RESUMEN
BACKGROUND: It has been recently mathematically demonstrated that the percentage increase in serum creatinine (SCr) can delay acute kidney injury (AKI) diagnosis in patients with previous chronic kidney disease (CKD). Based on creatinine (Cr) kinetics, it was suggested a new AKI classification using absolute increase in SCr elevation over specified time periods. However, this classification has not been evaluated in clinical studies. METHODS: A prospective cohort study evaluated myocardial infarction patients during the first 7 days of hospital stay with daily SCr measurements. They were classified using Kidney Disease Improving Global Outcomes (KDIGO) and Cr kinetics systems. Both classifications were compared by net reclassification improvement (NRI) and area under the receiver operator characteristic (AuROC) curve regarding hospital mortality. RESULTS: A total of 584 patients were included, of which 34.1% had previous CKD. Patients had more AKI by KDIGO than by Cr kinetics criteria (25.7 versus 18.0%, P < 0.001) and 81 patients (13.9%) had different AKI severity classification. Patients with AKI by KDIGO criteria and non-AKI by Cr kinetics had higher hospital mortality rates than patients with non-AKI using both classifications [adjusted mortality odds ratios (ORs): 4.753; 95% confidence interval (CI): 1.119-9.023, P = 0.014]. In patients with previous CKD, NRI analysis was 6.2% favoring Cr kinetics criteria. However, there was no difference using the AuROC curve analysis. In patients with no previous CKD, NRI analysis was 33.0%, favoring KDIGO, and this was in accordance with a better AuROC curve (0.828 versus 0.664, P < 0.05). CONCLUSIONS: AKI classification proposed by a Cr kinetics model can be superior when diagnosing patients with previous CKD. However, KDIGO had a better performance in patients with no previous CKD.