Rapid responses are electronic comments to the editor. They enable our users
to debate issues raised in articles published on bmj.com. A rapid response
is first posted online. If you need the URL (web address) of an individual
response, simply click on the response headline and copy the URL from the
browser window. A proportion of responses will, after editing, be published
online and in the print journal as letters, which are indexed in PubMed.
Rapid responses are not indexed in PubMed and they are not journal articles.
The BMJ reserves the right to remove responses which are being
wilfully misrepresented as published articles or when it is brought to our
attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not
including references and author details. We will no longer post responses
that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
We acknowledge the important point made by Dr Jacobs. Due to
restrictions on space this was an issue we were unable to address in our
paper. We now do this below.
Conjoint analysis assumes that individuals have continuous
preferences such that there is always some improvement in one attribute
that can compensate for deterioration in the level of another attribute.
In the context of the orthodontic study by Ryan and Farrar,1 this implies
that respondents would be willing to make trade-offs between location and
waiting time. Individuals who always chose either the option with the
lowest waiting time or the option with the same clinic location (local or
central), i.e. were unwilling to trade, were defined as having dominant
(discontinuous) preferences. Fifty-nine respondents (49%) were excluded
from the regression analysis on the basis that they gave responses that
were consistent with having dominant preferences.
There are two points we would like to make.
Firstly, the orthodontic study used a common method to identify
dominant preferences, and a large number were found. The method, combined
with the levels of the attributes included in the questionnaire, may have
identified as dominant respondents who were willing to trade but had very
strong preferences for either location of treatment or waiting time.
Changing the levels of the attributes to elicit at what point these
seemingly dominant respondents would be willing to trade would allow
inclusion of their responses in the regression analysis. Whilst such
refinements may allow trading to take place, consideration needs to be
given to the realism of the levels presented. For example, in a conjoint
analysis study concerned with women’s preferences for alternative in vitro
fertilisation (IVF) programs, it would have been unrealistic to elicit
respondents’ preferences for a 100% chance of leaving the service with a
child.2 Therefore, a balance needs to be found between theoretical
validity and realism.
Secondly, even after refinement of the questionnaire, dominant
preferences may still be observed. Such preferences are important and
should certainly not be excluded from policy decisions. Although the data
from a conjoint analysis study cannot be used to estimate trade-offs
between attributes for dominant respondents (since such individuals are
not willing to trade), there are other ways of using the data. For
example, from the orthodontics study, it is useful for policy makers to
know that 51% of respondents were willing to trade between the attributes.
Further, of the 49% who were unwilling to trade, 18% had dominant
preferences for waiting time, 24% for a central clinic and 7% for a local
clinic. Issues are obviously raised here for policy makers concerning the
weights to be given to these different groups. Information on dominant
preferences can also be used to predict the take-up of new services.3 For
example, when planning services, it is useful for policy makers to know
how different groups will respond to the introduction of a given health
promotion programme. However, the data from conjoint analysis studies
cannot be used to value the benefits to dominant respondents of different
health services. Different techniques should be used here (M. Ryan,
Measuring benefits in health care: the role of discrete choice conjoint
analysis, International Health Economic Association conference, 1999).
1 Ryan M. and Farrar S. Using conjoint analysis to elicit preferences
for health care. British Medical Journal; 2000;320;1530-1533.
2 Ryan M. Using conjoint analysis to go beyond health outcomes: An
application to in vitro fertilisation. Social Science and Medicine
1999;8:535-546.
3 Ryan M, McIntosh E, Dean T, Old P. Trade-offs between location and
waiting times in the provision of health care: the case of elective
surgery on the Isle of Wight. Journal of Public Health Medicine
(forthcoming).
Competing interests:
No competing interests
05 July 2000
Mandy Ryan
MRC Senior Fellow (Mandy Ryan), Research Fellow (Shelley Farrar)
Shelley Farrar
Health Economics Research Unit, University of Aberdeen
Am I missing something here, or is the issue of dominant preferences
a lot more important than the brief paragraph given to them? In the
example used in the paper, about half the questionnaires were excluded
because of inconsistent responses and dominant preferences, so this is
clearly not trivial. It seems to me that including dominant preferences
could drastically affect the conclusions of the analysis.
This seems to be analagous to the difference between a per-protocol
analysis and an intent to treat analysis in a drug trial, where the per-
protocol analysis shows a particular drug in a favourable light, but only
because half of the patients taking that drug dropped out of the study
owing to unpleasant side effects.
Dominant preferences in conjoint analysis studies
We acknowledge the important point made by Dr Jacobs. Due to
restrictions on space this was an issue we were unable to address in our
paper. We now do this below.
Conjoint analysis assumes that individuals have continuous
preferences such that there is always some improvement in one attribute
that can compensate for deterioration in the level of another attribute.
In the context of the orthodontic study by Ryan and Farrar,1 this implies
that respondents would be willing to make trade-offs between location and
waiting time. Individuals who always chose either the option with the
lowest waiting time or the option with the same clinic location (local or
central), i.e. were unwilling to trade, were defined as having dominant
(discontinuous) preferences. Fifty-nine respondents (49%) were excluded
from the regression analysis on the basis that they gave responses that
were consistent with having dominant preferences.
There are two points we would like to make.
Firstly, the orthodontic study used a common method to identify
dominant preferences, and a large number were found. The method, combined
with the levels of the attributes included in the questionnaire, may have
identified as dominant respondents who were willing to trade but had very
strong preferences for either location of treatment or waiting time.
Changing the levels of the attributes to elicit at what point these
seemingly dominant respondents would be willing to trade would allow
inclusion of their responses in the regression analysis. Whilst such
refinements may allow trading to take place, consideration needs to be
given to the realism of the levels presented. For example, in a conjoint
analysis study concerned with women’s preferences for alternative in vitro
fertilisation (IVF) programs, it would have been unrealistic to elicit
respondents’ preferences for a 100% chance of leaving the service with a
child.2 Therefore, a balance needs to be found between theoretical
validity and realism.
Secondly, even after refinement of the questionnaire, dominant
preferences may still be observed. Such preferences are important and
should certainly not be excluded from policy decisions. Although the data
from a conjoint analysis study cannot be used to estimate trade-offs
between attributes for dominant respondents (since such individuals are
not willing to trade), there are other ways of using the data. For
example, from the orthodontics study, it is useful for policy makers to
know that 51% of respondents were willing to trade between the attributes.
Further, of the 49% who were unwilling to trade, 18% had dominant
preferences for waiting time, 24% for a central clinic and 7% for a local
clinic. Issues are obviously raised here for policy makers concerning the
weights to be given to these different groups. Information on dominant
preferences can also be used to predict the take-up of new services.3 For
example, when planning services, it is useful for policy makers to know
how different groups will respond to the introduction of a given health
promotion programme. However, the data from conjoint analysis studies
cannot be used to value the benefits to dominant respondents of different
health services. Different techniques should be used here (M. Ryan,
Measuring benefits in health care: the role of discrete choice conjoint
analysis, International Health Economic Association conference, 1999).
1 Ryan M. and Farrar S. Using conjoint analysis to elicit preferences
for health care. British Medical Journal; 2000;320;1530-1533.
2 Ryan M. Using conjoint analysis to go beyond health outcomes: An
application to in vitro fertilisation. Social Science and Medicine
1999;8:535-546.
3 Ryan M, McIntosh E, Dean T, Old P. Trade-offs between location and
waiting times in the provision of health care: the case of elective
surgery on the Isle of Wight. Journal of Public Health Medicine
(forthcoming).
Competing interests: No competing interests