Reply to ‘Interpreting the clinical utility and generalizability of a multitask perioperative prediction model’
摘要
We thank Wo and Zhang for their comments on our multitask prediction model for postoperative complications, including postoperative acute kidney injury, respiratory failure, and in-hospital mortality. Here, we responded by mapping specific threshold probabilities to specific clinical actions in the decision curve analyses, quantifying the net benefit to express clinical gains as the number of additional true events detected per 1000 patients, and providing detailed calibration metrics with recalibration guidance.