Introduction to: Discussion, questions and answers
The health sciences clinical researcher has a rich variety of research designs from which he or she may choose, including the experimental survey, single case and qualitative field approaches. Each approach has its unique strengths.
The experimental and single case designs are particularly suited to studies of the effects of health interventions. These are studies that are concerned with the clinical impact of administering different types of interventions (or non-interventions) upon people. In the classical two-group experimental study, similar groups of people are administered with either an intervention or a non-intervention condition and their outcomes are compared. The use of non-intervention with people who are ill has moral and ethical implications. So, often the experimental method is used to study the relative effects of two or more different types of intervention. There has been an affinity between the professions where health interventions are the norm and a form of enquiry which also involves intervention (i.e. the experimental and single case types of research design). However, much high-quality health sciences research involves other types of research design.
Sample surveys are frequently employed to study the opinions of large groups of people concerning health services and their experiences of health and illness. Epidemiological research, which is aimed at studying the distribution of health and illness and associated risk factors in populations, generally involves large-scale sample surveys. These surveys often draw upon hospital and other health agency records.
Qualitative field research involves the discip-lined examination of the personal meanings of individuals’ experiences and actions in the context of their social environment. The emphasis in such research is upon depth of interpretation rather than extent of sampling. Therefore such studies typically employ much smaller samples of participants than, for example, sample surveys.
It is useful to consider examples of the application of the various techniques to the same type of research questions.
Let us consider the situation of people with back injuries associated with manual labour. The major problem associated with back injuries is disability arising from pain. In countries such as the US and Australia older workers who come from non-English-speaking backgrounds are over-represented among people with these injuries. This is probably because such people are over-represented in jobs that have a greater risk of back injury and the effects are cumulative over a long period.
Let us consider an example of where a clin-ical researcher is interested in comparing the effectiveness of two alternate interventions for the treatment of back injury. There are two common approaches: a conservative approach such as physiotherapy, and surgery. So, if we took 20 workers with back injuries and, after random allocation, 10 of them were treated with a conservative intervention such as physiotherapy, and 10 were treated with surgery, we could compare their outcomes.
To achieve a true experimental design, the people need to be randomly allocated to surgery or physiotherapy. Incidentally, to implement this procedure in real life would require a lot of talking to the relevant ethics committees. All the participants would have to volunteer.
As far as the measurement of the outcome is concerned, most experimental studies would employ a quantitative measure of outcome. For example, the workers could fill in a pain questionnaire, perhaps on several occasions after the interventions, in order to compare the outcomes for the two groups. The use of a written questionnaire, especially if it is provided in English only to people from a non-English-speaking background, makes assumptions about the literacy of the participants which may not be valid.
If the clinical staff had chosen the interventions for the patients on a non-random basis, the study as described above would be a natural comparison study. Simply comparing groups does not mean we have an experimental design. Otherwise, the study could be structured in an identical fashion to the experimental example described above.
A single case design involves the administration and withdrawal of interventions in a systematic fashion in order to observe their impact upon the phenomenon under study. The person in the single case trial is usually challenged with varied interventions to compare the effects of each. This type of design closely approximates the nat-ural clinical history of interventions with many patients, particularly those with chronic illnesses. However, the interventions in a single case study are structured much more rigorously.
In the present context, while it would be possible to administer and withdraw physiotherapy and to have baseline phases of no treatment, the surgical intervention cannot be withdrawn. Once the surgery is done, it permanently and drastic-ally changes the person’s body. Therefore a single case example could involve the alteration of baseline (no treatment) and physiotherapy and then, as the last link in the chain of event, the surgery. This is not a methodologically strong design, as there could be carry-over effects from the previous interventions that interact with the surgery, and the order of the administration of the surgery could not be readily altered.
An alternative way of conducting the study of workers with back injuries would be to select a large group of such people and survey them for the different types of interventions they have had for their injury. The outcomes for different groups could then be compared because, as in a natural comparison study, we can compare men and women respondents. We could even give them the same pain questionnaire as in the types of study previously discussed.
Surveys do not, of course, necessarily involve the use of questionnaires. One could bypass the patients and survey their medical records (with necessary approvals) using a coding schedule. Alternatively, the information could be collected in the form of a structured survey interview.
An additional way of studying this issue would be to conduct in-depth interviews with a small number of injured workers to study their interpretations of their situation. Workers who had surgery, as well as those who had physiotherapy, could be interviewed. The interviews would normally be recorded verbatim, transcribed and then exhaustively analysed for the theoretical constructs needed to describe the experiences of the participants. The use of checklists, survey inventories, etc. is normally avoided, although it is usual to have a list of issues to be introduced by the interviewer. The respondents typically describe their experiences and perceptions in their own natural discourse.
In studies which use questionnaires, there is limited interaction between the researcher and the respondent. As suggested by their name, the respondents respond to the questions framed solely by the researcher. Such an approach can be useful in eliciting factual information in an economical manner. However, it places the respondent in a relatively passive role. If the researcher and the respondent do not share the same constructs, ideas, feelings and motives, the ‘wrong’ questions may be asked. Schatzman & Strauss (1973, p. 57) note that the researcher ‘harbours, wittingly or not, many expectations, conjectures and hypotheses which provide him with thought and directives on what to look for and what to ask about.’ The respondent in most questionnaire studies has very little opportunity to contribute to the research agenda. Generally, a token ‘any other comments’ section is the extent of the invitation for the respondents to contribute to this agenda.