Summary: A new study applied machine learning to identify the health and lifestyle factors most closely linked to cognitive performance across adulthood. In a sample of 374 adults aged 19–82, the strongest predictors of performance on a focus-and-speed attention test were age, blood pressure (especially diastolic), and body mass index (BMI). Diet and physical activity showed smaller but meaningful associations, often moderating the negative impacts of high BMI or elevated blood pressure.
By analyzing multiple health indicators simultaneously, the research offers a clearer, data-driven view of which modifiable and non-modifiable factors support attention, processing speed, and inhibitory control as people age.
Key Facts:
- Top predictors: Age, diastolic blood pressure, and BMI most strongly influenced cognitive performance on a flanker-style attention task.
- Diet and exercise: Higher diet quality and greater physical activity were associated with better focus and reaction speed, though their predictive strength was smaller than age and blood pressure.
- Machine learning advantage: Supervised learning models revealed nuanced relationships among multiple variables that conventional statistics might overlook.
Source: University of Illinois
Overview
Researchers from the University of Illinois used machine learning to determine which demographic, lifestyle, and health markers best predict a person’s ability to respond quickly and accurately on an attention task. The study focused on performance in a version of the Eriksen flanker task, which measures how well participants can focus on a central stimulus while ignoring distracting flankers.
Published in The Journal of Nutrition, the study examined how demographic factors, anthropometrics, blood pressure, dietary quality, and self-reported physical activity together relate to cognitive outcomes. Using supervised machine learning allowed the team to rank predictors by importance while accounting for interactions among variables.
“Machine learning lets us evaluate many factors at once and highlight which ones most closely align with cognitive performance,” said Naiman Khan, the study’s lead author and a professor of health and kinesiology at the University of Illinois Urbana-Champaign. Traditional statistical approaches struggle to untangle this level of complexity in a single analysis.
Data came from 374 adults (age range 19–82, 227 female). The dataset included age, BMI, systolic and diastolic blood pressure, several dietary indices (Healthy Eating Index, DASH, Mediterranean, and the MIND combination), self-reported physical activity, and flanker task reaction time and accuracy. The models were trained and tested using an 80:20 split and validated via cross-validation and hyperparameter tuning.
Among the machine learning methods tested—decision trees, random forest, AdaBoost, XGBoost, gradient boosting, and linear and regularized regressions—the random forest regressor performed best. Feature importance assessed by permutation ranked age as the highest contributor to reaction time, followed by diastolic blood pressure, BMI, systolic blood pressure, and the Healthy Eating Index. Ethnicity and sex had minimal predictive influence in this sample.
Although diet quality and physical activity were less influential than age and blood pressure, they were still associated with improved cognitive outcomes. The results suggest that healthier eating patterns and regular exercise may partially offset the cognitive risks associated with higher BMI or elevated blood pressure. Physical activity, in particular, appeared to interact with other lifestyle measures to influence processing speed and reaction time.
The authors note prior literature linking adherence to high-quality diets—such as DASH, Mediterranean, and the hybrid MIND diet—with better executive function and slower cognitive decline. Diets rich in antioxidants, omega-3 fatty acids, and essential vitamins have repeatedly been associated with superior cognitive health in older adults.
“This study shows how machine learning can add precision and nuance to nutritional neuroscience and brain health research,” Khan said. “By identifying the relative importance of multiple factors, these methods can help tailor preventive strategies and personalized interventions for people at different stages of life or with metabolic risk factors.”
Funding came from the Personalized Nutrition Initiative and the National Center for Supercomputing Applications at the University of Illinois. Khan is a registered dietitian and affiliate faculty in the Division of Nutritional Sciences, the Neuroscience Program, and the Beckman Institute for Advanced Science and Technology.
About this AI and brain health research news
Author: Diana Yates
Source: University of Illinois
Contact: Diana Yates – University of Illinois
Image: Image credited to Neuroscience News
Original Research: Open access. “Predicting cognitive outcome through nutrition and health markers using supervised machine learning” by Naiman Khan et al., Journal of Nutrition. DOI: 10.1016/j.tjnut.2025.05.003
Abstract
Predicting cognitive outcome through nutrition and health markers using supervised machine learning
Background
The use of machine learning in health research is expanding, but its application to predicting cognitive outcomes from diverse health and lifestyle markers remains limited.
Objectives
This study applied supervised machine learning to identify which health and behavioral factors most strongly contribute to cognitive performance, with the goal of informing personalized prevention and intervention strategies.
Methods
Researchers analyzed data from 374 adults aged 19–82 years (227 females) to predict reaction time on a modified Eriksen flanker task. Features included demographics, anthropometrics, dietary indices (Healthy Eating Index, DASH, Mediterranean, and MIND), self-reported physical activity, and systolic and diastolic blood pressures. The dataset was split 80:20 for training and testing. Multiple predictive algorithms were tuned and validated; feature importance was assessed using permutation importance, and model performance was evaluated using mean absolute error (MAE) and mean squared error (MSE).
Results
A random forest regressor delivered the best performance (training MAE: 0.66 ms; testing MAE: 0.78 ms). Age was the most influential predictor (importance score: 0.208), followed by diastolic blood pressure (0.169), BMI (0.079), systolic blood pressure (0.069), and Healthy Eating Index (0.048). Ethnicity and sex showed minimal predictive effect.
Conclusions
Age, blood pressure, and BMI are strongly associated with attention and processing speed measured by the flanker task, while diet quality and physical activity have subtler but meaningful associations. Machine learning offers a valuable approach for integrating multiple health markers to guide personalized strategies for preserving cognitive function across the lifespan.