In the course of developing a new technique for medical decision making, we used the data from an article that appeared in the Archives1 as the basis for a decision model. Our results did not agree with the published results1 for several of the management strategies. For example, our software (and decision trees drawn by hand) indicated that the expected cost of strategy A ("No tests or treatment") was $1905, not $2220 as published. Our decision-tree representation of strategy A (Figure 1) and one that explains the study1 results (Figure 2) are enclosed.
The mortality and cost of several other strategies differ from our calculations as well. In fact, our analysis, which evaluated a superset of the 24 strategies evaluated by Hillner et al1 showed that, with the article's "standard model" assumptions, the most cost-effective management strategy is not strategy I ("If real-time ultrasonography [RUS] suggests deep