Back
Acute kidney injury (AKI) is a major cause of morbidity and mortality, particularly among patients with sepsis. To improve the outcome, early prediction tools that can aid clinical decision-making are imperative. Nomograms are visual predictive tools used to calculate risk of outcomes and evaluate prognosis. An AKI risk prediction nomogram in sepsis patients was previously developed by Yang and co-workers using retrospective data from the public critical care database MIMIC III v.14. To enhance the clinical applicability of the nomogram, further validation using external data is essential. Here, We further validated the nomogram among patients admitted at a Philippine tertiary hospital as independent external validation cohort.
Prospective clinical and laboratory data were collected from 244 patients admitted at Cebu Doctors’ University Hospital, Cebu City, Philippines, between April 2022 and March 2023. Patients with sepsis identified with the sepsis-3 diagnostic criteria upon admission were included. Receiver operating characteristic curve (ROC) analysis was performed to assess the clinical utility of the nomogram. Statistical analyses further determined the association of various parameters with AKI.
The area under the ROC curve >0.5 (0.648) indicated that the nomogram can provide a modest discrimination between positive and negative patient outcomes and is thus clinically useful. ROC analysis further determined the following: sensitivity, 65.68%; specificity, 64%; positive predictive value, 80.43%; and overall accuracy, 65.16%. Using the nomogram, 56.55% of the patients have total risk points >28.5 cut-off value compared with the 69.26% actual AKI patients diagnosed based on KDIGO criteria. The calibration curve shows that the nomogram‘s risk estimation performance generally agrees with actual AKI frequency. Decision curve analysis indicates that the nomogram has a better predictive effect than using only one predictor. Additionally, our use of prospective data allowed us to mine patient information for potential novel predictors. Statistical analysis indicated that levels of BUN and bicarbonate were among the variables showing strong association with AKI, hence could be added among the predictors to further enhance nomogram performance in future studies.
Successful prediction of AKI risk in septic critically-ill patients in a Philippine hospital further validated the AKI risk prediction nomogram. Inclusion of more patients and external data from other hospitals will further verify and improve the nomogram’s clinical applicability.