Prediction of outcome in acute renal failure.

In an attempt to predict outcome in acute renal failure (ARF) we have utilized multiple logistic regression to analyze clinical data from 151 patients with ARF seen over a 15-month period. Recovery of renal function occurred in 60% of patients with a 58% survival. Our analysis demonstrated sepsis, respiratory failure, and oliguria to be the major predictors of nonrecovery of renal function. A logistic equation was generated for prediction of outcome and was validated in a second independent group of patients with ARF. Prediction of outcome could be achieved with a sensitivity of 75% and a specificity of 80%. Maximum sensitivity (100%) was associated with a 17% specificity, while maximum specificity (98%) yielded a sensitivity of 20%.