Abstract
Emerging evidence suggests that minor changes in serum creatinine concentrations are associated with increased hospital mortality rates. However, whether serum creatinine concentration (SCr) on admission and its change are associated with an increased mortality rate in mechanically ventilated patients is not known. We have conducted an international, prospective, observational cohort study enrolling adult intensive care unit patients under mechanical ventilation (MV). Recursive partitioning was used to determine the values of SCr at the start of MV (SCr(0)) and the change in SCr ([Delta SCr] defined as the maximal difference between the value at start of MV [day 0] and the value on MV day 2 at 8: 00 AM) that best discriminate mortality. In-hospital mortality, adjusted by a proportional hazards model, was the primary outcome variable. A total of 2,807 patients were included; median age was 59 years and median Simplified Acute Physiology Score II was 44. All-cause in-hospital mortality was 44%. The variable that best discriminated outcome was a SCr(0) greater than 1.40 mg/dL (mortality, 57% vs. 36% for patients with SCr(0)<= 1.40 mg/dL, P < 0.001). Among patients with SCr(0) less than or equal to 1.40 mg/dL, Delta SCr greater than 0.31 discriminated mortality (56% vs. 34%, P < 0.001). In multivariate analysis, geographic area, advanced age, severity of illness, reason for MV, and cardiovascular and hepatic failure were also associated with mortality. Our study suggests that SCr(0) greater than 1.40 mg/dL and, in patients with low baseline SCr, a Delta SCr greater than 0.31 are predictors of in-hospital mortality in mechanically ventilated patients.