Systematic reviews: Search strategy

Your search terms determine what you find. Identify the most appropriate and relevant terms that cover each aspect of the research question. Think creatively, analytically and consider if you need to use very specific terms or broader terms.

Searching "free text" means searching in the language of the document or a representation of the document. One problem with free text searches is that language is ambiguous, and the same concept can be expressed in many ways. Therefore free text searches can lead to inaccurate searches. It is important to consider all synonyms, related concepts, opposites, super- / sub-concepts, abbreviations, grammatical and linguistic variations of all free text keywords when you are searching.

Each individual database or resource that allows free text searching, allows the search in different fieds and to a different extent. To make the search more specific you can restrict the search to certain bibliographic "fields" e.g. the title or abstract. 

Controlled vocabularies are the preferred terms applied to each reference by the producer of the database the reference is indexed in. The terms denote the major themes or subject of the reference. Controlled vocabularies ensure that references on the same topic are indexed uniformly, regardless of the author's word choice in the title, abstract and the author's own keywords.

Controlled vocabularies differ from database to database. The list of controlled keywords in the vocabularies can be hierarchically ordered, and is then called a thesaurus, where you can choose between broad, general terms (BT broader terms) and narrower, specific terms (NT narrower terms). An example of a thesaurus is Pubmed's "MeSH database"

The choice between "free text" searching and searching using terms from a controlled vocabulary depends on the subject you are working with, the source you are using and the aim of the search. 

• A free text search results in a large set of references or "hits". There will also be alot of "noise" in the results (noise = non-relevant results). In all, high recall, but with low precision. 
• A search using terms from a controlled vocabulary will result in a narrower search and more precise results. In all, low recall and high precision. 

The search in a systematic review aims to be comprehensive and to find "everything". The recommendation is to use both approaches together - free text search and search using controlled vocabularies. 

Block searching is a strategy for streamlining and structuring the search using keywords effectively. The research question is divided into concepts and organised in blocks, where each block covers one concept. In each block, controlled index or thesaurus terms and free text terms are combined using OR. The differnet blocks are combined together using AND. 

Example from Pubmed:

Block searching allows you to:

• Expand or narrow the search by adding or removing search terms i the individual block, remove blocks entirely (expland the search) or add blocks (narrow the search). 
• Document the search in a manageable and transparent way and accordingly ensure consistency when updating or replicating the search. Read more in the section "search protocol".

The PRESS guideline is an evidence-based guideline for the Peer Review of Electronic Search Strategies. 

The complete guideline is found in the article below: 
​McGowan J, Sampson M, Salzwedel DM, Cogo E, Foerster V, Lefebvre C. PRESS Peer Review of Electronic Search Strategies: 2015 guideline statement. J Clin Epidemiol. 2016 Jul;75:40-46 

Detailed explanation of guideline og PRESS checklist:

PRESS – Peer Review of Electronic Search Strategies: 2015 Guideline Explanation and Elaboration (PRESS E&E). Ottawa: CADTH; 2016 Jan.

The search in a systematic review aims to identify everything that has been written about a problem or question.

Therefore it is important to search in many different databases and other information sources to identify as many relevant studies as possible, The choice of databases and other sources depends on the subject of the review. 

Search in both subject-specific databases, such as Pubmed, Embase, Psycinfo, and in cross-disciplinary sources such as Web of Science and Scopus. 

Grey literature referes to literature that has not been published. There are two types of grey literature:

• File drawer: which is academic research that has never been published - rejected, or not written up, e.g. conference posters/abstracts or negative results that are not attractive to journal.
• Practictioner: research not intended to be published such as internal reports, consultancy reports and government reports.

Some of the most commonly used databases and search engines include contributions to conferences (Embase, Web of Science). There are also databases dedicated to grey literature, e.g. Grey Matters (Health) and BASE (multidisciplinary) 

Read more about Grey literature

Clinical trial registers register information about clinical trials, and can be important sources of information if you want to identify current and past e.g. randomized clinical trials (RCTs):

ClinicalTrials.gov 

WHO ICTRP

Define your inclusion and exclusion criteria whilst you design your search. Criteria associated with the topic should be reflected in the keywords. Other criteria can be built into the search strategy by using the tools ("Filters", "Limits" etc.) that many databases have, and which can filter the search based on various criteria, such as year of publication, language and document type.

Sometimes, depending on your subject, it is wise to filter during the forst screening rather that during the search to avoid loosing relevant references.

Which ever appraoch you choose, inclusion and exclusion criteria must be well-founded and stated in the protocol.

Websites with inspiration for searches  that can be used to include references based on e.g. type of study, method, etc .:

ISSG Search Filters Resource

McMaster University

University of Alberta