Literature Searching and Reviews

In addition to conducting the searches correctly, it is also important to search multiple databases and other sources of information to identify as many relevant studies as possible. Selection on databases and information sources depends on the topic.

Bibliographic databases
The most commonly used bibliographic databases in e.g. health sciences are:  PubMed, Medline (Ovid), Embase (Ovid) and Cochrane Library.  Interdisciplinary bibliographic databases are e.g. Academic Search Premier

Topic specific databases
There are subject-specific (bibliographic) databases with a specific focus, such as Cinahl (Ebsco), MathSciNet (Ebsco), Business Source Complete (Ebsco) and EconLit (Proquest).

Multidisciplinary search engines

More interdisciplinary search engines includes: Web of Science, Scopus and Google Scholar. . Web of Science and Scopus register citations and can be used to conduct a forward citation search, identifying which publications have cited a given relevant publication.

In addition to the traditional databases and search engines, there are other sources of information and literature that may be relevant.

Scientific databases often use controlled subject headings, and these are primarily associated with and describe reviews in the databases.
In health science, PubMed and Medline (Ovid) are examples of this and they use MeSH (Medical Subject Headings). Embase (Ovid) uses Emtree thesaurus and PsycINFO (Ovid) uses Subject Headings see more about the Ovid databases on https://libguides.sdu.dk/OVID 
Cinahl and Business Source Complete and other Ebsco databases uses Subject Headings.

Other databases use controlled keywords as well, e.g., Academic Search Premier has Subject Terms and Business Source Complete has Thesaurus.

The advantage of controlled keywords is that they link all articles on a topic, regardless of synonyms or alternative spellings - this is their very purpose.

Databases such as Web of Science and Scopus do not use controlled keywords.

The weakness of controlled keywords is typically that they are automated and do not allow for long and thorough considerations and are not always based on sufficient professional knowledge in the field.

In practice, searching using controlled keywords alone will result in very high precision but only medium recall of the relevant literature.

If you want to conduct a literature search that provides 'the most important information about...' or 'the essential information about...', it is fine to simply use the controlled keywords. However, if you want to find all literature on a given topic in a database (increase recall), it is necessary to use free text search as well.

Free text words are typically searched in title, abstract and keywords, but there can be huge differences in which fields you search for in different databases. You can often limit the free text words to search only the title or abstract (or both at once).

The advantage of free text search is that you use the authors' own word choice. Unfortunately, authors do not necessarily have the same agreement on terms, abbreviations, etc. It may therefore be appropriate to include all the technical terms used by the authors if you want a high recall of your search - that is, if you want to find the most relevant literature possible. 

Optimal free text searching may require knowledge of the structure and content of bibliographic databases. The classic scientific databases, such as PubMed, Web of Science, etc. mainly index only the author(s), title and abstract from the original articles. Depending on the database, subject and citation data may also be included. Newer search engines such as Google Scholar also automatically (albeit partially) index the content of the full text of the articles themselves. This enables a much more thorough search for more specific text or data strings. The problem is that the extent of this automatic indexing process is kept hidden from users, unlike the classic databases whose indexing process is clearly and completely declared.

Boolean operators

OR: The Boolean operator OR is used to include keywords/terms that are considered to be synonyms or closely related. This means that there must be at least one of the entered keywords, e.g. crime OR violence. Entering this into a database will return search results containing either crime or violence or both crime and violence.

AND: The Boolean operator AND is used to combine keywords / subject terms that both occur. This means that both of the entered keywords must be mentioned, e.g. crime AND abuse. Entering this will return search results containing both crime and abuse.

NOT: Using the Boolean operator NOT will omit search results that contain the following keyword/phrase. This means that if you enter crime NOT women, only search results containing crime but not women will appear.
It is rarely appropriate to use NOT, as this function is unfortunately very likely to exclude relevant literature.

Truncation and wildcards refers to the use of * ? # or other characters used to include all endings or compound terms of a search term. For example, crime*, which returns search results that include crimes, criminology and other similar terms.
This method can be used in multiple databases. To see which characters are used in the different databases, it is recommended that you read the guide or help function in the individual databases.

Phrase searching: Search for words that are compound terms or must be in a specific order with quotation marks around the words, e.g., "artificial intelligence". You will only get results for words in that exact order.

Be aware, however, that using quotation marks gives different results depending on which database you are using. For example, it will disable controlled keywords (MeSH) in PubMed, whereas phrase searching is a default option in Ovid databases. In Scopus, on the other hand, you should always use quotation marks for compound terms/phrases.

Proximity refers to how close keywords in a term can be to each other, e.g. heroine addicted. Here, the term must be grouped together. By using proximity operators, you allow the words in the search term to not necessarily be completely together, but other words can be placed in between these two. This allows for a more flexible search, which can be useful in some cases.

The exact proximity operator varies from database to database. In the OVID databases, the proximity operator ADJ is used. For example, if you search for heroine ADJ4 addicted, you allow the word heroine to occur within four words of the word addicted (the order does not matter).
In the EBSCO databases, N is used with number for how many words can be between (regardless of the order of the words). The default setting allows five words between words, for example heroine N2 addicted.
In other databases e.g. ProQuest, the proximity operators can be NEXT/ EXACT, NEAR or PRE.

Search filters are used in addition to database filters or limit functions. In some databases limits may be inadequate or not have the content that some people are looking for.

A search filter is a search string that can capture certain study types or specific topics.

Be aware that using search filters can result in a number of relevant studies being omitted from the search results (Cooper, Varley-Campbell, & Carter, 2019; Leeflang, Rutjes, Reitsma & Bossuyt, 2006). Therefore, search filters should be reviewed and possibly evaluated before applying them (Glanville et al., 2008). There may be selected databases that are described in detail regarding search filters (Glanville et al 2019 and 2020). Furthermore, factors such as sensitivity and precision should be considered when selecting a search filter (Beale et al., 2014).

If you want to use a search filter, we refer to the websites below, which can be used for inspiration, they are not necessarily validated.

References on search filters:

Beale, S., Duffy, S., Glanville, J., Lefebvre, C., Wright, D., McCool, R., . . . Smith, L. (2014). Choosing and using methodological search filters: searchers' views. Health Info Libr J, 31(2), 133-147. doi:10.1111/hir.12062

Cooper, C., Varley-Campbell, J., & Carter, P. (2019). Established search filters may miss studies when identifying randomized controlled trials. Journal of Clinical Epidemiology, 112, 12-19. doi:10.1016/j.jclinepi.2019.04.002

Glanville, J., Bayliss, S., Booth, A., Dundar, Y., Fernandes, H., Fleeman, N. D., . . . Welch, K. (2008). So many filters, so little time: the development of a search filter appraisal checklist. J Med Libr Assoc, 96(4), 356-361. doi: 10.3163/1536-5050.96.4.011.

Glanville J, Foxlee R, Wisniewski S, Noel-Storr A, Edwards M, Dooley G. Translating the Cochrane EMBASE RCT filter from the Ovid interface to Embase.com: a case study. Health Info Libr J. 2019;36(3):264-277. doi:10.1111/hir.12269

Glanville, J., Kotas, E., Featherstone, R., & Dooley, G. (2020). Which are the most sensitive search filters to identify randomized controlled trials in MEDLINE?. Journal of the Medical Library Association, 108(4), 556–563. doi:https://doi.org/10.5195/jmla.2020.912

Leeflang, M. M., Scholten, R. J., Rutjes, A. W., Reitsma, J. B., & Bossuyt, P. M. (2006). Use of methodological search filters to identify diagnostic accuracy studies can lead to the omission of relevant studies. J Clin Epidemiol, 59(3), 234-240. doi: 10.1016/j.jclinepi.2005.07.014.

Grey literature
Grey literature refers to literature such as research reports, dissertations and conference papers.

Some databases contains conference contributions and dissertations as Scopus, Web of Science and Proquest Dissertations & Theses Global, but it is possible to find grey literature in Google Scholar, SSRN Working Papers or OpenGrey. If you are using Google Scolar, you can search from Publish or Perish, which allows you to download articles (set it to a maximum of 200) and makes the search more transparent. You can then transfer the results to e.g. EndNote (choose RIS format), get the articles updated, and then transfer them to Covidence.

Preprint servers
A preprint refers to a version of a scientific article that precedes peer review and publication. From different subject areas you can find e.g.. medRxiv (health science), bioRxiv (biology) and SocArXiv (social science). Within Humanity there is Hprint, However, preprint servers are not quite as widespread. A list of preprint servers can be found at Wikipedia, but it's not necessarily comprehensive.

Updating database Searches McGill