Risk prediction of acute kidney injury in paraquat poisoning patients / 中华急诊医学杂志

ABSTRACT

Objective:To establish a risk prediction model of acute kidney injury in paraquat (PQ) poisoning patients.Methods:A retrospective observational cohort of adult patients with acute PQ poisoning between September 10, 2010 and January 16, 2020 from the Emergency Department of West China Hospital, Sichuan University were conducted. Data on demographics, clinical records, and laboratory results were collected from electronic medical record. The patients were divided into the AKI group and the non-AKI group according to whether AKI occurred during hospitalization. The patients were randomly divided into the training and validation groups (7:3). Multivariate logistic regression analysis was used to screen the independent risk factors of AKI and the nomogram was used to establish a prediction model. Receiver operating characteristic (ROC) curve and calibration curve were used to evaluate the differentiation and calibration of the prediction model. Decision curve analysis (DCA) was used to evaluate the clinical validity of the prediction model.Results:A total of 718 patients were included in this study. AKI occurred in 323 (45%) patients in hospital and 378 (52.6%) patients died. The mortality rate of the AKI group was higher than that of the non-AKI group (72.8% vs. 36.2%, P < 0.05). Multivariate logistic regression analysis showed that the time from poisoning to treatment ( OR=1.018, 95% CI:1.006-1.030), white blood cell count ( OR=1.128, 95% CI: 1.084-1.173), aspartate aminotransferase ( OR=1.017, 95% CI:1.006-1.027), cystatin C ( OR=516.753, 95% CI: 99.337-2688.172), and PQ concentration ( OR=1.064, 95% CI:1.044-1.085) in blood on admission were independent risk factors of AKI in patients with PQ poisoning ( P<0.01). The area under the ROC curve was 0.943 (95% CI: 0.923-0.962) in the training cohort, and the sensitivity and specificity were 82.4% and 93.6%, respectively. The calibration curve showed optimal agreement between prediction by nomogram and actual observation. Decision curve and clinical impact curve analysis indicated that the nomogram conferred high clinical net benefit. Conclusions:The time from poisoning to treatment, white blood cell count, aspartate aminotransferase, cystatin C, and PQ concentration in blood on admission were independent risk factors of AKI. The predictive model based on the above indicators has high sensitivity and specificity in evaluating AKI after PQ poisoning. Whether this prediction model can be applied to other PQ poisoning patients needs to be further expanded for verification.