LibGuides: Systematic Reviews: Where to Start

Scoping the Topic

Initial searches are scoping exercises, giving an overview of the topic, placing your work in the context of already published studies and working out:

if it has already been carried out
the size of the published literature
gaps in the literature
types of studies and research designs
geographical spread of publications

Initial searches also identify:

the relevance of various contexts to your topic
concepts that appear in the results which are peripheral to your topic
concepts within your topic which are missing from the results

At this stage, skim the results to identify the above. Look at titles, abstracts, subject headings. Initial search strategies will be amended for relevancy and accuracy following feedback from this appraisal of results.

Database search process flow chart

Putting Search Terms Together

You are ideally looking for a combination of keywords and subject headings to describe each aspect of the topic. Use OR to combine the words and subject headings which describe each aspect.

Finally combine the aspects together with AND.

To help clarify the logic of the search, it may help to consider it schematically. Identify and group the main keywords and subject headings.

This is a search derived from a systematic review entitled: “Systematic review of risk prediction models for falls after stroke.”

Use the thesaurus of subject headings within the various databases to identify relevant headings.

For example, in PubMed Medline, select the MeSH database to find additional relevant Medical Subject Headings to add to the keywords.

‘Systematic review of risk prediction models for falls after stroke’[1]

In the published review, searches were conducted on seven databases (Medline, Embase, CINAHL, PsychINFO, Web of Science, Cochrane Library and SCOPUS. In accordance with best practice the published review presents all search strings as executed on all databases as supplemental data.

The table below gives a summary search strategy for this topic for PubMed Medline.

*Aspect*	*Keywords*	*Subject headings*
Stroke / brain trauma	stroke[Title/Abstract] OR poststroke[Title/Abstract] OR hemiplegia[Title/Abstract] OR brain hemorrhage[Title/Abstract]	“Cerebrovascular Disorders"[Mesh] OR "Cerebrovascular Trauma"[Mesh] OR "Stroke"[Mesh] OR "Brain Infarction"[Mesh]
	AND
Falls	falls[Title/Abstract] OR falling[Title/Abstract] OR accident[Title/Abstract] OR accidents[Title/Abstract] OR slips[Title/Abstract] OR trips[Title/Abstract] OR tripped[Title/Abstract] OR stumble[Title/Abstract]	“Accidental Falls”[Mesh] OR “Accidents”[Mesh]
	AND
Risk	risk[Title/Abstract] OR prediction[Title/Abstract] OR predictive[Title/Abstract] OR score[Title/Abstract]	"Risk"[Mesh] OR "Risk Assessment"[Mesh] OR "Risk Management"[Mesh]

Appendix 2 of this guide gives the full versions of this full search for PubMed Medline and EMBASE databases.

[1] Walsh ME, Horgan NF, Walsh CD, et al Systematic review of risk prediction models for falls after stroke J Epidemiol Community Health 2016;70:513-519.

“Topical non-steroidal anti-inflammatory drugs for analgesia in traumatic corneal abrasions’[1]

OBJECTIVES: To identify and evaluate all randomised controlled trials (RCTs) comparing the use of topical NSAIDs with placebo or any alternative analgesic interventions in adults with traumatic corneal abrasions, to reduce pain, and its effects on healing time.

*Aspect*	*Keywords*	*Subject headings*
Randomized controlled trial	Randomized OR Randomised OR Placebo OR Randomly OR Trial OR Groups	Randomized controlled trial
	AND
Eye, Eye Injuries	Eye(s) OR Cornea(s) OR injury(ies) OR abrasion(s) OR erosion(s) OR trauma(s) OR wound(s) OR foreign	Cornea OR Corneal Diseases OR Wounds and Injuries OR Eye Injuries
	AND
Anti Inflammatory Agents	Nsaid OR nonsteroidal anti‐inflammator OR non‐steroidal anti‐inflammator OR diclofenac OR fenoprofen OR flurbiprofen OR indomethacin OR ketoprofen OR ketorolac OR piroxicam OR bromfenac OR nepafenac OR oxyphenbutazone OR suprofen	anti inflammatory agents non steroidal/ OR diclofenac OR indometacin OR ketoprofen
	AND
Analegsia	Analgesia OR Pain	Analgesia OR Pain

Appendix 3 gives the published version of the full search as run on the OVID Medline database.

[1] Wakai A, Lawrenson JG, Lawrenson AL, et al Topical non-steroidal anti-inflammatory drugs for analgesia in traumatic corneal abrasions. Cochrane Database Syst Rev. 2017 May 18;5:CD009781. doi: 10.1002/14651858.CD009781.pub2.

‘ Wound Cleansing for Pressure Ulcers’ [1]

Identify and group the main keywords and subject headings

*Aspect*	*Keywords*	*Subject headings*
Wound cleansing	Normal saline OR Povidone OR Water OR Soap*	Solutions OR Alcohols OR Water OR Soaps
	AND
Pressure Ulcers	Pressure ulcer* OR Pressure sore* OR Bed ulcer* OR bed sore* OR bedsore* Decubitus ulcer*	Pressure Ulcer

Appendix 4 gives the published version of the full search as run on the CINAHL Ebsco database.

[1] Moore ZEH, Cowman S. Wound cleansing for pressure ulcers. Cochrane Database of Systematic Reviews 2005, Issue 4. Art. No.: CD004983. DOI: 10.1002/14651858.CD004983.pub2

Search strategies are created according to the specifications of the database, for example whether * or $ is the truncation symbol; whether subject heading searching is possible; range of filters/limits offered etc. See the on-screen help tutorials for each of the databases. Give yourself plenty of time to become familiar with the structure of the different databases.

Translate Search to Other Database

Before running searching in other databases, be confident that you have refined your ideas and terms sufficiently to retrieve good results in the primary core database.

It is a complex task to “translate” a search from one database to another and requires a lot of time and effort. Essentially, you need to start again with each database. Find out about the database structure and use your exploration of keywords, subject headings and filters on one database to start exploring the next.

Each database provider (eg ProQuest, Ebsco) structures the databases differently.

For example:

ProQuest databases (Medline): default search is by subject heading
EBSCO databases (Medline, CINAHL, PsychInfo): default search is by keyword
PubMed database (Medline): default search is a combination of subject heading and keyword
Elsevier Science Direct database: default search is a combination of subject heading and keyword

Each database (Medline, CINAHL, PsycInfo,) has a different set of subject headings.

Some databases (Web of Knowledge, SCOPUS) don’t use subject headings, so you are relying on searching authors’ words – remember to think of different words to describe the topic.

Develop Question

It can be challenging to formulate the ‘right’ question for your topic, but it’s important that you spend time making sure the question:

Relates to what you really need to know about your topic
Is answerable
Is specific and focused
Is relevant to your field
Has not (recently) been answered by anyone else

Your question will guide your search strategy and data analysis, so it’s very important to figure it out early in the process.

Three types of questions were proposed by Eldredge (2002). These question types can also be related to more clinical questions:

Exploration/Etiology	(Why does D change with E in population F? Or What causes D to change in population F?)
Prediction/Prognosis	(Will X affect Y? Or How will X affect Y? Or How will X progress over time?)
Intervention/Therapy/Diagnosis	(Will J perform better than K at achieving outcome L?)

It’s important to work out what type of question you are asking.

A good question will combine several concepts (e.g. the D, E and F above). Identifying the relevant concepts is crucial to successful development and execution of your systematic search.

Break your topic down into constituent concepts. Depending on your topic, the PICO format may be used to identify the concepts:

Patient / Population / Problem
Intervention: cause, prognostic factor, treatment
Comparison /control (if appropriate)
Outcome

As well as PICO there are many other frameworks for conceptualising your question:

ECLIPSE - Expectation/Client group/Location/Impact/Professionals/Service (Evaluating services)

Wildridge, V., & Bell, L. (2002). How CLIP became ECLIPSE: a mnemonic to assist in searching for health policy/management information.

Health Information and Libraries Journal, 19(2), 113-115. doi: 10.1046/j.1471-1842.2002.00378.x

SPIDER - Sample/Phenomenon of Interest/Design/Evaluation/Research type (Qualitative studies, especially with samples rather than populations)

Cooke, A., Smith, D., & Booth, A. (2012). Beyond PICO: The SPIDER tool for qualitative evidence synthesis.

Qualitative Health Research, 22(10), 1435-1443. doi: 10.1177/1049732312452938

SPICE - Setting/Perspective (or Population)/Intervention/Comparison/Evaluation (Evaluating outcomes of a specific intervention)

Booth, A. (2006). Clear and present questions: formulating questions for evidence based practice,

Library Hi Tech, 24(3), 355-368. doi: 10.1108/0737883061069212

Identify Search Terms

Initial searches explore various keywords and subject headings and combinations of both.

Using key papers already known to you:

Find the record on a database such as PubMed
What words/phrases (keywords) does the author use in the title and abstract?
What subject headings are used on the record?

Search a relevant database and narrow your words to the titles of the articles. Look closely at the records:

What words/phrases (keywords) does the author use in the title and abstract?
What subject headings are used on the record?
Are there different subject headings to describe the same concept? Do you need to use several subject headings in your strategy?

Adapting the search strategy from a similar study:

Be critical
Amend for your topic

Search Filters

Search Filters – sit on top on your subject searches

Carry out your subject searches first, evaluate and make sure you are happy with the relevancy of the results. Then apply an appropriate filter to focus or narrow the results further. Filters are often inbuilt into databases menus and search screens. In PubMed Medline for instance you can refine searches from a menu of “article Types” or “Clinical Query” study types. In the CINAHL database on EbscoHost for example you can select from a variety of “Clinical Queries or research designs such as “Randomized Controlled Trials.”

Cochrane for instance presents some highly sensitive search strategies for identifying randomized trials in PubMed, and PsycInfo. Available at https://work.cochrane.org/pubmed/

Expanding Your Results

Using Core Articles

Once you have identified the core articles, you can expand your results further by using:

Reference lists: check out the reference lists in your core articles for other articles which may have been missed from your search strategy
Cited by: use the citation search in Web of Knowledge and SCOPUS databases and the cited by links in GoogleScholar to find out who has cited your core articles in their work
Related citations: a weighted algorithm is used to retrieve other items which are closely related to your item. Treat with caution, however, as the results may be on a tangent. (Not available in all databases)
Contact authors and organisations: for further information on their research
Handsearching: manually searching relevant journals to capture articles which may have been missed in a database search. See the Cochrane Handbook (6.2.2.1) for more details.[1]

[1] Lefebvre C, Manheimer E, Glanville J. Chapter 6: Searching for studies. In: 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 http://www.cochrane-handbook.org/