Machine Learning Highlights Key Predictors of Cognitive Performance
The research assessed how variables like age and diet related to performance on a task that tested attentional control.
Complete the form below to unlock access to ALL audio articles.
A study published in The Journal of Nutrition has used machine-learning techniques to determine which health and lifestyle factors best predict cognitive performance across the adult lifespan. The research assessed how variables such as age, blood pressure, body mass index (BMI), diet, and physical activity relate to performance on a task that tests attentional control and response speed.
The task, known as the flanker test, requires individuals to identify the direction of a central arrow while ignoring distracting flanking arrows. It is a commonly used measure of attention and inhibitory control.
Age and cardiovascular metrics emerged as dominant predictors
Data were collected from 374 adults aged 19 to 82, who were evaluated on a wide range of measures including BMI, dietary habits, blood pressure, and physical activity levels. Participants also completed the flanker test, which measured both accuracy and response time.
Among the machine learning models evaluated, those that most accurately predicted performance highlighted age as the strongest single predictor. Diastolic blood pressure, BMI, and systolic blood pressure followed in influence. While dietary quality, measured by adherence to the Healthy Eating Index, was less predictive than age or cardiovascular metrics, it was associated with improved task performance.
Lifestyle patterns may mitigate certain risk factors
The study noted that while high BMI and elevated blood pressure were generally linked with poorer outcomes, higher physical activity and better dietary adherence could partially offset these effects. Physical activity in particular emerged as a moderate predictor, and its interaction with other lifestyle factors suggested a complex relationship with cognitive performance.
Machine learning provides analytical depth
Traditional statistical methods often examine single variables in isolation. In contrast, machine learning allows for simultaneous evaluation of multiple, potentially interacting variables. This approach revealed nuanced associations among lifestyle factors that may otherwise remain obscured.
The researchers tested multiple algorithms to identify those best able to predict cognitive performance based on the collected data. Models were validated using standard machine-learning practices to assess robustness.
The research was supported by the Personalized Nutrition Initiative and the National Center for Supercomputing Applications at the University of Illinois Urbana-Champaign.
Reference: Verma S, Holthaus TA, Martell S, et al. Predicting cognitive outcome through nutrition and health markers using supervised machine learning. J Nutr. 2025. doi: 10.1016/j.tjnut.2025.05.003
This content includes text that has been generated with the assistance of AI. Technology Networks' AI policy can be found here.
