Developing a search strategy

Before developing a systematic search strategy, one needs to do some preparatory work: Searching for background information, reading textbooks and essential papers; sometimes also discussions and interviews with experts in the field.

Before embarking on specific search activities, it is essential to understand what the subject of the search really means - to understand the medical, social, human and economical implications of the topic under investigation.
Generally speaking, one needs to form a cognitive concept of the subject area and the directions a search should take.

Background searching

Background searching has to precede the development of any systematic database search strategy.

It helps to develop a contextual concept of the search topic in question and to collect a number of most promising search terms for the topic-searching parts of a systematic strategy.
It also guides the further strategy development regarding the necessity of filtering, which is dependent on the amount of research on a given subject (like any diagnostic test, search strategies are dependent in their "predictive value" on prevalence rates). Thus, the background search helps to get an idea about the expected amount of published material on a subject - and one can refine the search strategy to suit the "publication prevalence".

A seemingly simple question may not be so simple after all.
As a doctor, what you are interested in is something like "Which patient will profit most from a certain intervention?"
This is not a simple question at all, and no database will give you the answer to this kind of question right away.
You need to re-arrange your question.
What databases can tell you are "things that are"; - i.e.:

• Intervention "A" is better than intervention "B" for patients with characteristics "xy".
• Intervention "A" is similarly effective as intervention "B", but some patients with characteristics "xy" experience less negative effect when treated with "A".
• Intervention "A" did not prove effective in randomized controlled trials.

Thus, you have to check your patient's characteristics - and if you want answers for a large group of patients (like "everyone" in your country), you have to form classes or groups of patients with clinical characteristics or any characteristics that may be of importance - or that correspond to patient parameters generally reported in trials relating to the health problem under investigation.
Next, you can start looking for trials that included patients similar in all - or at least many - important aspects to "your" patients.
Looking at the literature in this way will often lead to the discovry of important omissions in research - like trial results, that were gathered from a completely or overwhelmingly male trial patient population, missing study results for the elderly, teenagers or children; no trial information on ambulatory care conditions in relation to the trial intervention etc. ...