Referee Review

This paper reports an empirical qualitative study of decision making concerning the allocation of resources with regard to new drug treatments. The paper is original in two ways: it is the first empirical study focusing on committees and committee members that are involved in making such decisions. Second, it presents six domains of priority setting. Although some of these have been identified from non-empirical studies, not all have been. The proposal that these are the six key areas is novel.

I think therefore that this is sufficiently original for publication in the BMJ. It is also clearly written.

The methodological approach is using the qualitative method of grounded theory. This is a suitable approach given the aims of this study. One of the key issues for such methodology is that of reliability. The paper raises this issue but does not give much information. In particular it is unclear to what extent the two investigators in addition to the primary analyst agreed or disagreed with the primary analyst in the initial coding of the raw data; and how disagreements were then handled. I think that a little more should be said about this process since it relates to the question of the extent to which the six categories were ones that the authors had already identified, and the extent to which they genuinely emerged from the data. I think that this issue should also be brought up in the discussion of potential limitations.

A second issue that needs to be raised in the discussion, if not in the methods is that of "data saturation". When applying grounded theory the idea is to identify the issues that arise, and then to classify those issues. That is what these authors have done. However, the question arises as to whether there are sufficient numbers of participants. It is usual to tackle that issue by continuing to recruit participants until one can be fairly confident that no more key issues will emerge. The evidence for this is that the final interviews do not add any issues that have not already been identified. In the present study the number of participants is limited by there being just two relevant committees with a limited number of people on each committee. In this situation the authors should I think comment on whether new issues or themes were continuing to emerge; and the absence of being able to recruit until data saturation should be raised as a limitation in the discussion.

I think that it would be worth a table listing the six domains with a very brief description of each, since this is what the average reader will want to take away from this paper.

Dr T. Hope