Prediction of acute renal failure by 'bedside formula' in medical and surgical intensive care patients

Background: Prediction of which intensive care unit (ICU) patients are likely to develop acute renal failure (ARF) would be useful. However, scoring systems such as APACHE have been disappointing in this regard. We previously developed a bedside formula to predict ARF using only 3 parameters: serum albumin, urine osmolality, and presence of sepsis. Methods: We prospectively evaluated 115 consecutive medical ICU (MICU) patients, comparing the bedside formula to APACHE II AND APACHE III as predictors of ARF or death and looking at nutritional parameters such as iron binding capacity, triceps skin fold, mid-arm circumference, and total lymphocyte count. We then evaluated 123 additional consecutive MICU and 98 consecutive surgical ICU (SICU) patients, comparing the bedside formula to APACHE II. Results: The bedside formula was consistently more accurate than APACHE II in predicting ARF or in-hospital death in MICU patients. However, in SICU neither formula predicted ARF, and APACHE II predicted in-hospital death slightly better. No nutritional parameter other than albumin correlated with ARF. Conclusion: The bedside formula appears superior to APACHE II in predicting ARF or death in MICU but not SICU. This suggests that these two ICU populations are different.