Judgment Analysis of Prioritization Decisions within a Dialysis Program in One United Kingdom Region

Background . Some previous research on rationed clinical services has confused the conceptual differences underpinning prioritization decisions on the one hand and assessments of individual need on the other. The balance of the clinical and nonclinical drivers of these decisions can be different. Our objective was to study the basis and extent of variation among nephrologists in one NHS region in their views concerning prioritization for dialysis. Design and methods . In a clinical judgment analysis, multiple regression analysis was used to express the impact of clinical and nonclinical cues on nephrologists’ decisions to offer dialysis and attribute priority to 50 “paper patients.” Cues were selected for the decision-making models using stepwise (backward) elimination of variables. Further “policy” models for priority were derived by forcing in the doctors’ views about the capacity of dialysis to extend life expectancy or improve its quality. Results . Comparison of “propensity to offer” and “prioritization” decision models showed a modest degree of correspondence. Among the nonrenal cues, the patient’s mental state made the single greatest contribution to the priority decision models (mean contribution to R2 = 12.1, with temporary or permanent confusional states in patients changing the priority [1-50] by an average of 15 rank places). Although patient age significantly influenced the decision models of half of the doctors, the beta-coefficients were very modest, suggesting a change in rank order of no more than one place. There was a significant improvement in the overall explained variance (R2 )of the models when varying perceptions of the capacity of dialysis to improve the quality or extend the duration of the patient’s life were forced into the model. Although, in general, temporary or permanent confusion in the patient downgraded the priority for dialysis by between 10 and 20 places, this tendency was largely unchanged when the doctors’ perceptions of benefit were forced into the priority model. Among renal cues, the presence of uremic symptoms had the greatest impact on priority (mean contribution to R2 = 49.1, mean beta-coefficient -17.1), whereas the presence of other comorbid disease had relatively little effect. Conclusions . When forced to rank patients, the nonrenal factor that had the most significant bearing on perceived priority for dialysis was the patient’s mental state. However, the impact of the patient’s mental state on priority did not appear to be driven by its influence on the doctors’ perceptions of how dialysis might improve quality of life.