Temporal effects of empirical round-up of serum creatinine on the accuracy of estimated kidney function after critical illness

This study aimed to evaluate the temporal effect of a serum creatinine (SCr) correction intervention in critically ill patients with daily muscle wasting. A total of 5,355 critical care days in 574 patients with creatinine clearance (Clcr) and glomerular filtration rate (GFR) measured and estimated simultaneously were included. The primary endpoint was the difference in accuracy over time of the estimated Clcr/GFR relative to the measured Clcr. The CG equation was used to estimate Clcr, and the MDRD and CKD-EPI-equations were used to estimate GFR. Estimated Clcr(round-up)/GFR(round-up) indicates the estimated Clcr/GFR calculated using the rounded up value if SCr was <0.6 mg/dL. Bias was analyzed using Bland-Altman analysis, correlation using Spearman’s rank correlation coefficient, and accuracy using percentage within 30% of the measured Clcr (P30). The CKD-EPI equation showed a large underestimation bias in the early period, whereas the CG and MDRD equations indicated large overestimation biases in the later period. Round-up correction increased the underestimation bias in the early period of the CG(round-up) and MDRD(round-up) equations, but decreased the overestimation bias in the later period. The limits of agreements for the CG(round-up) and MDRD(round-up) equations were narrower, although not consistently acceptable. The correlations were stronger for the CG(round-up) and MDRD(round-up) equations. Accuracy did not improve with the corrective intervention. Conclusion: SCr round-up correction increased the underestimation of kidney function in the early phase, but mitigated overestimation in the later phase. The limits of agreement remained unacceptable, and the accuracy did not improve.