Summary
Models that predict a person’s age using health data sometimes show differences when applied to sick patients, hinting at disease effects. The authors developed a way to figure out which training data samples most influence these disease-related changes in the model’s predictions. They found that removing certain influential healthy-looking samples actually increased the model’s ability to highlight disease differences for conditions like type-2 diabetes and multiple sclerosis. This shows that some hidden health issues in the training data can affect how well the model works. The authors also created a tool for others to use their method.
Normative age modelsCohen's dTraining data attributionInfluence functionDisease risk predictionMetabolomic ageBrain ageType-2 diabetesMultiple sclerosisUK Biobank
Abstract
Normative age models are trained to predict chronological age in a nominally healthy cohort. Applied to patients, they deviate, and the gap between predicted and chronological age is read as disease risk. Here, we attribute the disease-related effect size of the age gap directly to individual training samples, rather than using a prediction-level loss as the attribution target. For Cohen's $d$, the resulting closed-form influence functional, validated against leave-one-out retraining, ranks training samples by their effect on held-out case-control separation. Across four diseases and two biomarker modalities in UK Biobank, removing the 10% most influential training samples raises held-out disease-related effect size in every seed. It more than doubles the metabolomic-age effect for type-2 diabetes and raises the brain-age effect for multiple sclerosis by roughly a third. Random removal leaves effect size flat even at 50% removal, confirming the gain comes from which samples are removed, not how many. Flagged subjects carry subclinical cardiometabolic burden that diagnosis-based exclusion misses, on markers the model never sees. For type-2 diabetes, where the method gains most, the marker recovered is HbA1c, the standard measure of blood sugar control. We release pyinfluence, our influence-function package, for reproducibility and reuse.