Key result
ECG models trained on selective revascularization failed in untested patients, while OMI-trained models retained discrimination for 30-day major cardiovascular events.
Why the study?
The biases of selective labels from clinician testing decisions are well known, but it is less understood what happens when a target diagnosis includes cases that a machine learning model's inputs cannot detect.
Does training ECG machine learning models on selectively tested patients using poorly recoverable targets (like unstable angina or broad ACS) impair risk discrimination for 30-day MACE in untested emergency department patients compared to training on ECG-recoverable targets (like OMI or STEMI)?
Observational (n=61,844)
Yes
Does training ECG machine learning models on selectively tested patients using poorly recoverable targets (like unstable angina or broad ACS) impair risk discrimination for 30-day MACE in untested emergency department patients compared to training on ECG-recoverable targets (like OMI or STEMI)?
ECG machine learning models trained to predict broad ACS or revascularization in selectively catheterized patients fail and even reverse risk predictions in untested patients due to collider bias from ECG-undetectable cases like unstable angina, whereas training specifically on ECG-recoverable targets (STEMI, NSTEMI, OMI) preserves clinical discrimination.
Answers are grounded in this paper and cite their sources.
Revascularization-trained ECG models should not deploy in untested patients; hypothesis-generating for OMI-specific targets in ACS risk stratification.
Medical machine learning models often learn from diagnoses observed only in patients clinicians chose to test. The biases of such selective labels are well known, but less understood is what happens when a diagnosis includes cases the model's inputs cannot detect. Because testing decisions draw on information the model does not see, these cases can appear predictable among tested patients, producing relationships that fail, and can even reverse, among the untested. We formalize this mechanism and show in simulations that it worsens as testing becomes more selective or undetectable cases more common, while models trained only on detectable cases remain informative. We empirically study electrocardiogram (ECG)-based prediction of acute coronary syndrome (ACS), which is confirmed only by invasive testing and whose subtypes differ in how clearly they appear on the ECG. Three ECG foundation models fine-tuned on revascularization, our ACS proxy, performed well among catheterized patients (AUROC 0.740-0.770) but worse than chance for 30-day major adverse cardiovascular events among untested patients (AUROC 0.381-0.478). Models trained on ECG-visible subtypes retained discrimination (AUROC 0.680-0.709), whereas a model trained on unstable angina, which can occur without ECG changes, was inversely related to risk (AUROC 0.310). Under selective labels, effective learning therefore requires prediction targets that can be inferred from the model's inputs and evaluation in the intended deployment population, which often extends beyond the patients who were tested.
No takes yet. Share an insight, caveat, or question.
Schubert et al. (2026) conducted an observational in Suspected acute coronary syndrome (n=61,844). ECG machine learning models trained on ECG-recoverable ACS subtypes (OMI, STEMI, NSTEMI) vs. ECG models trained on selective revascularization labels or unrecoverable subtypes (unstable angina) was evaluated on 30-day major adverse cardiovascular events (MACE) discrimination (AUROC) among untested patients (AUROC 0.709 (OMI model) vs AUROC 0.310 (UA model), 95% CI 0.702-0.716). ECG models trained on selective revascularization failed in untested patients, while OMI-trained models retained discrimination for 30-day major cardiovascular events.
Synapse has enriched 10 related papers. Consider them for comparative context: