Risk factors to predict renal failure and death in the medical intensive care unit

The ability to predict outcomes on admission criteria has important implications, both prognostically and for assessing interventions on comparable groups. Use of severity of disease scoring systems such as the APACHE II score for predicting mortality has become widespread. There is no comparable formula for acute renal failure. We prospectively evaluated 115 consecutive admissions to the medical intensive care unit to define risk for renal failure from admission data and to assess the impact of admission hypoalbuminemia levels on outcome. Diagnosis, age, serum creatinine and albumin levels, urinary electrolyte concentrations and osmolality, daily serum creatinine levels, and urine output were recorded admission APACHE II score was calculated. Admission hypoalbuminemia (57% of patients) was associated with both acute renal failure and death (odds ratios, 16, 19 and 8.06, respectively). The Glasgow coma score, distinguished between patients in whom acute renal failure developed and in those it did not. Low urine osmolality (<400 mOsm/kg) was the most significant factor in predicting mortality (odds ratio, 9.87). Mortality was lowest in the normal albumin group (2%), intermediate in the low albumin/no renal failure group (12%), and highest in the low albumin/acute renal failure group (53%). The APACHE II score was accurate in 3 of 14 deaths in the hypoalbuminemic population and in the one normal albumin patient who died. We conclude that at admission, hypoalbuminemia, urinary hypo-osmolality and abnormal creatinine levels are predictive of acute renal failure and death, diagnosis, and mental status impact on the risk for acute renal failure. APACHE II lacks predictive value in hypoalbuminemic patients.