A good research problem best arises naturally in the context of practice, rather than being contrived for the sake of an impressive data display. At ®rst, the idea is very general. Don't arti®cially restrict your enquiries at the outset through what you perceive as being `analytical limitations'. Analytical limitations can be of two types.
1You are apprehensive about your own grasp of statistics and hesitate, therefore, to embark on a project which you suspect will involve sophis- ticated techniques.
2 You ask a question for which mathematics has not yet devised a satisfactory analytical approach.
Let neither of these deter you. With respect to (1), you just need to have the faintest idea of what you want, and any number of statistics handbooks are available to tell you exactly what to do. As far as (2) is concerned, go ahead and formulate the question. If the limitation you perceive is real, none of the handbooks will help you. You then go to a statistics lecturer (but retain your copyright of the question ± just in case!). If they cannot help you, the question will be referred to a mathematician dealing with that ®eld. If you really have found a gap, you will be informed and thus have the basis for an interesting communication to a statistical research journal. It is not often that simply the question leads to a publication, but it has happened!
Think right now of some very general research questions that can arise in your ®eld of healthcare, e.g. eect of training sessions on interpersonal skills in engendering patient con®dence in the practitioner; does osteopathy reduce the incidence of pre-eclampsia in pregnancy, etc.? There are no limits! Notice that the phrasing of the initial general question totally ignores such strategic considerations as how to measure the variables. Get your priorities right; let your genuine professional curiosity govern the questions you ask and only re®ne them later in the light of restrictions imposed by data collection
A brief introduction to designing a research project 183
problems, ethics, etc. This `re®ning' process is best carried out in consultation with more experienced colleagues, a statistician or two, the Ethics Commit- tee, etc. But the important thing to remember is that the ®rst step in research design is to ask a question.
Now, even before re®ning the question, consider your resources. Do you have access to the data? If you need clinical data, do you (or your colleagues) see enough patients to obtain it? If a `before' and `after' situation is envisaged, does your situation permit collection of data at frequent enough intervals?
After you have satis®ed yourself with respect to these details, you need to seek professional advice to ascertain whether the question has already been asked and answered. Just because the idea is original to you does not necessarily mean that it is original to the profession. Often, all the profes- sional advice you need in this regard can be provided by a good medical librarian. More often, you will have to check with your research director ®rst, who, in collaboration with the librarian, can see whether the matter has already been resolved. This step is called `surveying the literature'.
At this point, if you are honestly sure that the ground has not already been trodden, you set down and elaborate your research protocol. This includes asking the question as unambiguously as possible and stating in precise detail how you propose to answer it. It is possibly the most dicult part of the whole process and the restrictions will come at you thick and fast. For instance, is it ethical? You would think that such a question would be easy. Not so. Suppose you had reason to believe that a particular 5-minute manipulation of a pregnant patient's feet signi®cantly reduced the incidence of discomfort caused by oedema. OK, how would you ®nd out? I suppose you could do it to every second patient you saw from the ®rst trimester through to delivery and then analyse the dierence between those to whom it was done and those to whom it wasn't.
But, hold on! Suppose it does work? Then, you have knowingly deprived certain patients of bene®ts by withholding a technique that certainly would not have harmed them, even if it had not worked. Was that ethical? Well, you would have to be a Solomon or a Daniel come to judgement to determine such things. Fortunately, you rarely have to do so unaided. The people who run the facility in which you are taking the data have to make those decisions, so you consult them when you have worked out your protocol. In most cases they, in turn, will be able to consult an Ethics Committee. In reviewing your proposal, the people who run the facility will not only consider the ethics involved, although that is the most crucial consideration; they will also advise as to whether or not the proposal is ®nancially feasible; whether it can be carried out in the time allocated, etc. Your research protocol needs to be written up so well (including a literature review) that anyone else with your
184 Basic concepts in statistics and epidemiology
training could carry out the experiment precisely as you would. The next section will tell you how to do this.
Thus, to summarise: 1Ask a general question.
2 Look into your resources for answering it. 3 Do a literature survey.
4 Re®ne your research protocol in consultation with your supervisors. 5 Write up the research proposal.