Comparison of outcomes by modality for critically ill patients requiring renal replacement therapy: a single-centre cohort study adjusting for time-varying illness severity and modality exposure.

Prolonged intermittent renal replacement therapy (PIRRT) is a recently defined acute modality for critically ill patients, and in theory combines the superior detoxification and haemodynamic stability of continuous renal replacement therapy (CRRT) with the operational convenience and low cost of intermittent haemodialysis (iHD). We performed a retrospective cohort study for all critically ill adults treated with renal replacement therapy at our centre in Auckland, New Zealand from 1 January 2002 to 31 December 2008. The exposure of interest was modality (PIRRT, CRRT, iHD). Primary and secondary outcomes were patient mortality determined at hospital discharge and 90 days post renal replacement therapy inception, respectively. Co-variates included co-morbidity and baseline illness severity measured by Acute Physiology and Chronic Health Evaluation IV and Sepsis-Related Organ Failure Assessment (SOFA) and time-varying illness severity measured by daily SOFA scores. We used Marginal Structural Modelling to estimate mortality risk adjusting for both time-varying illness severity and modality exposure. A total of 146 patients with 633 treatment-days had sufficient data for modelling. With PIRRT as the reference, the adjusted hazard ratios for patient hospital mortality were 1.31 (0.60 to 2.90) for CRRT and 1.22 (0.21 to 2.29) for iHD. Corresponding estimates for mortality at 90 days were 0.96 (0.39 to 2.36) and 2.22 (0.49 to 10.11), respectively, reflecting the poorer longer-term prognosis of patients still on iHD at hospital discharge with delayed or non-recovery of acute kidney injury. Our study supports the recent increased use of PIRRT, which within limits can be regarded as safe and effective.