Abstract
BACKGROUND: This study aimed to explore the performance of renal resistive index (RRI), semiquantitative power Doppler ultrasound (PDU) score and renal venous Doppler waveform (RVDW) pattern in predicting acute kidney injury (AKI) in critically ill patients and establish prediction models. METHODS: This prospective observational study included 234 critically ill patients. Renal ultrasound was measured within 24 h after intensive care unit admission. The main outcome was the highest AKI stage within 5 days after admission according to the Kidney Disease Improving Global Outcomes criteria. RESULTS: Patients in the AKI stage 3 group had significantly higher RRI, RVDW pattern and lower PDU score (p < 0.05). Only lactate, urine volume, serum creatinine (SCr) on admission, PDU score and RVDW pattern were statistically significant predictors (p < 0.05). Model 1 based on these five variables (area under the curve [AUC] = 0.938, 95% confidence interval [CI] 0.899-0.965, p < 0.05) showed the best performance in predicting AKI stage 3, and difference in AUC between it and the clinical model including lactate, urine volume and SCr (AUC = 0.901, 95% CI 0.855-0.936, p < 0.05) was statistically significant (z statistic = 2.224, p = 0.0261). The optimal cut-off point for a nomogram based on Model 1 was ≤127.67 (sensitivity: 95.8%, specificity: 82.3%, Youden's index: 0.781). CONCLUSIONS: The nomogram model including SCr, urine volume, lactate, PDU score and RVDW pattern upon admission exhibited a significantly stronger capability for AKI stage 3 than each single indicator and clinical model including SCr, urine volume and lactate.