%0 Conference Paper
%A Jockwitz, Christiane
%A Mendl-Heinisch, Camilla
%A Miller, Tatiana
%A Dellani, Paulo R.
%A Caspers, Svenja
%T Prediction of individual cognitive test scores from brain and non-brain data across the adult lifespan
%M FZJ-2025-05032
%D 2025
%X Predicting cognitive decline in aging remains a challenging but important topic. Existing results are heterogeneous, potentially due to the non-linear nature of both, cognitive decline and the factors that influence it. We here aimed to systematically examine the predictability of cognitive abilities based on brain and non-brain data across five decades of the adult lifespan in the large German National Cohort (NAKO; N = 23,863; 25 to 75 years). Brain summary statistics (e.g. total grey matter), health (e.g. body-mass-index) and demographic (i.e. age, sex, education) data were used to predict four cognitive scores using a machine learning (ML; repeated nested cross-validation; four regression algorithms) approach.Current results emphasize that demographics tend to outperform brain and health factors in predicting cognitive abilities in a large sample spanning the whole adulthood, with better predictability for episodic memory and interference compared to verbal fluency and working memory. Contrary to the hypothesis of a worse prediction at older ages, prediction appeared to be similarly low in each decade. Hence, sample size seems to matter even more than sample homogeneity. Including a wide age range for reaching large sample sizes, though, could come at the cost of predicting a hidden age effect.
%B Aging and Cognition Conference
%C 7 May 2025 - 10 May 2025, Pavia (Italy)
Y2 7 May 2025 - 10 May 2025
M2 Pavia, Italy
%F PUB:(DE-HGF)6
%9 Conference Presentation
%U https://juser.fz-juelich.de/record/1048935