Data-driven approaches to executive function performance and structure in aging: Integrating person-centered analyses...

Area of Science:

• Neuroscience
• Psychology
• Gerontology

Background:

• Executive function (EF) is often studied using variable-centered approaches.
• Person-centered analytics offer unique insights into subgroups by examining EF performance and structure concurrently.
• Machine learning can identify risk predictors for these subgroups.

Purpose of the Study:

• To detect distinct subgroups (classes) of EF performance and structure in nondemented aging.
• To test multiple risk predictors for discriminating these identified subgroups.

Main Methods:

• Utilized data from the Victoria Longitudinal Study.
• Employed a classification sample (n=778) and a prediction subsample (n=570).
• Assessed three EF dimensions (inhibition, updating, shifting) using eight neuropsychological measures and fifteen predictors across five domains (genetic, functional, lifestyle, mobility, demographic).

Main Results:

• Identified two distinct classes: Class 1 (lower EF performance, unidimensional structure) and Class 2 (higher EF performance, multidimensional structure).
• Class 2 was predicted by younger age, increased cognitive activity, higher education, lower BMI, lower pulse pressure, female sex, better balance, and more physical activity.
• Machine learning analyses indicated that the higher-performing, multidimensional class was associated with multiple brain health-protective factors.

Conclusions:

• Data-driven modeling identified distinct EF aging classes differing in performance level and structure.
• The higher-performing, multidimensional EF class is associated with numerous brain health-related protective factors.