Effect of discussion and deliberation on the public's views of priority setting in health care: focus group studyBMJ 1999; 318 doi: https://doi.org/10.1136/bmj.318.7188.916 (Published 03 April 1999) Cite this as: BMJ 1999;318:916
- Paul Dolan, research fellow (, )
- Richard Cookson, research fellow,
- Brian Ferguson, senior research fellow
- Correspondence to: Dr P Dolan, Sheffield Health Economics Group, University of Sheffield S1 4DA
- Accepted 18 December 1998
Objective: To investigate the extent to which people change their views about priority setting in health care as a result of discussion and deliberation.
Design: A random sample of patients from two urban general practices was invited to attend two focus group meetings, a fortnight apart.
Setting: North Yorkshire Health Authority.
Subjects: 60 randomly chosen patients meeting in 10 ugroups of five to seven people.
Main outcome measures: Differences between people's views at the start of the first meeting and at the end of the second meeting, after they have had an opportunity for discussion and deliberation, measured by questionnaires at the start of the first meeting and the end of the second meeting.
Results: Respondents became more reticent about the role that their views should play in determining priorities and more sympathetic to the role that healthcare managers play. About a half of respondents initially wanted to give lower priority to smokers, heavy drinkers, and illegal drug users, but after discussion many no longer wished to discriminate against these people.
Conclusion: The public's views about setting priorities in health care are systematically different when they have been given an opportunity to discuss the issues. If the considered opinions of the general public are required, surveys that do not allow respondents time or opportunity for reflection may be of doubtful value.
Emphasis is being placed on involving the public in decisions about healthcare priorities
Different techniques used to obtain the public's views may give different results
people's views on setting priorities differ systematically when they have been given opportunity to discuss the issues
people become less willing to discriminate against smokers, heavy drinkers, and illegal drug users when they have had an opportunity to reflect on and discuss their views
If considered opinions are required, the value of surveys that do not allow time or opportunity for reflection may be in doubt
Since the government's white paper Local Voices, it has been the official policy of the NHS that the views of the public should play a greater role in setting healthcare priorities.1 The many different techniques that can be used to obtain the public's views have recently been classified according to whether respondents are given detailed information about the choices they face and whether they are able to engage in discussion and deliberation when arriving at their responses.2 The most common approach to has been the “uninformed and undeliberated” one, involving one-off surveys of the public. 3 4 This approach is relatively cheap and can generate large sample sizes, but it can be criticised, firstly, on the grounds that people may take time to form a view or opinion and, secondly, because views can be shaped appreciably by the way that survey questions are framed.5 Hence, more deliberative approaches such as “uninformed” group discussions and “informed” citizens' juries6 are increasingly being used. However, given the high costs of these approaches (one citizens' jury can cost as much as =A320 0007), there has been surprisingly little investigation of the extent to which people change their views as a result of being exposed to the arguments of others. One of the main aims of our study was to address this issue.
We aimed to recruit 10 groups of six people who would meet for about two hours on two occasions separated by a fortnight. This would allow comparisons to be drawn between people's initial views at the start of the first meeting and their considered views at the end of the second meeting, when they had had an opportunity to discuss and deliberate.
After ethics committee approval for the study was obtained, letters of invitation were sent out to 1000 people randomly chosen from two general practice lists in York. These practices were selected because their catchment areas were convenient and provided sufficient scope to generate a broadly representative sample.
Potential respondents were given a list of times and venues and asked to indicate which pair(s) of dates they were willing to attend on. Each was told that they would be paid =A330 for attending. They were asked to indicate their sex and to which of three age groups they belonged (18-34, 35-55, or 55-70 years). There were 207 (21%) positive replies, from which it was possible to invite at least one person from each of the six sex-age categories to attend each group. Altogether 72 people (12 from each category) were invited to attend.
Each discussion group was moderated by the same two researchers. They made sure that each group had a full discussion of each of the items on the agenda and that all respondents were given sufficient opportunity to air their views. Apart from this, the moderators' role was a relatively passive one. The discussion groups aimed to allow respondents time and opportunity to consider their own responses rather than to provide them with additional information upon which to make their judgments. Therefore, the moderators' main objective was to facilitate discussion between respondents. To allow external validation, all group discussions were tape recorded and subsequently transcribed.
To facilitate discussion at the beginning of the first meeting, and to focus attention on the central issues, respondents were initially asked to discuss the question, “How would you set priorities in health care?” To obtain people's spontaneous views, we asked respondents to fill in an initial questionnaire without discussion with anyone. All questionnaires in this study were developed by the research team and piloted on their colleagues. Once the questionnaires were completed, a discussion about the questions took place. For the remainder of the first meeting, respondents were asked to arrive at group decisions, after discussion, on several hypothetical questions about setting priorities between groups of patients on the basis of differing quality or length of life, or both.
At the beginning of the second meeting, respondents were asked to rank four patients (who differed in a number of important respects, including the extent to which they could benefit from treatment) in order of priority. Then, on a group basis, they were asked to make three pairwise comparisons between groups of patients; these had been chosen to look at access and deprivation issues and to assess whether a large benefit to a few people is seen differently from a small benefit to many people. Finally, to obtain their views after deliberation, respondents were given a second questionnaire in which they were asked to answer again three of the questions they had answered in the initial questionnaire.
A five point scale—”much more,” “more,” “the same,” “less,” and “much less”—was used to determine responses in the initial and final questionnaires. For ease of analysis and exposition, the first two and last two categories were combined. Wilcoxon sign rank test was used to compare responses to the two questionnaires, and the chosen level of significance was <0.01.
Sixty people attended the meetings (83% of those invited). Table 1 describes the characteristics of respondents. table 2 presents the answers to the first question about how much involvement different groups should have in decisions about priority setting. Initially, most respondents wanted more involvement for the general public and doctors and nurses, but less involvement for managers and politicians. They were divided in their views on pressure groups. In the final questionnaire, however, the proportions who wanted more public involvement and less involvement for managers fell while views about doctors and nurses, politicians, and pressure groups were largely unchanged. Taking responses to all five groups as a whole, 52% of responses were unchanged; in 40% there was a shift of one category on the five point scale and in 8% a change of two categories. Only four respondents did not change their response to any of the five groups.
In both questionnaires around a third of respondents (17 (28%) in the first questionnaire and 21 (35%) in the second) wanted a lower priority to be given to people who are deemed largely responsible for their own illnesses. Although the overall views of the group did not change (40 (67%) in the first questionnaire and 39 (65%) in the final questionnaire gave these people the same priority as others), the final responses of 18 respondents differed from their initial ones. Nine switched from giving everyone the same priority to giving lower priority to people responsible for their own illness and six moved in the opposite direction. The remaining three respondents wanted initially to give higher priority to people responsible for their own illness, but in the final questionnaire, two wanted to give these people the same priority and one wanted to give them lower priority.
table 3 shows the changes in response to the final item in the questionnaire. This question listed 21 prioritisation factors which were considered by the research group to represent all of the personal characteristics to which people might, in principle, wish to give more or less healthcare priority. Respondents were asked to indicate how far they wanted to discriminate either in favour of or against each of the 21 factors. In both the initial and final responses, people generally wished to give greater priority to the more vulnerable members of society (for example, old people, young people, and disabled people) and lower priority to those whose illness may be partly caused by their lifestyle (for example, drinking, smoking, or taking illegal drugs). However, the key finding was that there was a general trend towards giving the same priority to a particular factor after discussion and reflection, particularly the three lifestyle factors, for which the final responses of the group were significantly different from their initial ones.
Across all 21 factors, there were 105 changes from a lower or higher priority to the same priority between the initial and final responses, and only 17 changes from the same priority to a lower or higher one. Only 22% of responses to individual factors changed between the initial and final questionnaires, but nearly three quarters of these changes were related to the first nine factors (table 3). For these factors, the response was unchanged in 63%, changed by one category in 32%, and changed by two categories in 5%. The number of respondents who gave the same priority to all factors rose from six in the initial questionnaire to 11 in the final one. The six people who initially gave the same priority to all factors were the only respondents who did not change any of their responses to this question in the final questionnaire.
Reassuringly, responses to the initial questionnaire were not affected by the group that respondents were placed in. Idiosyncratic features of the groups themselves did not seem to have any systematic effect on changes in responses: the respondents who changed their views, and the direction of change, were randomly distributed across the groups. The views that respondents held across the range of factors were unrelated to most of the background characteristics listed in table 1. The only significant differences have intuitive appeal: smokers were less willing to discriminate against those who are largely responsible for their own illness (in the final questionnaire, only 20% of the smokers wanted to give lower priority to these people, compared with 40% of the non-smokers) and also less willing to discriminate against those who drink heavily or who smoke (13% of smokers wanted to give lower priority to drinkers and smokers; for non-smokers the figures were 44% and 38%, respectively). Changes in views did not seem to be systematically related to any of the background characteristics listed in table 1.
These results suggest that the public's views about setting priorities in health care are systematically different when they have been given an opportunity to discuss the issues. About half of the respondents initially wanted to give lower priority to smokers, heavy drinkers, and illegal drug users, and a similar number wanted to give higher priority to children, disabled people, and elderly people. All of these results except the one regarding elderly people are consistent with other studies,8 but it is difficult to tell from these data why people wanted to discriminate in favour of elderly people when many other studies found the opposite.9 After reflection and discussion, many people no longer wished to discriminate against smokers, heavy drinkers, and illegal drug users. Equally, however, fewer people wanted to discriminate in favour of elderly people in the final questionnaire.
These results are consistent with the intuitive belief that we often find it harder to give definite answers to questions to which we have given considerable thought. It seems that people initially feel that the public should have more say in setting priorities in health care and that managers should have less.10 But when people discuss complex issues related to setting these priorities, they realise that things may not be straightforward and, as a result, are more reticent about the role that their views should play and more sympathetic towards the part that managers play.11 Likewise, after considering the various background characteristics that they could use to prioritise in favour of or against, people are more reluctant to make clearcut decisions—not because they think that choices in health care do not have to be made, but because they recognise that the choices are complicated ones.
There is some evidence that surveys, which largely elicit people's initial reactions, and focus groups, which aim to elicit more considered opinions, generate similar results. 8 10 The findings reported here suggest otherwise. This leaves a puzzle for future research, but if the within-subject results reported here are found to be generalisable, and if the considered opinions of the general public are required, then doubt is cast on the value of surveys that do not allow respondents the time or opportunity to reflect on their responses.
The study was designed and conducted in collaboration with Professor Mark Baker, Sue Baughan, and Rachel Johns at North Yorkshire Health Authority and Professor Alan Williams at the University of York. We thank them for their input. The authors are indebted to Kerry Atkinson, Vanda King, and Helen Parkinson for painstaking transcription of the group discussions.
Contributors: PD coordinated the project. RC and PD conducted the focus group discussions. BF provided access to the general practice lists and attended some discussion groups. All authors shared equally in the design of the research reported in this paper. PD acts as guarantor.
Funding The project was funded by Northern and Yorkshire Region Research and Development Office.
Conflict of interest None declared.