Predicting renal replacement therapy after cardiac valve surgery: external validation and comparison of two clinical scores.

Abstract

Acute kidney injury requiring renal replacement therapy is an infrequent but dangerous complication of cardiac surgery. Different clinical prediction scores have been developed and validated in American and European countries. Due to the lack of validation in China, the purpose of our study was to validate and compare two clinical scores externally in a Chinese valve surgery population. A retrospective analysis was conducted using data obtained from the database of all cardiac valve operations performed during a 5-year period (2010-2014) at a university hospital in Shanghai, China. The primary outcome was defined as the need for renal replacement therapy after cardiac surgery. For evaluation of the scores, discrimination and calibration were measured. After surgery, 52 (1.6%) patients received renal replacement therapy. Patients with different Cleveland scores were found to have significantly different incidences of renal replacement therapy of 0.01, 2.12 and 11.5%. Discrimination of both models was only fair, with values for the area under the receiving operator characteristics curve of 0.68 (95% confidence interval 0.61-0.75) and 0.68 (95% confidence interval 0.61-0.76), respectively. Calibration of the Cleveland score was excellent (P= 0.81) with validity at low score levels; in contrast, calibration of the simplified renal index score was poor (P< 0.001). Both models provided a certain clinical significance, allowing one to discrimination between patients with higher or lower risks of the postoperative need of renal replacement therapy within the present study population. However, it was not suitable for estimating the real incidence of the need for postoperative renal replacement therapy with sufficient precision in the study sample. Direct clinical use in our centre should be considered cautiously.