Multifactorial 10-Year Prior Diagnosis Prediction Model of Dementia.

Area of Science:

• Neuroscience and Gerontology
• Computational Biology and Bioinformatics

Background:

• Dementia is a progressive neurodegenerative disorder impacting the elderly, with no current cure or disease-modifying treatments.
• Brain changes associated with dementia can manifest a decade before clinical symptoms appear, highlighting the need for early prognosis research.

Purpose of the Study:

• To investigate a comprehensive multifactorial decision tree approach for predicting dementia.
• To identify predictive factors and potential prognostic subgroups for dementia over a 10-year timeframe.
• To address limitations in previous machine learning studies by incorporating a broader range of variables and a longer prediction window.

Main Methods:

• Utilized a broad decision tree multifactorial approach analyzing 75 variables.
• Included demographic, social, lifestyle, medical history, biochemical, physical examination, psychological, and health instrument data.
• Employed data from the Swedish National Study on Aging and Care (SNAC), including 726 subjects with 91 developing dementia within 10 years.

Main Results:

• Achieved an Area Under the Curve (AUC) of 0.745 and a Recall of 0.722 for 10-year dementia prognosis.
• Identified modifiable risk factors and physical strength as significant predictors across all age groups.
• Noted a scarcity of variables related to routinely used health instruments for dementia diagnosis within the selected factors.

Conclusions:

• A broad multifactorial approach can effectively predict dementia up to 10 years in advance.
• Modifiable lifestyle factors and physical strength are crucial for early dementia risk assessment.
• Future research should integrate more health instrument data for enhanced diagnostic and prognostic accuracy in dementia prediction.