Michael’s Guide to Qualitative Research

Put simplistically, if you want to know how often GPs prescribe antibiotics for sore throats, you need to do quantitative research. If you want to know why GPs do or don’t prescribe antibiotics for sore throats, you need to do qualitative research.

The latter is probably more useful if we want to work out how to promote change in GPs’ behaviour. Neither method is “better” than the other or is less “scientific” - researchers choose the method that will best answer the research question.

Because of our scientific backgrounds, quantitative research is what we’re more used to, and perhaps more comfortable with. We’re used to the jargon.

The research questions start with a likely rationale or explanation;
The study is then designed to test that in the real world, and
The data are analysed statistically and interpreted in an attempt to prove or disprove the theory;
The research protocol is kept to rigidly throughout the study.

Qualitative research, on the other hand, has a grounding in sociological research, so we find some of the jargon impenetrable. However, it’s a commonly used form of research in General Practice, so it might be asked about in the MRCGP. It's worth learning a little about it, and you can get away with knowing a minimum of jargon. It is used to:

Develop theories and generate hypotheses (why do/don’t GPs prescribe for sore throats?);
Define key variables (what do patients think are the markers of a good GP?).

Qualitative research:

Doesn’t need to start with a theory or explanation;
Research takes a holistic approach - the research question looks broadly;
Data usually gathered in the form of interviews or observation;
The design of the research can change as the context changes;
The feelings and insights of the subjects are considered important - subjects play a key role in interpreting outcomes;
The theory is developed from the data (“grounded theory”), rather than the other way round - the themes and issues emerge during the study.
Data, and the resulting theory, may arise that the researcher hasn’t expected.

Key methods for generating qualitative data are:

Participant observation - the researcher is a member of the group being studied - however, in this method the researcher may overly influence the views of the group being observed;
Non-participant observation - the researcher looks at what is happening from the outside;
Interview research - by means of questionnaires, either face-to-face or by ‘phone.

Interviews can be with individuals or groups. The interviewees are usually chosen because they are likely to be able to give useful, relevant answers (“purposeful sampling”), rather than a random cross-section (another difference from quantitative research).

Interviews can be:

Structured - all the questions are decided beforehand
Non-structured - there are no fixed questions;
Focused - the “Focus Groups” beloved of politicians, where the researcher stimulates group interaction and discussion.

Usually the interviews are recorded, transcribed and the key points summarised.

Sampling can be:

Random
Stratified - where samples from different "strata", or groups, are taken: for instance by ensuring that samples are taken of subjects in different age groups
Purposeful - where the researches choose subjects, or groups of subjects, that will help give them the range of data that they are looking for. This is quite legitimate in qualitative research, as researchers are usually looking for a broad range of data, and there's no point in selecting and interviewing people that can't help.

There is no minimum or maximum number of interviews or subjects. When new interviews don't reveal any new themes, there is no point in doing more interviews. Authors may use a phrase like "after 25 interviews, data saturation was achieved", or "after 10 interviews, no new data emerged" to explain when and why they stopped interviewing.

Validity can be checked by:

Giving the findings to the subjects to check that they are a reasonable account - known as “member checking”, or "respondent validation".
“Triangulation” - using more than one, independent source. In the antibiotics & sore throats example above, a researcher may get data from a focus group, as well as send written questionnaires to other individual GPs. If the results are broadly similar, there is more chance that the results are valid.

Written by: Michael Harris

Last update: 10 January 2008