Enhancing dementia risk prediction with heart rate and machine learning in the Canadian Longitudinal Study on Aging

Comprehensive Summary

Alaka et al. investigated whether the addition of resting heart rate, an easily measurable cardiovascular marker, would improve the prediction of dementia when added to the CAIDE model using machine learning techniques. Data from 18,013 participants of the Canadian Longitudinal Study on Aging were analyzed in order to predict the 3-year risk for dementia by using random forest and support vector machine algorithms for risk modeling. The risk was modeled based on CAIDE predictors, such as age, education, blood pressure, BMI, cholesterol, physical activity, and APOE status, to which RHR was added as a new variable. The performance of both models was assessed in terms of AUC, Matthews correlation coefficient, sensitivity, and specificity, as well as the Brier score, in both training and test datasets. With the addition of RHR, the precision of the risk modeling improved modestly but significantly. In random forest models, AUC increased from 0.65 to 0.67, and MCC increased from 0.29 to 0.32. In SVM models, there was a similar performance increase of 2-3%. Individuals with a resting heart rate of over 90 bpm have a 1.6-fold higher risk of dementia within three years compared with individuals with a lower resting heart rate. Both algorithms had higher sensitivity but lower specificity, therefore indicating their improved ability at correctly identifying individuals at risk compared to controls. The study has shown that the inclusion of RHR strengthens the predictive validity of dementia risk using data that is non-invasive and easily accessible. Though the discrimination of the model was moderate, its simplicity and accessibility are worthy of note in large-scale, community-based screening, pending further validation.

Outcomes and Implications

This is an important research because it enhances an established tool for dementia risk, using an inexpensive and widely available vital sign. Through the application of cardiovascular and cognitive health markers, it could help identify candidates for preventive treatment. Clinically, the CAIDE-RHR model can be used for performing dementia screening at primary care or public health level due to the low cost of testing on a large scale. It allows for early intervention by alerting individuals at risk using their regular check-up data. However, this study requires more research across diverse populations, with longer follow-up periods, before it can be applied into standard clinical practice.