Objective

To determine predictors of ICU mortality in end-stage renal disease (ESRD) patients treated with haemodialysis or peritoneal dialysis and requiring ICU admission.

Design and setting

Retrospective/prospective cohort study in an adult 10-bed medical ICU in a university hospital.

Mesasurements and main results

Over a 4-years period, out of 104 ICU admissions, 92 ESRD patients were studied. The etiologic diagnosis of ESRD was diabetes mellitus (n = 23), glomerulonephritis (n = 19), hypertension (n = 13), polycystic kidney disease (n = 11), pyelonephritis and obstructive uropathy (n = 6), interstitial nephritis (n = 5), congenital abnormalities (n = 5), others or unknown (n = 10). The prior mean duration of dialysis was 68 ± 83 months, 86 patients were on hemodialysis and 18 on peritoneal dialysis. The admission diagnosis was sepsis (n = 32), cardiac failure/fluid overload (n = 19), hemorrhage (n = 12), postoperative (n = 10), mesenteric ischemia and peripheral arterial thrombosis (n = 7), stroke (n = 6), cardiac arrest (n = 6), hyperkalemia (n = 5), others (n = 7). The mean length of stay in ICU was 6 ± 9 days. The overall ICU mortality was 29.8% (31/104). The survival rate for patients requiring mechanical ventilation was significantly less than for those not mechanically ventilated 13/36 (36%) vs 8/68 (12%), respectively (P < 0.0001). There was no significant difference between ICU survivors and nonsurvivors according to prior duration of dialysis, type of dialysis, and etiology of ESRD. In this target population, the mean SAPS II and APACHE II were 50.3 ± 20.9 and 24.9 ± 9.1, respectively. The discrimination as determined by the area under the receiver operating characteristic curve was not different between SAPS II and APACHE II: 0.859 vs 0.878, respectively (P = 0.62). For both models, the Hosmer–Lemeshow goodness-of-fit test revealed a poor performance. The H test result was P = 0.013 for SAPS II and P = 0.0006 for APACHE II. The C test result was P = 0.008 for SAPS II, and P = 0.005 for APACHE II. The ICU mortality among ESRD patients trend to be higher than that of other ICU admitted patients during the same period (21.8%), P = 0.063.

Conclusions

SAPS II and APACHE II were not well calibrated in ESRD patients. These models probably need to be customized to accurately predict mortality and analyse quality of care or performance among ICUs when applied to this target population.