Interpreting the clinical utility and generalizability of a multitask perioperative prediction model
摘要
This Matters Arising comments on Yoon et al.’s multitask gradient boosting model for predicting acute kidney injury, postoperative respiratory failure and in-hospital mortality after non-cardiac surgery. We ask for clarification on the clinical decision context and threshold ranges in the decision curve analyses, the absolute impact of the model for rare outcomes, and calibration and potential recalibration in external cohorts to support implementation.