Abstract: Abstract Objective: This study was designed to establish a clinical prediction score for the prognosis of acute kidney injury (AKI) in hospitalized patients. Methods: A total of 363 hospitalized patients were enrolled in Huashan hospital from January 2011 to March 2012. A prospective cohort study was performed on 261 patients of AKI in the development set, and 102 patients of AKI were selected for the validation set. Multivariate logistic regression analysis was applied to identify the risk factors of AKI and derivate a new prediction score based on the odds ratio . Thus the prediction curve for the mortality of AKI 90 days was established. The ROC curves and Hosmer-Lemeshow goodness-of-fit chi-squared test were used to assess the accuracy and efficacy of the scoring system. Results: (1)Five variables were identified as the independent risk factors of AKI: the total number of complications, vasopressor support (dopamine), mechanical ventilation, blood urea nitrogen(BUN) and serum prealbumin. (2) The total score was calculated for each patient. The incidence of AKI was 16.8%in the low-risk group(≤4 points), 48.0% in the moderate-risk groups(5-10 points), 76.0% in the high-risk group(11-16 points) and 100% in the very-high-risk group(17-30 points). The area under the ROC curve was 0.801(p<0.001). (3) Good discriminative power was found in the validation population, and the risk score was strongly correlated with AKI (χ2=4.149 , p=0.657). Conclusion: The scoring system could predict the mortality of hospital-acquired AKI patients accurately. Clinical application of this score may support decision making for protective kidney interventions and improve the prognosis of AKI patients in hospital.

Key words: Acute kidney injury, Prognosis, Prediction score