Graft loss following renal transplantation in Australia: Is there a centre effect?

Abstract

Background. Assessment of centre variation in renal transplantation outcome provides an opportunity to examine differences in quality of care between centres. However, differences in outcome may represent differences in patient factors between centres and be biased by sampling variability and inadequate data ascertainment. Methods. Differences in 12-month graft survival in 1986 primary renal transplant adult recipients from 16 centres in Australia between 1993 and 1998 were examined. Fifteen recipient and donor factors known prior to transplantation were examined to determine factors independently predictive of graft survival. Differences between centres in these factors were examined. Unadjusted and multivariable adjusted outcomes for each centre were compared to the average for all centres. Multivariable hierarchical modelling was employed to account for potential bias due to sampling variability. Results. Factors predictive of reduced 12-month graft survival on multivariable analysis that were significantly different between centres were time on dialysis prior to transplantation, donor age, organ source, and number of human lymphocyte antigen mismatches. Unadjusted 12-month graft survival for all centres was 91.7% and ranged from 83.1 to 96.4%. Although two centres performed significantly lower than average, this poorer outcome was accounted for in one of these two centres after adjusting for factors shown to be independently predictive of outcome. However, multivariable hierarchical modelling failed to identify any centre as performing significantly different to average, with 12-month graft survival ranging from 89.2 to 92.2%. Outcome in patients excluded from the study due to inadequate data ascertainment was significantly worse than patients who were included. Conclusions. There was no evidence of centre variation after accounting for variation in risk factors predictive of poor outcome between centres, as well as potential bias due to sampling variability. Exclusion of patients due to inadequate data remains an important source of bias in estimating accurate outcomes. Appropriate analytical strategies and consideration of sources of bias are important for the valid identification of centres with poorer outcomes.