Observational cohort of renal replacement therapy patients at a district general hospital ICU: case mix and outcomes

Introduction

Previous studies have shown the mortality of ICU patients requiring renal replacement therapy (RRT) to be high at 62.8% [1], and underpredicted by APACHE II scoring [2]. The ICNARC Case Mix Programme gives mortality of 59.5% for 2003 to 2004. We aimed to review our patients to see how we compare.

Methods

We prospectively collected data on all RRT episodes on the ICU from 2005 to 2007: patient age, APACHE II score in the first 24 hours, primary indication for RRT, RIFLE classification at start of RRT, ICU and hospital length of stay (LOS), mortality at 30, 60, and 90 days, and recovery of renal function at 30 days.

Results

We admitted 1,557 patients over the 3-year study period, of which 18% were elective. A total of 210 patients received RRT (data were available on 208). The median age was 66 years, 56% were male, and mean APACHE II score was 25.4. Our hospital mortality for all patients receiving RRT was 50.5%. One hundred and seventy-six patients had RRT for acute kidney injury with a hospital mortality of 51.1% (23 class Risk patients, 70 class Injury, 83 class Failure). No statistical difference in hospital mortality, ICU LOS, or renal function recovery existed by RIFLE class. By APACHE II score, the standardised mortality ratio was 0.98 (1.1 in 2005, 0.97 in 2006, and 0.86 in 2007). Main indications for filtration were: acidaemia 51%, oliguria 9%, uraemia 13%, fluid overload 7%, sepsis 6%, and hyperkalaemia <1%. The average ICU LOS was 9 days for hospital survivors (IQR 5 to 21 days) and 4 days in nonsurvivors (IQR 2 to 8 days). Only three hospital survivors were not alive at 90 days. Three out of 82 followed-up patients still required dialysis at 30 days.

Conclusion

Our hospital mortality compares favourably with other published work [1, 3]. We found the APACHE II score to predict mortality accurately, in contrast to published work showing underprediction [2]. This may represent a better outcome in our cohort. We were unable to demonstrate a correlation between the RIFLE score at initiation of RRT and hospital mortality. This could be due to small numbers, or to an equivalence of outcome for the RIFLE classes once RRT is established.