resources required to do CEA or CBAs on a routine basis, and to develop a range of estimates of the resources used to do published CEAS or CBAs. To a large degree, these goals were realized. Because of the survey's design, however, no firm conclusions are possible.

The scope of the survey was limited in the sense that the process used to select the sample population was restricted and arbitrary since the criteria used to identify the "quality" CEAS and CBAs were highly subjective. The survey sample was limited even further because its focus was restricted to CEAS and CBAs done only on health care topics. The survey sample may have been further distorted by the characteristics of the types of issues selected by researchers. Possibly, the CEA and CBA studies done to date have been "easy" ones. Analyses of health care topics may have been directed at areas that have had the most or best data available or that were the most highly visible in the public eye. On the other hand, a counterargument could be made. Since the field is relatively new, perhaps more work was required to develop the methods, more analysts were needed to perform quality work, and in essence, added effort was needed to establish the groundwork that other analysts can use. Whichever the case, it is difficult to know what effects these variables will have on the sample results.

The survey was broadly focused in that no attempt was made to differentiate between the various types of analyses. For example, no attempt was made to group the studies according to their technological focus (i.e., diagnosis v. therapy, procedure v. drug, or systems-based technology). Likewise, no attempt was made to determine if some analyses were considered more complex or sophisticated, or if some had more "value" to the research community or to the policy process, than others. Although these variables are very important, it was beyond the scope of this survey exercise to investigate them to the degree necessary to form estimates or conclusions.

At best, the survey results suggest a lack of consensus on most aspects of funding or staffing resources required for CEAs or CBAs. A good example of this lack of agreement are the estimates of what it might cost to perform a single CEA or CBA. In large part, however, the many reservations that the respondents included regarding the effect of data problems and the complexity of the issue(s) being examined on the cost of the study would explain the wide range of estimates provided. Although there were few surprises regarding the types of professionals or support staff required to do CEAS or CBAs, there was an in

teresting change in views in the responses received for the questions related to the types of professionals actually used to perform specific analyses and those related to the types of professionals recommended to staff the hypothetical research team. All but one actual study included a physician as part of the research group, but only seven included an economist. The hypothetical research teams, however, leaned more heavily toward the inclusion of economists. This apparent shift toward economists may not indicate any real change in the perceived need for physicians, though, because several respondents indicated a need for physicians and scientists to serve as consultants to the hypothetical research group. Thus, some of the respondents to the survey listed physicians in another category. The shift to more economists, however, was not explained.

Data needs and problems appeared to be a significant factor for all the respondents. The large range of cost-per-study estimates was directly tied to the availability and quality of data. Information needs were cited much more often than factors such as complexity of the problem being studied and stage of development of the technology as variables that will affect the cost of a given study. To date, however, the respondents seem to feel that many of the data range from being easy to difficult, but not impossible, to obtain. The data also were very inexpensive to obtain and were available from public sources.

The results of OTA's survey may simply provide a look at the types of resources analysts have used to perform health-care-related CEAS and CBAs and at the types of resources they feel are necessary to perform them on a routine basis. It is interesting to note the differences between the resources used and the resources that the respondents felt were required to do CEA and CBA on a regular basis. Several professions that were not used a great deal by the actual study teams in the past show up quite frequently on the "ideal" research team list. Disciplines such as economics, statistics, engineering, computer analysis, and computer programing were not frequently used in the actual studies, yet were cited several times by the respondents as being needed for the hypothetical research group. As the health care issues become more complex and as information needs become more demanding, the range of expertise needed to do analyses will broaden, and the costs of performing CEAS and CBAs will increase. These resource costs may become an important factor to be considered by governments and other institutions that must decide whether or not to do CEAS or CBAS or how many they can do in a given time period.