Predicting incident dementia in community-dwelling older adults using primary and secondary care data from electronic health records

2024

Predicting Dementia Risk in Older Adults Using Health Records

Sample size: 144113 publication 10 minutes Evidence: moderate

Author Information

Author(s): Georgiev Konstantin, Wang Yiqing, Conkie Andrew, Sinclair Annie, Christodoulou Vyron, Seyedzadeh Saleh, Price Malcolm, Wales Ann, Mills Nicholas L, Shenkin Susan D, McPeake Joanne, Fleuriot Jacques D, Anand Atul

Primary Institution: University of Edinburgh

Hypothesis

Can machine learning models using electronic health records predict the risk of future dementia in older adults?

Conclusion

The study found that machine learning models can moderately predict dementia risk in older adults over a 13-year period.

Supporting Evidence

8% of participants developed dementia over 13 years.
The data-driven model achieved better precision-recall scores than the clinically supervised model.
Age, deprivation, and frailty were key predictors of dementia risk.
40% of the highest risk group were under 80 years old.
Model performance improved with longer prediction windows.
High-risk stratification could improve targeting of healthcare resources.
Routine health data can be used to identify individuals at risk of dementia.
Machine learning models can help in early detection of dementia.

Takeaway

Researchers used health records to predict who might get dementia in the future, helping doctors find people who need care sooner.

Methodology

The study used a longitudinal retrospective cohort design analyzing electronic health records to develop machine learning models for predicting dementia.