Building a literature search
A literature search is built from search terms, the logic that links them and the syntax of the tool it runs in. Database interfaces change that syntax over time, a general search engine can be restricted to a single site such as a social network, and some fields have no single database that holds their literature.
Boolean search logic
Boolean logic refers to relationships between search terms: it links two or more terms with one of the words AND, NOT and OR, to search for what they have in common or for its lack, and so limits or widens a search. For example, migraine AND treatment returns only articles about migraine therapy, while migraine OR treatment returns a much longer list that may include the treatment of hypertension and chronic hepatitis. The more terms are combined with AND, the shorter the list of articles, and the more are linked with OR, the more articles are retrieved. Many, but not all, websites allow Boolean searching.
Google allows Boolean searching and its default is AND, so pink monkey finds every document containing both words, while the phrase in double quotes finds only documents containing the phrase pink monkey. The same searching strategies apply in Google Scholar.
Search terms
Search terms define the limits and the nature of a literature search, so they should be established comprehensively enough to select all the related articles while eliminating those that are not relevant. The key concepts are turned into keywords, choosing only the most distinctive terms, and a thesaurus such as the Medical Subject Headings (MeSH) used to index articles for PubMed may be consulted to select keywords directly related to the topic.
PubMed can be searched by title word, phrase, text word, author name or journal name, or any combination of these, and allows Boolean searches. A search yields a list of citations to journal articles, with the authors, title, publication source and generally an abstract, and covers published scientific articles but not books.
PubMed can also be searched with the National Library of Medicine's controlled vocabulary, MeSH, which indexes articles from 5400 of the world's leading biomedical journals for MEDLINE/PubMed. MeSH is a thesaurus of descriptors arranged in a hierarchy that permits searching at various levels of specificity: its broadest headings are ones such as Anatomy or Mental Disorders, and narrower levels of its eleven-level hierarchy hold headings such as Ankle and Conduct Disorder. The 2016 MeSH had 27,883 descriptors. One physician author does without MeSH, finding Google, Google Scholar and PubMed sufficient for his needs.
Misspellings should be included among the search terms. Drug names are frequently misspelt by healthcare professionals, and spelling errors are common in databases such as Medline/PubMed. One study searched for gentamicin, amitriptyline and other drugs that professionals commonly misspell, using y instead of i or the reverse, and found that 18 variants of amitriptyline returned 179 hits that a search on the standard name alone would have hidden. The study advises truncated search terms: the text word am#tr#pt#l*, truncated at the letter l, uncovers variants of the last few letters, such as endings in lin, line, llin and lline, without sacrificing specificity, and gives further hits.
Search sites
One physician author thinks it best to become really familiar with just a few websites and learn to use them well, rather than cruise a large number inefficiently. In his view, whichever sites a searcher chooses, they must learn to use them themselves rather than depend on the expertise of a medical librarian, and understanding Boolean searching helps them search effectively. For a quick link to a phrase, a book title, an author's name or an unfamiliar website, the same author advises trying Google first.
Where the lists from PubMed and other sites seem to hold too much dead wood, the same author suggests BioMedLib, a free website that offers a much more focused response to a query than PubMed.
Brief tutorials on searching ClinicalTrials.gov, including one on setting up an RSS feed for a search, were available in 2010, with their instruction broken into segments of one to four minutes, and tutorials on searching PubMed were available at the same site.
Changes to the PubMed interface
PubMed's 2008 overhaul added two features: a beta version of an Advanced Search option, and Citation Sensor, which recognises combinations of citation terms such as volume and issue numbers, author names, journal titles and publication dates, and activates automatically when they are typed in the main search box. Its most controversial change, to automatic term mapping, was giving some users an unworkable number of hits, including irrelevant ones, on subject searches. The workaround was to use Advanced Search, or to type [tiab] after a subject search term so that only papers using the term in the title or abstract were returned.
In 2009 PubMed's Advanced Search came out of beta, and two changes to the interface were announced. The Single Citation Matcher was to disappear from the bar at the left of the screen; Advanced Search allowed searches by author, journal and publication date, and, through a further link, by title, volume, issue or page number. The Limits, History, Preview/Index and Details tabs were also to disappear, because those features had moved to the Advanced Search page.
In 2010 PubMed restored the option of e-mailing large batches of search results. From a page of results, the Send to link led to an e-mail option with choices of format, sort order and number to send, and an optional additional text that appeared at the top of the e-mailed file as a reminder for oneself or information for a colleague, such as "48 citations for asthma project". At most 200 results could be e-mailed at once, and a larger batch was sent in a second e-mail starting from citation 201.
By 2011 the Type of Article limit in PubMed allowed Video-Audio Media to be selected, so that a search identified articles accompanied by online videos and other audiovisual content. Two subsets, Dietary Supplements and Veterinary Science, had been added, and selecting a subset restricted a search to journals in that specialty and certain citations that PubMed's staff had identified in other journals.
Social media searches
Social media conversations can give insight into hot topics and industry trends and lead to people and key opinion leaders with experience of a subject, though it can be difficult to find exactly what and whom a searcher is looking for. Social media search results can help a writer learn what people are saying about a subject, identify trends and topics to write about, discover what content already exists on a subject, find experts and key opinion leaders, and use more relevant hashtags to market their content.
Most social media sites allow people to be searched by name; in 2016 Twitter and Facebook could be searched without logging in, while LinkedIn's search required being logged in. On Twitter and Facebook in particular, where the people sought have common names or do not use their real names, the results are not very relevant, and results always depend on users' privacy settings, which limit what is public whatever the search tool.
Google's advanced search commands search an individual social media site without logging in to it, or even having an account. A search string that starts with site: followed by the site and then the search terms searches that site, inurl: limits the terms to web page addresses, and quotation marks give an exact phrase match. In 2016, site:www.YouTube.com Pfizer returned multiple YouTube channels associated with Pfizer, site:.com inurl:medicalwriting searched all sites ending in .com that include medicalwriting in their address, and site:.com "medical writing" searched all such sites whose content includes the phrase. A space after the colon removes the limitation by site or address. Any of these searches can also be restricted to a specific period of time, as with zika inurl:blog followed by the Past 24 hours or Past hour option under Search Tools.
UVRX Social Search uses Google's custom search technology to let a searcher choose which of a set of social media sites to search, LinkedIn among them, without logging in to LinkedIn or having a profile there. A 2016 search of it for medical writing returned 29 LinkedIn results among the first 30, some of them content that the searcher's other searches had not found.
Fields with no single comprehensive database
A veterinarian who writes regulatory dossiers for veterinary pharmaceuticals thought in 2021 that the veterinary field had no data source comparable to PubMed. She paid for access to a veterinary pharmaceutical database, knew of no veterinary database containing all the field's information, and said that a searcher would probably go to different sources depending on the kind of search. She sometimes used Google Scholar, because it very often found good articles in less common journals, since, she assumed, not all veterinary journals are in PubMed. Another participant in the same 2021 discussion named CAB Direct as the best database for veterinary journals, covering about 80% of them.
In her experience, terms such as cats or feline were very useful, also in PubMed, where they filtered well for animal or pet material, and she thought it really important to use the right keywords. For her daily work PubMed and Google Scholar worked well, but for regulatory text a searcher should know which pages to go to.