Public health policy makers increasingly rely on CEAs to help them sift through the many choices confronting them. When different CEA models give different answers to the same question, confidence in their usefulness may suffer, since it is unclear to what extent the disagreement arises from uncertainty about the underlying evidence, which affects all decision making approaches, or from the modeling methods used by different modelers. Understanding the reasons for differences among models is therefore an important first step in building the public’s confidence that CEA can provide objective and informative insights into the consequences of health policy choices.
The Institute of Medicine’s (IOM’s) National Cancer Policy Board (NCPB) convened the workshop, “Economic Models ofColorectal Cancer Screening in Average-Risk Adults” on January 26–27, 2004, to explore the reasons for differences among leading CEA models of CRC screening. Participants discussed the results of a collaborative pre-workshop exercise undertaken by five research teams that have developed and maintained comprehensive models of CRC screening in average-risk adults. The purpose of the exercise was to provide workshop participants with insights into each model’s structure and assumptions and possible explanations for differences in their published analyses. Workshop participants also examined the current state of knowledge on key inputs to the models with a view toward identifying areas where further research may be warranted.
In keeping with the purpose of IOM workshops, this summary of its proceedings presents the individual perspectives and research of people who made presentations at the workshop and of many other experts who participated. This summary does not contain consensus recommendations, nor does it represent a consensus opinion of the IOM’s NCPB. Nor is it intended as a guide for conducting or using cost-effectiveness analyses in CRC screening decisions.
It is particularly important to recognize that the purpose of the workshop was not to consider the relative merits of different strategies for CRC screening, or to suggest which CRC screening strategy is best. It was solely to consider the commonalities and differences among the CEA models bearing on the subject. The demand for more certain guidance from models by those who recommend or pay for screening strategies, while clearly a motivating force behind the workshop, was not its focus. More certain guidance may result in the future as modelers continue to grapple with and explain the differences in their findings.
The idea for collaboration among research teams that maintain published models of CRC screening grew out of a recent review by Michael Pignone and colleagues for the U.S. Preventive Health Services Task Force (Pignone et al., 2002). They systematically reviewed seven published CEAs of periodic CRC screening in average-risk adults. That review identified several aspects of model structure and underlying assumptions which, taken together, might account for most of the differences in cost-effectiveness rankings of CRC screening strategies. However, each model involves dozens of assumptions, and the reviewers concluded that the published reports provided insufficient information to determine which assumptions or aspects of model design were most important in explaining differences in conclusions across models.