Engaging medical librarians to improve the quality of review articles.

Review articles published in JAMA summarize various aspectsofmedicalpracticeandarewrittenbyknownauthorities.Anexpert’sopinionabouta topichasvalue,but often more can be learned by that expert’s assessment ofall thepertinent literature.Whenreading individual research articles, readers couldmiss subtle features of the studies that are more apparent to an expert clinicianresearcher. Readers benefit from the expert’s explanation of the validity and applicability of individual studies. When authors summarize the literature, a natural tendency is to select articles supporting their views. To minimize this potential for bias, JAMA editors ask authors to systematically review the literature and comment on as much of it as is practical. Extensive literature searches can be difficult to perform, given the complexityof the searchprocess andauthors’ timeconstraints. It is also unlikely that a content expert will be familiarwith the intricacies ofmore than one database. Empiricalevidenceprovidesastrongrationale forsearchingmultiple databases.1,2 For this reason, collaborating withmedical librarians is recommendedwhenwritingreview articles. A general approach for this collaboration is presented in the Box. Preliminary searches conducted by a librarian can help refine thequestionanddetermine its feasibilityand scope. These initial screening searches can also inform the development of the review’s prospective inclusion and exclusion criteria. The choice of database depends onseveral factors.Most important is thecontent (topic) of database. MEDLINE is produced by the US National Library of Medicine and focuses on articles in peerreviewed journalsofbiomedicineandhealth.Embase includesbroadercoverageofdrugsandpharmacologyand conference abstracts. Of the 4800 journals indexed in Embase, 1800arenot indexed inMEDLINE. Similarly, of the5200 journals indexed inMEDLINE, 1800arenot indexed inEmbase.3 CINAHL is anexcellent source for researchofnursing,alliedhealth,or interprofessionalareas. PsycINFO is the primary database for literature in psychology, psychiatry, counseling, addiction, and behavior. The Cochrane Collaboration’s CENTRAL is a critical database forusebycomparativeeffectiveness researchersbecausethisdatabaseattempts tocaptureall therandomizedclinical trials inmedicinethroughextensiveelectronic and hand searching of journals and conference proceedings. Regional and national databases are important in searching for certain topics (eg, searching LILACS, a database from Latin America and the Caribbean,when evaluating a tropical disease; searching the Chinese Biomedical Literature Databasewhen evaluatingacomplementarymedicinetopic).Searchingtrial registries (eg, ClinicalTrials.gov) reveals ongoing and asyet unpublished trials. The second consideration involves the database platform (ie, the vendor or interface/search engine). Forexample,MEDLINE isavailablewithoutcost through PubMed but can also be accessed through a vendor such as Ovid. Embase is only accessible through an institutional subscription via 1 or more vendors (eg, Embase.com, OvidSP, ProQuest). Each of these platforms offers different algorithmic indexing and search capabilities. PubMed, for instance, does not search for all phrases,whereasOvidSPMEDLINEwill search forany combination of words as a phrase, and can use sophisticated adjacency searching, identifying abstracts in which a word is mentioned within a certain number of wordsof another. PubMed,however, includesmoreupto-date recordsand records fromPubMedCentral, such as manuscripts from National Institutes of Health– funded research published outside of the journals indexed byMEDLINE. Once sources are selected, librarians reduce the research question to major concepts to determine appropriate subject headings or other controlled vocabulary terms for each database. Controlled vocabulary is the most valuable strength of databases like MEDLINE, Embase, CINAHL, and PsycINFO. It standardizes terminology and broadens a searcher’s ability to find all relevant materials on a given topic. Many controlled vocabularies are arranged into hierarchical trees to allow searching for a broad term (eg, heart diseases) and retrieving articles about all narrower concepts (eg, myocardial infarction or cardiomegaly) without having to search for each term individually. An expert searcher looking for information on stress management methods for patients with cancer would most likely have more success with CINAHL’s patient-centric controlled vocabulary terms than with MEDLINE’s disease-focusedMedical Subject Headings (MeSH terms). Embase has superior drug indexing via its Emtree vocabulary, allowing searchers to swiftly focus on drug administration routes as well as search more comprehensively for drugs overall. The unique terminology and controlled vocabulary structure give human-indexed databases their power, enabling faster, more comprehensive searching than that achieved through the use of solely computer algorithm–indexed databases like Scopus or even through the use of search engines that search full-text articles, like Google Scholar. In addition to controlled vocabulary terms, librarians also can determine what terms to use as free-text keywords, such as words in article titles or abstracts. These must be carefully considered because they can generate extensive search results that are difficult to manage.