library@imperial.ac.uk
+44 (0)20 7594 8810
Thanks to your scoping searches you should have a rough idea of what literature is out there to find and how the topic is defined and described in titles and abstracts. Going back to your PICO(S) or question framework will help you to see the main concepts to search for, providing an overall structure for your search strategy. You will then need to decide and how to translate those concepts into a searchable format.
Your search strategy will be based on the key concepts from your PICO or other framework. However, a good search does not necessarily need to include all the concepts, in fact your search should include as few concepts as possible to reduce bias and maximise sensitivity.
Including some concepts in your search may hinder finding what you need rather than help. Unless you are only looking for one or two very specific and well described outcomes, attempting to include outcomes (especially all or any outcomes) in a search strategy is often difficult because you do not necessarily know what they are, or if you do know them, there are likely to be many (both intended positive outcomes but also potential negative effects or no treatment effects) and it will be very lengthy and difficult to comprehensively include them in your strategy due to the variability of ways they will be described or reported in the literature or indexed in databases. If you did miss something it would not be included in your results, thus biasing your search, so it is usually best to leave this concept out of a strategy entirely. This also is the case for research questions focused on discovering or understanding other potentially unknown or unclear elements such as risk factors or barriers or facilitators.
In a 2024 paper Farhad Shokraneh suggests "search resistant concepts" for these known unknowns and unknown unknowns. The method paper discusses the theory and practice regarding search-resistant concepts to help researchers understand how to deal with such concepts. It suggests exclusion from the search, inclusion as a long search strategy, developing a new search filter or using an existing one, priority screening, and supplementary search methods as five ways of dealing with search-resistant concepts. It details when and how to use each of these potential solutions.
For some questions you may find you have to divide one concept into two smaller ones to search for it effectively. For instance if your Population was male thyroid cancer patients you would need to search men as a concept (using relevant keywords and subject headings), thyroid cancer as another concept and then combine them with AND to create the male thryoid cancer population concept.
For some topics, for instance for a new surgical technique, technology, drug, very rare disease or other such niche subject, one concept of the question framework (such as solely intervention or population) may be enough to find all available literature, as the number of results to find is small enough that focusing it further by adding other concepts would have no utility.
For wider or well researched topics (e.g Type 2 Diabetes Mellitus) you will always need to have several concepts in the strategy to keep the number of results manageable and the relevancy high.
library@imperial.ac.uk
+44 (0)20 7594 8810