Decision curves of “treat all,” “treat none,” eIPM_Pre, and eIPM_Post approaches to grade 3 to 4 ICAHT. By visualizing the net benefit (proportion of true positives, “weighted” proportion of false positives vs false negatives) of various strategies across a range of probability thresholds, decision curves enable comparison of the clinical utility (net benefit) of these approaches. Across all probability thresholds, a clinical approach based on either eIPM confers greater clinical net benefit than a “treat all” or “treat none” approach. At higher probability thresholds >20% (ie, in situations when the costs of false positives are higher than false negatives), a clinical approach based on eIPM_Post confers greater clinical net benefit than that based on eIPM_Pre.