imaging for emerging immunotherapies in the future.
This research has been an extraordinary collaborative effort, and we extend our deepest gratitude to everyone involved:
- The Data Lab – Innovation Centre
- Digital Health & Care Innovation Centre
- The University of Edinburgh:
- DataLoch (Data-Driven Innovation Initiative)
- The Sir Jules Thorn Charitable Trust
We are proud to contribute to this important field of study and invite you to explore the full article in Brain Communications here.
Stay tuned for more updates as we continue to explore the potential of AI in healthcare.