Background: Early diagnosis of acute kidney injury (AKI) is a major challenge in the intensive care unit (ICU). The\nAKIpredictor is a set of machine-learning-based prediction models for AKI using routinely collected patient information,\nand accessible online. In order to evaluate its clinical value, the AKIpredictor was compared to physiciansâ?? predictions.\nMethods: Prospective observational study in five ICUs of a tertiary academic center. Critically ill adults without endstage\nrenal disease or AKI upon admission were considered for enrollment. Using structured questionnaires, physicians\nwere asked upon admission, on the first morning, and after 24 h to predict the development of AKI stages 2 or 3 (AKI-\n23) during the first week of ICU stay. Discrimination, calibration, and net benefit of physiciansâ?? predictions were\ncompared against the ones by the AKIpredictor.\nResults: Two hundred fifty-two patients were included, 30 (12%) developed AKI-23. In the cohort of patients with\npredictions by physicians and AKIpredictor, the performance of physicians and AKIpredictor were respectively upon ICU\nadmission, area under the receiver operating characteristic curve (AUROC) 0.80 [0.69-0.92] versus 0.75 [0.62-0.88] (n =\n120, P = 0.25) with net benefit in ranges 0-26% versus 0-74%; on the first morning, AUROC 0.94 [0.89-0.98] versus 0.89\n[0.82-0.97] (n = 187, P = 0.27) with main net benefit in ranges 0-10% versus 0-48%; after 24 h, AUROC 0.95 [0.89-1.00]\nversus 0.89 [0.79-0.99] (n = 89, P = 0.09) with main net benefit in ranges 0-67% versus 0-50%.\nConclusions: The machine-learning-based AKIpredictor achieved similar discriminative performance as physicians for\nprediction of AKI-23, and higher net benefit overall, because physicians overestimated the risk of AKI. This suggests an\nadded value of the systematic risk stratification by the AKIpredictor to physiciansâ?? predictions, in particular to select\nhigh-risk patients or reduce false positives in studies evaluating new and potentially harmful therapies. Due to the low\nevent rate, future studies are needed to validate these findings.
Loading....