Statistical methods for dementia risk prediction and recommendations for future work: A systematic review

Area of Science:

• Neuroscience
• Biostatistics
• Epidemiology

Background:

• Numerous dementia risk prediction models have emerged recently.
• Methodological limitations in analytical tools may compromise dementia risk score reliability.

Purpose of the Study:

• To review and critically discuss methodologies used in dementia risk prediction models.
• To identify limitations in current analytical techniques for dementia risk assessment.

Main Methods:

• Systematic literature review from March 2014 to September 2018.
• Qualitative synthesis of 137 publications on dementia risk prediction models.
• Critical discussion of analytical techniques including machine learning, logistic regression, and Cox regression.

Main Results:

• Machine learning, logistic regression, and Cox regression are the most common methodologies.
• Identified three key methodological weaknesses: single data source reliance, inadequate statistical assumption verification, and lack of validation.

Conclusions:

• Over-reliance on single data sources limits generalizability.
• Thorough verification and adherence to statistical assumptions for logistic and Cox regression are essential.
• Enhanced validation strategies and diverse datasets are recommended for robust dementia risk prediction.