Objective To establish and verify the prediction model of acute kidney injury (AKI) in emergency inpatients.

Methods The clinical data of 310 inpatients admitted to the Emergency Department of Zhongshan Hospital, Fudan University from October 2014 to September 2015 were collected. Univariate logistic regression was used to calculate the OR value and P value of each index for AKI occurrence. Independent risk factors were selected by stepwise multiple factor logistic regression method and included in the AKI prediction model. 206 patients were included in the training group, and 104 patients were included in the validation group. The discrimination (ROC curve), calibration degree (calibration curve), and clinical applicability (DCA curve) of the model in the two groups were calculated. Multivariate logistic regression was used to draw a nomogram to predict the risk of AKI in emergency inpatients.

Results The incidence rate of AKI was 36.1% (112/310). There were significant differences in albumin, urea, creatinine, serum magnesium, and other 7 indexes between the AKI and non-AKI groups (P < 0.05). Multivariate logistic regression showed that albumin, creatinine, serum magnesium, and blood glucose were independent risk factors for AKI. The prediction model of AKI was logitic (PAKI)=-0.113albumin+0.021creatinine+3.837serum magnesium + 0.108 blood glucose -2.878, and AKI is considered if the result is higher than 0.398. The AUC of the model in the training group and validation group were 0.790(95%CI 0.722-0.859, P < 0.001) and 0.752(95%CI 0.646-0.858, P < 0.001), respectively. Calibration curve U-test showed that the calibration degree of the training group and validation group were good (P>0.05). DCA curve showed that the model had good clinical applicability.

Conclusions The model including serum magnesium has a good predictive value in the incidence of AKI in emergency inpatients and has a good guiding role in clinical practice.