Stewardship or clinical freedom? Variations in dialysis decision making

Background. It is generally agreed that acceptance criteria for dialysis have varied and changed over time and that implicit rationing, to some extent forced on clinicians by limited capacity, has been widely practised. Our objective was to study the basis and extent of variation in dialysis decision making among nephrologists in one NHS region. Design and methods. In a clinical judgement analysis, linear regression models were employed to reflect the impact of clinical and non-clinical cues on nephrologists' decisions to offer dialysis to 60 'paper patients' under current capacity constraints and under an assumption of no capacity limit. A short questionnaire was also completed by eight nephrologists to elicit their expressed decision drivers, which were subsequently compared with those tacitly derived from the appraisal of the 60 clinical vignettes. Results. Doctors showed substantial variation in their propensity to offer dialysis and in their perceptions of the benefits of dialysis. Even for the five patients where the discordance in propensity to offer dialysis was least, the range in perceived gain in life expectancy was from 24 to 264 months (mean 91 months). The decision models had relatively good explanatory power with an average r(2) Of 0.67 (0.39-0.90) and 0.70 (0.47-0.95) for decisions made under current capacity constraints and under an assumption of no limit capacity respectively. Surprisingly, for most doctors, the patient's age had very little impact on dialysis decisions but the magnitude of the beta-coefficients for the patient's mental state (mean -30.7) was of a similar order of magnitude to the coefficient for the principal 'renal' drivers (e.g. the mean coefficient for uraemic symptomatology under current capacity constraints was 47.7). The influence of other non-renal factors on the doctor's likelihood to offer dialysis was largely independent of the capacity assumption. A comparison of the doctor's stated decision drivers with those tacitly derived from their decision models showed only modest correlation Conclusions. The extent to which doctors vary in their propensity to offer dialysis is substantial. Very few non-clinical cues appear to influence the decision to offer dialysis. The most important non-renal factor in determining dialysis decisions was the patient's mental state.