Conducting a systematic literature search
Posted on July 18, 2017
Tags: database, literature search, PICO, search
Systematic reviews sit amongst the top of the evidence hierarchy. This is all well and good, provided they are conducted appropriately. To ensure a review can provide high quality, reliable evidence, they must be completed meticulously, following reporting guidelines such as the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and the Cochrane Handbook (1, 2) A key element of this is a robust search strategy.
Systematic reviews take existing literature and synthesise it, either qualitatively or quantitatively. In order to be ‘systematic’ and minimise selection bias, the search must be comprehensive, encompassing all of the relevant research. Bias will exist if the authors fail to include ALL of the pertinent research. “How can you be sure your conclusions are true if you have not explored all of the evidence?”
Of course, some articles may be excluded (for instance if they are methodologically flawed) once you come to screening the evidence, especially if this is a meta-analysis. However, the authors must ensure that initially all of the relevant evidence is captured by the search.
Performing an in-depth search is often overwhelming for authors setting out with a systematic review. It can seem confusing or difficult and is often substituted for a more manageable, less rigorous search. The following checklist should provide a clear framework for those wanting to ensure their search is truly systematic.
All illustrative example of database searching will be performed on Ovid Medline. Please note that this is a basic worked example to illustrate some of the key principles involved in searching. Using the search outlined below, you are likely to pick up key studies. However, to minimise your chances of missing relevant articles then do consultant a librarian who will be able to assist you with advanced search techniques. This is the best way to maximise sensitivity and reduce your risk of missing relevant articles.
BEFORE THE SEARCH:
Firstly, it is important that systematic review protocols (including search strategies) are prospectively registered to prevent deviation from the intended methods and minimise bias. A popular registration database is the International Prospective Register of Systematic Reviews (PROSPERO).
Also, it is a good idea to carry out a search for any similar systematic reviews. It may be pointless to repeat an existing systematic review. Although if an existing systematic review is outdated or features flawed methodology, then a new systematic review on that topic can be justified.
Defining the research question
Before a search can be performed, it is crucial that the research question is explicitly defined. There are many ways to do this, but the most common method is to divide your question according to the Patients, Interventions, Comparisons and Outcomes (PICO) model.
Patients:Which patient population do you want to explore?
Adults? Elderly? Paediatrics? Males? Females? Certain ethnicities? Inpatients? Community patients? Certain co-morbidities?
Interventions:What intervention(s) do you want to explore?
A certain medication? A certain surgical procedure?
Comparisons:What are you going to compare the intervention(s) against? (Some reviews may not have a comparison so this section is optional).
Are you going to compare the intervention with no intervention? A control? Usual treatment? A different medication? A different surgical procedure?
Outcomes:What outcomes do you want to explore?
Mortality rate? Morbidity or the development of a certain disease? Change in baseline parameters such as blood pressure, weight or cholesterol?
Defining the parameters of the search
Once a well-defined research question has been established, it is important to outline where you will search for the evidence. Systematic searches should aim to search as many different sources as possible. This can be broken down into the following:
Online databases* (this will be the major area for medical literature searches)
Grey literature* (this refers to unpublished material/published in a non-commercial form)
Ongoing trials (clinicaltrials.gov)
Types of articles
Randomized controlled trials (RCTs)
Case series studies
How far back do you wish to explore?
Perhaps the intervention you want to investigate was only developed in the last 10 years
Other languages too? Ideally you should be willing to translate these in order to fully understand their findings
*The following databases are useful for clinical systematic reviews:
PsychINFO – key database for mental health literature
MEDLINE – large medical database
EMBASE – large medical database
SCOPUS – includes many scientific disciplines
Cochrane Library – high-quality evidence
Web of Science – includes many scientific disciplines
CINAHL – includes biomedicine, healthcare, nursing and allied health articles
*Examples of grey literature databases include:
THE SEARCH ITSELF
Searching online databases
When searching online databases, the terms and their synonyms for each of the components of the PICO model must be written out, including abbreviations. It is also important to use alternate spellings and word endings. This can be done using a number of strategies within the database:
Firstly, truncation involves putting an asterisk (*) in a word with a variable ending e.g. toxic* will search for toxic, toxicity and toxicology.
Secondly, using a question mark (?) in place of a single letter for words that have alternate spellings (such as American vs British English) will search for both spellings, e.g. p?ediatric will search for both paediatric and pediatric. The question mark replaces a single character or no character, allowing you to pick up alternate spellings. It should be noted that these terms can vary between databases, and so should always be checked prior to conducting a search.
You should also search for hyphenated terms with and without the hyphen as different authors may title their work differently.
An example to illustrate a search in Ovid Medline:
Let’s imagine we want to conduct a systematic review to assess the effect of primary percutaneous coronary intervention (PCI) versus streptokinase on mortality in myocardial infarction (MI) patients. (Clicking on the screenshots below will enlarge them).
Patients:myocardial infarction, MI, myocardial necrosis, heart attack
Interventions:primary percutaneous coronary intervention, PCI, percutaneous coronary intervention, primary PCI, reperfusion, re-perfusion, angioplasty
Comparisons:streptokinase, medical reperfusion, medical re-perfusion
Outcomes:mortality, death, survival
As you can see, each line represents a synonymous term that is searched for, with the corresponding number of articles found in the results column. These terms can be combined using ‘OR’ to ensure the database captures all of the articles relating to these terms.
For the final screenshot, all of the groups of terms are combined with ‘AND’ to ensure the database only displays results that are related to the combination of terms (i.e. about PCI and streptokinase in MI patients). In this example, it yielded a total of 254 results.
Further refining or increasing your results
Medical Subject Headings (or MeSH terms) are terms predefined by the database using human indexers in concordance with thorough protocols. MeSH terms encompass ‘Headings’, ‘Subheadings’, ‘Supplementary Concept Records’ and ‘Publication Characteristics’. Further definitions of these categories and relevant examples can be found on the National Library of Medicine’s website (https://www.nlm.nih.gov/bsd/disted/meshtutorial/introduction/04.html). The same website also provides detailed descriptions of how the MeSH system is structured and how this can be searched to further refine or expand your literature search. Using Medline as an example, you must select ‘Map term to subject heading’ to search for MeSH terms.’
If your search yields no results then it is important to re-evaluate your PICO model and broaden the search terms. Conversely, if the results are very expansive this will take considerable effort to screen the relevant information. It may be worth narrowing the search terms in order to achieve a more focused search.
AFTER THE SEARCH
Anymore studies to include?
Once the formal search has been completed, you must now remove the duplicates and screen the titles/abstracts of the remaining results. This is a large task in itself, and information on how this can be done is beyond the scope of this blog.
Once competed, a further exploration for relevant studies can begin. There are a number of strategies that should be carried out at this stage.
- Citation tracking – This involves searching for studies that have referenced the included studies in their work, and as such, may be relevant to your systematic review. Most databases will have an integrated tool to perform this.
- Manual reference searching – Once you have finalised a list of applicable studies, you should go through their references individually and search for additional relevant studies.
- Contacting authors – Finally, authors of the included studies can be contacted to search for further results that may be pertinent to your review, or to ensure your search strategy has not missed any of their other work which may be useful.
At this point it is crucial that you finalise your list of included studies and document your search well so that you can refer to this in the future and during the rest of the review process.
Your SYSTEMATIC literature search is now complete!
Congratulations! If you have followed the various steps in this article, you are well on your way to completing a comprehensive systematic review! If you need further help, consult your librarian.
- Higgins JPT, Green S (editors). Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 [updated March 2011]. The Cochrane Collaboration, 2011. Available from cochrane-handbook.org.
- Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group (2009). Preferred Reporting Items for Systematic Reviews and Meta-Analyses:The PRISMA Statement. PLoS Med 6(7): e1000097. doi:10.1371/journal.pmed1000097
Having completed my fourth year as a medical student at the University of Liverpool, UK, I am currently intercalating in a Masters of Research (MRes) in Medicine at the University of Leeds, UK. I will be entering my sixth and final year of medicine in September 2017. My primary interests lie within Cardiology and Evidence Based Medicine.
More Posts - Website
Now that you have an answerable question and an idea of what type of database you need to search (at least to start), let's talk about the nuts and bolts of searching. For the purposes of this paper, we will use PubMed as the search engine.
A. Getting specific—the basics of “how to”
Type a word or phrase into the query box, including subject, author, and/or journal
Click on the search button or press the “enter” key
Results will be displayed in summary format:
To retrieve more information about the search results, use the display settings menu (upper left corner) to view the abstract or MEDLINE formats, change the number of items that appear per page, and sort by recently added, publication date, first author, last author, journal, or title.
PubMed also contains links to full-text articles (appears in upper right corner of page) at participation publishers' web sites as well as links to other third party sites such as libraries and sequencing centers.
2. Advanced searching in PubMed—MeSH terms and the MeSH database:
a. Medical subject headings (MeSH)
It is important to understand that PubMed uses a controlled vocabulary to index journal articles called MeSH and uses “automatic term mapping” to find MeSH terms when you search. MeSH terms are organized in a hierarchy called a tree, with more specific (narrower) terms arranged beneath broader terms. By default, PubMed includes in the search all narrower terms; this is called “exploding” the MeSH term. Inclusion of MeSH terms enhances and optimizes the search strategy. For example, if you looked up the term “Spine” in the MeSH database you would see:
Therefore, PubMed would retrieve every article containing any of the terms located under Spine in the hierarchy.
b. MeSH database features
MeSH vocabulary contains over 25,000 descriptors and is updated weekly and reviewed annually. You can only search citations that have been indexed for MEDLINE (92% of the PubMed database) using MeSH terms. Features include:
Allows you to identify and select appropriate MeSH terms for a search and to see their definitions
Builds a PubMed search strategy
Displays MeSH terms in the hierarchy (MeSH tree) allowing you to broaden/narrow a search
Limits MeSH terms to a major concept/topic heading for a search
Allows you to broaden your search by choosing not to explode a term
Attaches subheadings for a search creating complex search strategies
The list of subheadings includes terms paired at least once with a given heading in MEDLINE.
Focuses searches using other types of MeSH terms including publication types [pt], substance names [nm] or registry numbers [rn], and pharmaceutical actions [pa]
MeSH Brower for access to annotations: http://www.nlm.nih.gov/mesh/
To access MeSH from PubMed, click on MeSH Database on the PubMed homepage or click MeSH under “more resources” in “advanced search.”
Once in the MeSH database, if you entered cancer into the search bar and clicked Go (or hit Enter) you would see:
Clicking on “neoplasms” will bring up the page where you have the option of selecting any of the features listed above to help you refine your search.
Also, clicking “links” adjacent to the MeSH term desired, will give you a drop-down menu which offers several options:
PubMed: search PubMed with the term
PubMed—Major topic: search PubMed with the MeSH term, retrieving only citations where the term is a major focus
Clinical queries: put the MeSH term into the Clinical Queries box where the search may be further refined
NLM MeSH browser: show the MeSH browser descriptor data for this term including scope note, allowable qualifiers, and the MeSH tree
⇛ Anything which appears in blue and is underlined is a link that reveals more information. Clicking on the title would bring up the abstract (Abstract format). Clicking on “Related articles” would provide a link to other similar articles that might be of interest.
⇛ The Mesh database homepage includes three brief tutorials on how to search with the MeSH database, combine MeSH terms, and apply subheadings and other features of the MeSH database.
B. Too much information! Refining your search
Replace general search terms with more specific terms (the MeSH database would be a great resource for this)
Add terms or combine search terms with connector words: AND, OR, or NOT using upper case letters (called Boolean logic)
AND between terms returns only records that contain all of the search terms
OR between terms returns all records that contain any of the search terms
NOT between search terms returns only records that contain the first term and not the second
Truncate terms. Place an asterisk (*) at the end of a string of characters to search for all terms that being with that string. PubMed searches the first 600 variations of a truncated term.
Example: mimic* will find all terms that begin with the letters m-i-m-i-c-; eg, mimic, mimics, mimicking
Use a wildcard. Use a “?” to replace a letter or denote an extra letter where spelling or word variation is possible.
Example: behavio?r will find behaviour or behavior
Use the “limit” option in PubMed to limit citations by age group, language, publication type, date, human studies, etc.
Use the “advanced search” option to look up a term as it is indexed in PubMed
Use the MeSH database features
Let's say we are interested in what the best surgical treatment is for osteoporotic spine fractures. Using some of the tips above, the chart below shows how a typical search might go:
|Search term(s)||# citations|
|“Spinal fracture” [MeSH] OR vertebral compression fracture||16023|
|“Spinal Fracture” [MeSH] OR vertebral compression fracture AND “osteoporosis” [MeSH]||3718|
|“Spinal Fracture” [MeSH] OR vertebral compression fracture AND “osteoporosis” [MeSH] AND “surg*”||911|
|“Spinal Fracture” [MeSH] OR vertebral compression fracture AND “ osteoporosis” [MeSH] AND “Surg*”|
Limits: only items with abstracts, humans, clinical trial, English, publication date from 1990–2010
View it in a separate window
By combining terms (using Boolean logic), truncating a term, and using the limits option we were able to narrow our search down from 16,023 articles to a more manageable and relevant 54 articles. The “details” tab in the PubMed search window shows the complete search expression (ie, query translation) employed by PubMed, similar to what is represented in the table above.
One of the best resources that PubMed provides for users new to the database is the online tutorials. They are brief but informative and because they are interactive you are guided step-by-step through each process. Perhaps consulting the online tutorials and the fact sheets on PubMed would be a next step for you. Give it a try! In fact, here is the link to the PubMed Tutorial homepage created by the National Library of Medicine: http://www.nlm.nih.gov/bsd/disted/pubmedtutorial/. Also, check-out the PubMed help page which contains a plethora of information regarding all aspects of PubMed: http://www.ncbi.nlm.nih.gov/bookshelf/br.fcgi?book=helppubmed&part=pubmedhelp