Preparing to search
Before you begin - are you doing a search for a systematic review or a scoping review?
Systematic reviews and scoping reviews are both part of the evidence synthesis family (a group of methods that also includes mixed method reviews, rapid reviews, umbrella reviews, realist reviews, evidence maps and more) and confusion can arise about which type of review should be undertaken.
If you are unsure about the differences between a systematic review and a scoping review (or any other type of evidence synthesis), there are tools which can help you distinguish the differences and decide on the right type of review for your research question. For example:
There are also numerous articles which cover this topic, including:
Defining your research question
It's important to have a well-defined research question and to be clear about your objectives when you start work on a systematic review. If your research question is too broad, you may be overwhelmed with a large number of irrelevant results; equally, if it's too specific you may miss relevant evidence. The School of Health Sciences have a useful introduction outlining why ‘Asking the right question’ can help in the search for evidence, and explaining how to construct good questions using the PICO technique discussed below.
Frameworks to help define your research question
You will first need to define your research question. Using a research question framework can help you structure your question and enable you to find the information you need most effectively. The PICO (Population/Patient, Intervention, Comparison, Outcome) framework is commonly used in systematic reviews and works well for reviews of effectiveness using a single comparative study design.
You can find out more about PICO from this 5-minute tutorial from the Cushing/Whitney Medical Library at Yale University. See also this example PICO which we have used as the basis for running a search in Medline (Ovid) and other bibliographic databases.
There are a number of different frameworks available, and the one you use will depend on whether your review plans to explore quantitative or qualitative evidence and the specific purpose of your review:
- PCC (useful for scoping reviews) - Population; Concept; Context.
- SPICE (useful for qualitative evidence synthesis) - Setting; Perspective; Intervention/Interest, of Phenomenon; [Comparison]; Evaluation.
- SPIDER (useful for mixed methods reviews) - Sample; Phenomenon of Interest; Design; Evaluation; Research type.
If your topic doesn't fit into a framework, that’s not a problem – just use the parts that do. Many clearly defined questions do not have a comparison or control to consider, so don't worry if this category isn't applicable to your topic.
Identifying your search concepts, keywords and synonyms
When thinking about your search, you will need to break down your research question into its key concepts. For example, if your research question is ‘Which complementary therapies work for acne?’, you should start by breaking your research question down into the separate concepts.
For this question there are 2 concepts: ‘people with acne’; and ‘complementary therapies’.
Next, try to think of as many different ways as possible to express the terms associated with each concept in your research question. Think about variant word endings, hyphenated words, different spellings and which way round terms might appear.
For the Population in the example above (people with acne), your list might include additional terms like acne vulgaris, comedones, pustules, etc.
For the Intervention (complementary therapies), you might consider including terms like alternative therapy, aromatherapy, homeopathy, holistic health, etc.
To help you break down the question into its key concepts and consider related synonyms, you might find a logic grid useful. You can also read our example of a Medline (OVID) search, which was developed using the PICO framework.
Refining your search: keywords, subject headings, Boolean operators, nesting, phrase searching, truncation, wildcards, proximity operators, limits and filters
Most of the searching techniques described below can be found in our example searches for key bibliographic databases.
Keywords are the natural language (free text) terms that you include in your search. You will need to identify keywords relevant to each of your search concepts then consider as many synonyms or related terms as possible, including commonly used abbreviations.
For example, if one of your concepts of interest is type 2 diabetes then you could consider including the following in the search: diabetes mellitus type 2, type 2 diabetes, type 2, diabetes mellitus, diabetes type II, diabetes mellitus type II, type II diabetes, type II diabetes mellitus, T2DM, T2D, non-insulin dependent diabetes, late onset diabetes, etc.
For an overview of all aspects of keyword identification, watch the video on Search skills: thinking about keywords.
Subject headings are standardised terms with consistent spellings; they may also be referred to as controlled vocabulary terms or thesaurus terms. They are used in some databases to index the content of articles, regardless of the keywords used by the authors or the language in which the article is written.
Including subject headings in your search helps to improve precision and the relevancy of the results. You can use them to supplement your keyword searches: an effective search strategy is one that uses a combination of subject headings and keywords for each of the key concepts.
If the database uses subject headings, you can search the thesaurus to locate narrower or broader terms. Exploding the subject heading means you will automatically retrieve records indexed with the narrower terms as well.
Subject headings include MeSH in Medline and Emtree in Embase.
Examples:
- the MeSH for type 2 diabetes is: Diabetes Mellitus, Type 2/
- the Emtree for type 2 diabetes is: non insulin dependent diabetes mellitus/
Boolean operators and nesting
Boolean operators (AND, OR, NOT) are used by databases to combine search terms, helping you to build your search strategy and refine your results.
Parentheses (brackets) can be used to group together words which relate to the same concept. Grouping words in this way is called 'nesting'.
Nesting can be used in conjunction with Boolean operators to create an effective search strategy.
For example, ("type 2 diabetes mellitus" OR T2DM) AND (exercise OR "fitness training" or walking or swimming) will find records that contain any of words/phrases in the first set of brackets and any of the words/phrases in the second set of brackets.
See this short tutorial on using brackets and Boolean operators.
When you enter two (or more) words into a database search, it will look for those words anywhere in the title/abstract/keywords, etc. However, if you want to force the database to run an exact phrase search, then you should enclose the words within quotation marks "...".
For example "diabetes mellitus" will ensure the database looks for exact matches of this phrase (the two words next to each other and in the same order), which in turn means the information found should be more relevant.
"complementary therapy" will find complementary therapy (note: it will not find the plural complementary therapies - see the section below on truncation).
See this short tutorial about phrase searching.
Truncation allows you to search for different word endings (singular, plural and other variations). The symbol most commonly used in databases for truncation is the asterisk (*) but check the database's help pages as other symbols may also be used.
Examples:
- "complementary therap*" will find complementary therapy or complementary therapies.
- exercis* will find exercise, exercises, exercising or exercised (note: be careful not to truncate the word too early – e.g. exer* would find the words already mentioned but also irrelevant words such as exercycle, exergonic or exertion).
See this short tutorial about truncating terms.
A wildcard symbol can be used to replace a character within a word or to add an extra (optional) character. This is helpful when you want to find US/UK spelling variants or irregular plurals. The symbols used for wildcards vary between different databases, so check the database's help pages first.
Examples (applicable to databases on the Ovid interface):
- wom#n will find woman or women.
- randomi#ed will find randomized (US) or randomised (UK).
- tumo?r will find tumor (US) or tumour (UK).
- behavio?r will find behavior (US) or behaviour (UK).
See this short tutorial about wildcard searching.
Proximity searching (also called nearness searching or adjacency searching) allows you to find two or more keywords which occur within a specified number of words of each other in the text, in any order. This type of searching is particularly useful when you have a number of phrases that may be used to describe the same thing.
Databases use different proximity operators (ADJ, NEAR, N) – check the database's help pages.
Examples (applicable to databases on the Ovid interface):
- complementary ADJ3 therap* will find the terms complementary and therap* within three words of each other, in any order, e.g. complementary and alternative therapies, or therapy as a complementary alternative.
- diabetes ADJ4 "type 2" will find diabetes within four words of type 2 (diabetes type 2, diabetes mellitus type 2, type 2 diabetes, type 2 diabetes mellitus, etc.).
See this short tutorial about proximity operators.
You can combine proximity operators and Boolean operators in the same line to make searches even more effective. This involves using parentheses (brackets) to create 'nested' search terms.
Example:
(complementary ADJ3 therap*) OR acupuncture will find records containing the terms complementary and therap* within three words of each other and then combine these records with any that contain references to acupuncture.
See this short tutorial on using brackets and Boolean operators.
A database may offer predefined limits which you can apply to your search (e.g. date, age group, study design, language). You should use limits with care and only if appropriate. For example, if you know a drug or intervention has only been available from a specific date, then you can justify the use of a date limit.
A better way to focus your search is to use a tried-and-tested search filter. For example, there are filters validated by Cochrane for finding randomized controlled trials (RCTs). You can copy and paste the filters line by line into your own search.
The King's College London website has some useful guidance on using limits and filters in systematic reviews, including links to pre-tested filters for RCTs, qualitative research, geographic regions, populations, etc.
Ready to search
The terms you identify during this initial preparation stage will form the basis of your strategy when you come to search bibliographic databases. It is worth spending time at this point:
- Clearly defining your research question;
- Identifying the major search concepts in the question;
- Developing a robust set of search terms, including truncated terms or those that need a wildcard; and
- Identifying the appropriate subject headings in databases which use a controlled vocabulary (e.g. MeSH in MEDLINE).
Using a combination of subject headings and keywords will ensure the most comprehensive set of results.