AI Insight
Researchers developed an audit method to identify specific subsets of patients where prediction models systematically over- or under-predict outcomes, and whether these errors differ by demographic groups. The method groups similar observations into "pockets" using machine learning techniques, then tests whether prediction errors vary within these pockets and across demographic categories like race or sex. Testing on multiple medical datasets including care-management algorithms and ICU mortality predictions revealed local miscalibrations that were hidden by overall accuracy metrics, such as under-prediction for the sickest patients and demographic disparities within specific patient groups.
Why it matters
This approach could help identify hidden biases and errors in clinical prediction tools used for resource allocation and medical decision-making, potentially reducing healthcare disparities. By pinpointing exactly where and for whom algorithms fail, healthcare systems can make targeted corrections rather than relying on overall performance metrics that may mask systematic errors affecting vulnerable populations.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
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
Source: Where Is a Prediction Model Miscalibrated, and for Whom? Local Audits in Outcome-Learned Pockets