Systematic reviews: finding and managing the evidence

Top 10 tips for database searching

1. Take time to gather your search terms in a list for each concept in your search before you begin searching. Think about alternative spellings, terminology and the different ways topics are referred to, such as synonyms or colloquialisms. Be aware of UK (paediatric) and US (pediatric) spellings and include those as well. Your aim to produce a highly sensitive search in order to not miss relevant papers.

2. Research which databases to search and search each database separately

3. Search for each concept in your question separately

4. The optimal search method for systematic reviews is to, where possible, perform both text word and subject heading searches for each concept

5. Learn how to combine terms together using AND and OR

6. Take time to learn about and use advanced search techniques such as truncation, wildcards and proximity searching

7. Remember that searching is an iterative process. Be prepared to test, validate and refine your search strategy to achieve the best possible search to underpin your systematic review

8. Perfect your search in one key database such as Medline or Embase,  before moving on to other databases, where you may have to tweak your search slightly to reflect different subject headings used or database syntax

9.Save each search strategy as you go along so you can document your search at the end

10.Decide in advance how you are going to manage your search results- in Word, Excel or reference management software such as RefWorks, Mendeley or Endnote.

Identifying your search terms

Searching for research evidence is like shopping: if you don't make a shopping list before you start shopping, you may forget to buy something essential you need. When searching for evidence, if you don't make a list of your concepts and terms, then you might miss some relevant information that you need. Preparation is key.

In the table below you will see a list of terms expanding on our earlier P I C O S framework. Note how for each of the concepts we have developed a list of alternative words and phrases- these terms may form the basis of the search strategy.

PICOS framework in practice

Patient/Population	Teenage Diabetics, Teen diabetics, adolescent diabetics, adolescence and diabetes, adolescents and diabetes, teenagers and diabetes
Intervention	Patient education, patient information, health education, health information
Comparison	No comparison
Outcomes	Well managed blood glucose levels, blood sugar levels, hypoglycaemia, hypoglycemia, hyperglycaemia, hyperglycemia, hypoglycaemic, hypoglycemic, hyperglycemic, hyperglycaemic
Study	Any studies

As a general rule, using a SINGLE term for each of your concepts is insufficient to capture all of the relevant papers. At the same time, generating a list of terms can be daunting. Luckily there are a number of strategies you can employ to help with this aspect, as outlined below:

• Examine the titles and abstracts of known relevant papers to identify more terms
• Use truncation to search for plurals and different endings, thus streamlining your search
• Identify indexing and controlled vocabulary terms (known as 'subject headings')
• Check with your colleagues, experts, books or even Wikipedia to discover alternative terminology and refine the different parts of your search. For example, 'cardiac rehabilitation' consists of education, exercise therapy and counselling, so all three of these components would need to be searched to identify articles on 'cardiac rehabilitation'
• Use freely available Frequency Analysis software such as Write Words Word Frequency Counter to help generate a list of frequently used terms. Cut and paste 200-300 records from an initial list of results downloaded into word and hit submit to get a breakdown on terms appearing in the text and their frequency.

You'll find more details on the above strategies in the sections on text word or subject heading searching.

Selecting your databases

Once you have created your list of search terms, think about which databases you need to search; as a rule you should search as many databases as possible in order to target as wide a pool of articles as possible.See below for an overview of databases available, all of which can be accessed via the library website http://library.sgul.ac.uk

Search modes

There are two modes of searching databases:

1. Text word searching, also known as free-text or keyword searching

2. Subject heading searching, also know as thesaurus or controlled vocabulary searching

For systematic reviews, it's important to employ both modes when searching on your terms for all of  your concepts, to make the search is comprehensive as possible.

See Steps 3 & Steps 4  in the above tabs for more information about these two modes of searching.

Text word searching

In this search mode, the database will search papers for the words you type into the search box.  Searches are normally limited to specific fields such as title or abstract or across the whole paper. This is also known as keyword or free-text searching.

Once you have generated a list of terms using frameworks such as PICOS or SPIDER, you then need to decide which of these  text words will ultimately become part of your search strategy.  It's possible that for some searches  you may decide not to search on every term identified in your framework due to redundancy  and duplication or you might want to use the terms as guidance for when you are evaluating references for inclusion/exclusion.

As discussed earlier, it remains important to think of all alternative terms such as spellings, plurals and synonyms related to each of the concepts  that you decide to search on.

 This ensures that you capture all the variant ways your topic may be discussed and described in the literature by different authors. The advanced search tips below can help you with this aspect of searching databases.

---

5 Advanced search tips

Tip 1: Use Truncation

Truncation is where you use a symbol, such as an asterisk *,  to represent alternate word endings. For example, typing nurs* would mean that your search would bring back articles containing nurse or nurses but also nursing, nursed, nursery etc.

Using truncation helps broaden your search without you having to type lots of variations of the same word, but be careful when using it to make sure you are not overwhelmed with too many irrelevant papers.

Tip 2: Use proximity searching

You can use special commands in databases to specify how close words appear to one another. This is known as proximity or adjacency searching and can aid in increasing the sensitivity of your search.

Some examples below include:

• To search for a distinct phrase of 2 or more words, put it in "quotation marks" to tell  the database to only find articles with those words next to each other, in that order.

• Adj(number)  (Ovid or NHS HDAS platforms)  e.g. acute adj5 "otitis media" will find the term 'acute' within 5 words of "otitis", in any order.

• N(number) (Ebsco platform) e.g. "middle ear" N2 inflam* will find the phrase 'middle ear' within two words of 'inflamed' or 'inflammation' or 'inflammations', in any order.
  
  If you're not sure which command to use, check each database platforms help pages.

Tip 3: Use Wildcard Symbols

Wildcard symbols  (normally ? or $) stand for zero or one character.  They are useful for picking up American vs UK spellings of words or other word variants.

Examples include the following:

p?ediatric = pediatric or paediatric

 wom?n = woman or women

Tip 4: Use Field Labels or Field Searches

Specify where the database looks for your terms by either limiting the search to  particular fields or using field label identifiers; for example in Ovid,  you can use the following field label identifier  to only find terms that appear in the title or abstract field (text words in Ovid): .tw. 

Tip 5: Use Brackets

Use brackets to group together (similar or synonymous) search terms together and run the search in a logical manner. 

Using some of the above tips, your text word search strategy in Ovid could look something like this:

Subject heading searching

This mode of searching is when the database maps your terms to its own unique, controlled vocabulary that is used in a standardized way to tag and identify the contents of articles held within a database. They can help your search immensely as they use one consistent term for any given concept and are used in a very precise way. In Medline, the controlled vocabulary system is called MeSH which stands for Medical Subject Headings.

To Map in Ovid, first make sure the box next to 'Map term to subject heading' is selected before you click search once you've typed your search term; mapping works best if you enter a  simple word or phrase such as 'heart attack' without all the additional advanced search techniques described in the section on Text word searching:

 

As you will see below,  'heart attack' maps to 'Myocardial Infarction' in Medline:

Subject headings come from a hierarchically arranged thesaurus- for example MeSH in Medline-allowing you to explore associated terms in a subject tree around your topic.   Click on a subject heading  such as 'Myocardial  Infarcation' to see narrower subordinate terms that can also be included in your search:

When you find a relevant subject heading, tick the Explode (to include all narrower relevant subject headings) and Include All Subheadings (to include all qualifiers for your tem) boxes to make your search as comprehensive as possible:

 

If you are unsure about how to carry out a subject heading search in a particular database platform, check the help pages. Note that not all databases will have subject headings available to search, for example,Web of Science (WOS). In these cases you will have to rely on text word searching alone.

Before you carry out your final search, it’s useful to scope out a comprehensive list of potentially relevant subject headings. You can do this by searching for a few relevant papers and looking at which headings have been assigned to describe their content. Examining the Scope note attached to each subject heading  within the database is also a very useful way of capturing alternative or narrower terms it is used for and for identifying any previous indexing term used for your concept:

Combining search terms using OR or AND

OR & AND are search operators used by databases and search engines to group search concepts together.  The diagrams below illustrate the use of AND or OR.

OR is used to combine similar terms together  - this tells the database to show you articles where any of your search terms appear, thus broadening the scope of your search and giving you more results:

AND is used to combine different concepts together- this tells the database to only show you articles where all of your search terms appear, thus narrowing your search:

This is also known as Boolean searching or boolean logic.

In addition, you can also use NOT to eliminate terms and concepts that are not of interest you which are significantly throwing your search out of context; however, use NOT with caution to avoid inadvertently eliminating papers you might need.

Putting your text word and subject heading searches together

The optimal search for each concept will be a combination of a text word and subject heading searches for each element combined at the end with AND.  To achieve this, follow the steps below, but note that we are using a very simplified version searching for information with very few alternative terms.

1. First, for each of your text words, you need to attempt to find a corresponding subject heading, then  combine both text word and subject heading for that element of the search together with OR. Below, we use the 'Combine with option'  that appears at the bottom of your search history to combine the similar searches for articles about heart attack with OR:

2. Second, you need to combine the different OR set of searches with AND to get a final set of results, as illustrated below in a simplified version of this search:

 

Refining & limiting your search

Searching is an iterative process and you might find you’ll need to try different combinations of your search terms to improve the relevancy of your results. Try applying Limits (as shown below) or additional keywords if you are retrieving too many results and want to focus them, or adding alternative terms if you are retrieving too few. As a rule, however, try not to use too many database limits so as not to reduce the numbers of results to too small a pool.

If your results aren’t quite relevant, check to see if any of your terms are throwing your search off track.

If you find useful articles, look at the text words and subject headings used in the articles and build them into your search if you haven’t already included them.

Search filters

If finding clinical trials or other study designs is important to your review, there are a number of tested search filters available to help you hone in on the particular types of papers you need.  Below are some links to some key search filter resources. An example of a filter to find RCTs in Ovid Medline is below:

1 2 3 4 5 6 7 8 9 10 11 randomized controlled trial.pt. controlled clinical trial.pt. randomized.ab. placebo.ab. drug therapy.fs. randomly.ab. trial.ab. groups.ab. 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 exp animals/ not humans.sh. 9 not 10  Cochrane Highly Sensitive Search Strategy for identifying randomized trials in MEDLINE: sensitivity-maximizing version (2008 revision)

 

Further guidance on search filters can be found on the following resources:

The InterTASC Information Specialists' Sub-Group Search Filter Resource
ISSG Search Filters Resource (google.com)

Cochrane Filters
4.S1 Technical Supplement to Chapter 4: Searching for and selecting studies | Cochrane Training 

SIGN Filters:
Search filters (sign.ac.uk)

Methodological Filters- compiled by Library staff from University of Alberta- easy to cut and paste filters

 Methodological Filters - Health Sciences Search Filters - Subject Guides at University of Alberta Libraries (ualberta.ca)

In all database platforms search results can be saved, emailed and exported, using the options available. Select the references you wish to save using the tick boxes to the left of the article title, or  find options to save all of your results.

The Print, Email and Export options will allow you to customise the amount of information you want to save/display. You can also opt to include a record of your search history, as well as any full-text links to articles if they are available via St George’s Library or NHS OpenAthens.

The export feature typically will allow you to export to reference management software such as RefWorks, Mendeley or EndNote as well as to other formats such as a Word document or Excel spreadsheet.

Saving your search history will allow you to run it again at a later date, or you can save it as an alert to receive email updates when new papers are published on your topic. Saving your search strategy will also be important when you want to document your search.

When saving your search strategy, you may be  be prompted to login to a personal account – if you don’t already have one you can create one at this stage. This is a really useful option if you are working on the search at different times or working on different versions and you don't want to lose your work. You can run, edit or delete your searches and alerts by logging back into your database personal accounts.

In Ovid, for example, simply select the ‘Save all’ button underneath your search history. You’ll be prompted to login to your personal Ovid account

You can run, edit or delete your searches and alerts by logging into Ovid, selecting ‘My Workspace’ and then clicking on ‘My Searches & Alerts’.

Translating your searches to other databases

Once you find are happy the search is giving you what you need, apply the same search strategy to other databases. It's not possible to safely rerun your search from, for example, Medline Ovid to Embase Ovid however tempting, so you will need to start from scratch in each new database and following the same method of creating and combining text word and subject headings for the different elements in the search. In Ovid, click Change to select a new database: