Abstract: Purpose： Analyze AKI patients who undertook renal replacement therapy (RRT) to screen the important indicators to decide the appropriate timing of RRT and its influence on prognosis. Method： The clinical data of 507 AKI patients were reviewed. 102 AKI patients who undertook RRT with complete medical records were further analyzed to reveal the relationship between clinical data and prognosis. Statistical methods were used to screen the important indicators of the appropriate timing of RRT. Results： TCO2 and urine volume were screened as predict factors of prognosis after the comparsion of the relationship of data at the onset of RRT and the prognosis. Further analysis revealed that MODS was also a powerful factor. The prediction model for AKI patients who undertook RRT was established after multiple variable Logistic regression analysis. The sensitivity and specifity of the model was respectively 78.8% and 78.0%, and the area under ROC was 0.809. The Hosmer-Lemeshow goodness-of-fit tests of the model was P=0.295. Conclusion： (1) Under the senario of the same disease and clinical background, the recovery of kidney function may benefit from initiating RRT before urine output＜600ml/d and/or TCO2＜22.2mmol/L. (2) As an independent risk factor, MODS indicates a worse prognosis; the determinant of improving clinical outcome of AKI is initiating RRT before MODS occurs.

Key words: Acute kidney injury, acute renal failure, renal replacement therapy, prognosis