Methods: A decision model (decision tree in TreeAge) was constructed in collaboration with pediatric hematologist experts, and a detailed questionnaire was presented to a nationally representative committee of 25 pediatric hematologists. Enough data was elicited to specify a full model for each respondent.
Results: 19 (76%) experts responded; based on preferred initial treatment, there were 12 treaters and 7 non-treaters. Out of 60 variables, there was statistically significant differences on 3 variables: the relative risk reduction of ICH by the three treatments (Anti-D: treaters 64% vs non-treaters 27% (P .012, min Bayes Factor against null hypothesis 13); IVIG: treaters 76% vs non-treaters 33% (P .0018, BF 69); steroids: treaters 71% vs non-treaters 28% (P .001, BF 120)). A non-inferiority study would require 11,000 patients to change the opinion of treaters to non-treaters. There was agreement on values: Median number needed to treat (NNT) to avoid ICH was 100; median NNT to avoid treatment side effects was < 1. The derived NNT of treatment for treaters was 10,000; for non-treaters, 30,000. For treaters, although anti-D was the reported preferred treatment, the steroids strategy was cost saving and anti-D conferred an incremental cost effectiveness ratio of $10.5 million.
Conclusion: Decision-analytically na´ve physicians provided enough data to specify complete individual models. The disagreement among these investigators concerns probabilities and not values. These decision-analytically results helped these investigators to clarify their goals on a formal basis