This research by Velmovitsky et al. aims to study how a recommendation engine that uses both rule-based logic and machine learning can support university student mental health care through a stepped care model. The researchers analyzed longitudinal data from incoming undergraduate students collected between 2018 and 2023, which included standardized self-report measures such as the PHQ-9 and GAD-7, as well as family and personal mental health history, childhood struggles, and substance use behaviors. They trained XGBoost machine learning models with 10-fold cross-validation to predict students at high risk of developing clinically significant anxiety or depression, then combined these predictions with a rules-based algorithm to recommend appropriate levels of intervention. The models achieved high sensitivity and strong negative predictive values, though specificity and positive predictive values were lower, indicating a higher number of false positives. Analysis revealed that factors such as childhood trauma, family history of mental illness, and functional impairment were among the strongest predictors of future mental health risk. The authors note that the system’s false positives generally led to students without significant mental health risk being recommended lower-intensity interventions, which are considered acceptable within stepped-care models, and emphasize the importance of ensuring equitable access and validation across a more diverse population.