Research design and methods
The main features of the study design and methods you select follow the question that has been posed. For example, you may be using a survey by questionnaire or interview; a case study of one or more person or organisation; a trial, which may be randomly controlled or use matched or waiting-list controls. It may be ethnographic, requiring time in the field to look at relationships, culture or communication, or action research, where practical problems are considered and where the feedback, changes and subsequent evaluation of change are all part of the research – rather like audit in its cyclical design.
Action research is a common mode of educational research. It appeals to those who like the idea of change and the feeling that the research they do actually contributes to it. It joins together both research and implementation, and is a much messier, more participative research method than most; so people who start it should not be those who demand precision or decimal points in their answers, and they should enjoy being involved with the teams and individuals who actually put the changes into practice (Somekh, 1995).
Research may also be cross-sectional or longitudinal; depending on what is the right way to answer the question and what resources are available. For example, cross-sectional designs can provide answers to the immediate success of a programme if success is seen in terms of, for instance, course satisfaction. They can also be used to see, for example, if different people are more or less satisfied depending on their learning preference, or to consider any other differences between them or between course presenters, or an interaction between both. But if you want to assess what really changed as a result of your course, then it has to be longitudinal.
Longitudinal design is expensive, not only because you have to have the researchers available over the length of the study, but also because you might have to spend funds on tracking your subjects over time. However, the benefits of longitudinal cohort studies are that you can answer more complicated questions. For example, if you were studying job satisfaction in doctors using a cross-sectional design, you would probably find that around a quarter were pretty dissatisfied. If you had a large enough sample, you could break it down into specialties and see that one particular group (e.g. surgeons) were very satisfied, while another (e.g. psychiatrists) were very dissatisfied. You would be likely to conclude that the work role of psychiatrists has more factors likely to produce dissatisfaction than the role of surgeons does. However, a longitudinal study might have assessed these doctors as students or house officers and looked at personality or at satisfaction then, and would probably have found that many of those who are disgruntled now were disgruntled then. This then allows us to see how much of the dissatisfaction is due to the person and how much is due to the job, which in turn means we may not need to continue to plough resources into the job alone, but also into individual intervention or selection issues.
Your research question and available resources will determine over what period you should assess and how many assessments you will need to make. Although it will vary according to the intervention made, where clinical interventions are the subject of research, patients are usually assessed during a baseline period, immediately after the intervention, a few weeks or months later, and after a year or two. Ideally, educational interventions would not differ much from this, but a minimum would be before, immediately after and 6–12 months later.
- Taking the exercise above, think about the types of design that might be necessary for each level of question. The last one – a randomly controlled trial design – has been described, but what designs and methods would best fit the others?
Please see our suggested answers here.