The steps of a proper search
Once you have formulated your PICO or PIRO question, you need to start looking for literature. When constructing a guideline, the initial search should focus on already existing guidelines and systematic reviews that potentially have already answered your question. If these do not exist or if they are of bad quality, then a search for randomized and/or non-randomized studies is required. As such, the search is conducted in 3 steps:
- Systematic reviews
- Randomized/non-randomized trials (primary literature)
If you happen to find a systematic review, that however needs an update, you can do a search starting from the latest search date of that particular review. By default, the search should only extend 10 years back. However, this should be discussed for each individual question.
A further search to get the full picture
Searching for unpublished data is difficult. However, if you want to assess if there is the potential of publication bias, it may be relevant to include a search for unpublished data in validated database such as clinicaltrials.gov. In addition, the search may extended to include posters, abstracts ect from conferences.
When relevant, a search may include a search for adverse events concerning medical and technical interventions. This can be done by searching in relevant databases such as the European Medical Agency (EMA) and the Danish Medicines Agency.
Document what you do
The search terms and relevant databases should be determined in the beginning of project. Here, constructing a search protocol is essential as it provides and overview and transparency of the search process. The search protocol is a description of what you have been searching for, where you made the search and how you did it.
Consider it a cook book, everyone reading it should be able to follow the description and largely get the same result.