Dementia Risk Prediction Models in Older Adults

Dementia risk prediction models aim to identify individuals at elevated likelihood of developing cognitive decline and dementia, enabling targeted prevention and early intervention. These models integrate demographic factors (such as age, sex and education), clinical variables (including comorbidities and medication use), lifestyle measures (for example physical activity, smoking and diet) and, increasingly, genetic and biomarker data. Traditional approaches deploy statistical risk scores derived from longitudinal cohorts, whereas contemporary methods harness machine learning to optimise feature selection and predictive accuracy. Model performance is commonly evaluated by discrimination (how well the model distinguishes future cases from non-cases) and calibration (how closely predicted probabilities match observed outcomes). Key challenges remain: external validation across diverse populations, balance between model complexity and clinical feasibility, and transparent interpretation of algorithmic outputs. Despite these hurdles, validated risk tools hold global significance for public health planning, individualised counselling and stratification in clinical trials, and they form the foundation of emerging Brain Health Services that prioritise primary prevention.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Dementia Risk Prediction Models in Older Adults publication trend

The graph below shows the total number of articles in dementia risk prediction models in older adults across all publications each year (not limited to Nature Index journals).

Technical terms

Logistic regression: A statistical technique for modelling the probability of a binary outcome based on one or more predictor variables.

Recursive feature elimination (RFE): An iterative method that removes the least significant features to improve model performance and reduce overfitting.

SHapley Additive exPlanations (SHAP): A model-agnostic framework that assigns each feature an importance value for a particular prediction, enhancing interpretability.

Area under the curve (AUC): A metric of a model’s discriminative ability, representing the probability that a randomly selected case is ranked higher than a non-case.

Cox LASSO regression: A survival analysis model that applies a least absolute shrinkage and selection operator penalty to select relevant predictors and prevent overfitting.

References

1. A novel integrated logistic regression model enhanced with recursive feature elimination and explainable artificial intelligence for dementia prediction. Healthcare Analytics (2024).
2. CogDrisk, ANU-ADRI, CAIDE, and LIBRA Risk Scores for Estimating Dementia Risk. JAMA Network Open (2023).
3. Development and validation of a dementia risk score in the UK Biobank and Whitehall II cohorts. BMJ Mental Health (2023).
4. Modifiable risk factors for dementia and dementia risk profiling. A user manual for Brain Health Services—part 2 of 6. Alzheimer's Research & Therapy (2021).
5. Current Developments in Dementia Risk Prediction Modelling: An Updated Systematic Review. PLOS ONE (2015).