Prediction model audit finds where and for whom errors appear
Where Is a Prediction Model Miscalibrated, and for Whom? Local Audits in Outcome-Learned Pockets
Machine LearningArtificial Intelligence
Summary
Prediction models can sometimes make mistakes that only show up in certain groups or types of cases, and these errors can be hidden by overall averages. The authors created a method that finds exactly where a model over- or under-predicts and whether these errors differ between groups like race or gender. Their approach groups similar cases into "pockets" using how important different features are for predictions, then tests these pockets to find miscalibration with confidence. They showed it works well on simulated scenarios and real medical data, revealing hidden errors that other methods miss.
What this means in practice
- •For hospital data teams: Identify specific patient groups where clinical risk prediction models systematically over- or under-predict, to improve care decisions and fairness.
- •For machine learning engineers: Diagnose local prediction errors by subgroups in models to guide targeted model improvements and fairness auditing beyond overall performance metrics.
Authors
Bernardo Modenesi, Taoyi Chen, Sima Najafzadehkhoei, Lucia Wang, Ameya Diwan, Beatrice Knudsen
Abstract
Overall and group-level averages, and ranking metrics such as the AUC, can hide sets of observations for which a prediction model systematically over- or under-predicts. We present an audit that asks where a model is miscalibrated (in which sets of similar observations it over- or under-predicts on average) and for whom (whether its average prediction error differs between demographic groups, such as by race or sex, within the same set). It groups observations into pockets using random-forest similarity, which weighs each feature by its importance for predicting the outcome, with a ridge-regression grouping alongside. Pockets are learned on one sample and tested on another, with simultaneous confidence intervals that allow for correlation within units such as hospitals; the audit needs only features, outcomes and fixed predictions. In 17 simulation scenarios with an error planted for one group, the combined audit located it better than the strongest alternative grouping in 15, with false-alarm rates near or below 5%; with a local error planted for everyone, it won in 12 of 15 scenarios at the smaller error size and lost when the error was unrelated to the outcome. In public synthetic data from the care-management algorithm studied by Obermeyer et al., it found no detectable overall error but under-prediction of illness in the sickest pockets and over-prediction elsewhere, and distinguished Black patients' concentration where the score under-predicted for everyone from a Black-White gap within pockets. For APACHE IVa, an exploratory analysis located its previously reported over-prediction of ICU mortality in a mostly ventilated pocket and a low-risk pocket, with none detected in a moderate-risk one. Two chest X-ray classifiers showed offsetting local errors behind no detectable overall error. Code: https://github.com/modenesi/pocket-audit