| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Features Section |
1 University Dental Hospital of Manchester, UK
2 Liverpool Dental Hospital, UK
S. R. Bickley, Cochrane Oral Health Group, University Dental Hospital of Manchester, Higher Cambridge Street, Manchester M15 6FH, UK. Email: Sylvia.r.bickley{at}man.ac.uk
Introduction
One of the most important and possibly time-consuming areas of research is literature searching. The availability of electronic databases allows us to search and access the results very quickly, but the quality of our search will be dependent upon a number of factors. In this paper we will provide information on some of these and consider how they impact on search results. We hope that this will be helpful and lead the reader to develop a more systematic approach to literature searching.
Electronic searching
At first glance, searching electronic databases appears to be an easy way of tracking down literature, but you should be aware that even the most experienced searchers will miss some relevant literature. For example, it has been shown that when a search is carried out for controlled clinical trials approximately half of the relevant trials on a topic may be missed in an electronic search, even though most of the missed citations are in the databases!1
Electronic searches chiefly rely on two things:
There is, fortunately, an increasing trend for journals to use structured abstracts. In this format the author systematically describes the objective, design, setting, subject, interventions, outcomes, results, and conclusions of a study. This should improve indexing of records and the quality of electronic searches in the future.2–4 Structured abstracts are also one of the recommendations for the CONSORT guidelines.5
So how do you do a search?
An effective search will involve the following steps:
Identifying the sections of your research question
It is important to avoid the temptation to start searching before properly thinking through appropriate search terms for the question. A structured search strategy will be more controllable, and easier to check for inclusions and omissions than a jumbled list of search terms. Time spent identifying search terms and organizing them into the framework of the search strategy will be time well spent, and will save having to trawl through masses of papers that are not relevant to the research topic.
First, identify key section headings from the research question. These might include participants, their condition of interest, the intervention or exposure of interest, and outcome. Under the section headings, list appropriate terms to create the structure on which the search strategy will be built. An example of this can be seen in Figure 1, where the subject of the search is the effectiveness of powered versus manual toothbrushes for orthodontic patients.
|
Most of us will probably start searching with MEDLINE. This is the US National Library of Medicine’s (NLM) premier bibliographic database that contains over 11 million references to journal articles and can be accessed free of charge via website http://www.ncbi.nlm.nih.gov/PubMed. It should be stressed that effective searching is an art that can only be learned through practice, and gaining knowledge and understanding of the rules that must be applied to searching individual databases. Information and guidance can be obtained from medical libraries or database providers’ point of access, e.g. NLM, PubMed, OVID, Silver Platter, etc., via Internet webs sites or directly from the database help files. Ideally, a good first step is to seek the guidance of a medical librarian or information specialist. Their expertise in literature searching, combined with the clinician’s subject knowledge, will provide the complementary skills needed to build the most appropriate and effective search strategy.
The mechanics and technicalities of electronic searching
The search terms should include controlled vocabulary and free-text terms. Controlled vocabulary refers to the subject headings (indexing terms) that are used in electronic databases. Some databases, such as MEDLINE, EMBASE, and CINAHL, use subject headings that are arranged in a hierarchically structured format like the branches of a tree, with broader concepts near the top and more specific terms lower down. In MEDLINE this is called the MeSH® Tree (MeSH standing for ‘Medical Subject Headings’) and in EMBASE–EMTREE®. Other databases may use a structured thesaurus of concepts arranged alphabetically. Using the hierarchically structured trees or thesauri will allow you to broaden or narrow your search. You will also find the NLM’s MeSH Browser, which has full details of MeSH terms, their indexing and hierarchy, a valuable resource. This can be found at www.nlm.nih.gov/mesh/MBrowser.html or a hard copy may be found in the reference section of medical libraries.
The section of the MeSH tree presented in Figure 2 shows a dental example where the broadest subject heading is ‘orthodontics’ with more specific headings in the branches below. Any point in the hierarchy can be searched to include the terms indented beneath it by applying the ‘explode’ function in the database’s search engine. Linking the instruction ‘explode’ to the term ‘orthodontics’ would automatically include all the terms presented in the orthodontic hierarchy below. Alternatively, the search can be focused on a specific area of interest within orthodontics by applying the ‘explode’ function to a branch lower down the tree. For example, applying ‘explode’ to the subject heading ‘orthodontic-appliances-functional’ would focus the search to retrieve only articles indexed with the terms ‘orthodontic-appliances-functional or activator-appliances’ (Figure 2).
|
However, do not be misguided into thinking that using free-text only is a ‘catch all’ option, because the search will be confined to finding the exact match of the text word/s in the title, abstract, or indexing field of the reference. Your search will also miss things that exploded indexing will pick up. For example, take the phrase ‘maxillofacial abnormalities’; used as free-text this will only pick up articles where this exact phrase appears. On the other hand, if the same phrase is used as an ‘exploded’ MeSH term, not only will articles be picked up where this term appears, but also articles containing the subordinate terms beneath it in the MeSH tree as presented in Figure 3.
|
Truncators and operators
‘Truncators’ can be used to expand search terms, and ‘operators’ to include or exclude specific terms and locate terms that are close to each other. If a truncation symbol is inserted at the end of a free-text word it will expand the search to retrieve multiple suffix variations of the word. An example of this would be when placing a truncation symbol (in this example an asterisk) at the end of a word, e.g. orthodontic*, which will also retrieve articles containing the words orthodontics or orthodontically (see Figure 4a).
|
give examples. Placing the operator ‘NOT’ between two terms will have the search retrieve papers that contain the first word, but not the second term. For example, ‘neoplasm NOT breast’. Used in a similar way, proximity searching, using the operators ‘NEAR’, ‘NEXT’, or ‘ADJACENT’ can be helpful for focusing a search. Be aware that operators might be used or expressed differently across databases.
Building the search strategy
Returning to the example in Figure 1 and the list of controlled vocabulary terms, (identified by upper case text) and free-text terms (lowercase), the next stage is to enter the search strategy into the search engine. In the worked example the search was applied to the Cochrane Controlled Trials Register (CCTR). Figure 5 shows how the search terms from the initial framework have been presented and linked using truncators and operators. Once all the search terms or strings from one section have been searched line by line, the results of these are then ‘OR’d’ together to combine the search for the section (lines 3, 8, 11, and 21). When all the sections have been searched, the cumulative searches for each section are linked with the operator AND (line 22). This focuses retrieval of citations by ensuring that at least one search term from each section will feature in the retrieved citations and should therefore be relevant to the research topic. Remember though, that the more sections you include in your search framework, the more focused the search will be. To protect the sensitivity of the search, keep the search framework to essential section headings and ensure that a wide range of search terms for each section is included.
|
Running, reviewing, and revising your search strategy
Once the search strategy has been developed and run, the retrieved articles should be examined to identify whether the search needs revision or not. In general, if the search produces large numbers of irrelevant references it should be refined further. Likewise if fewer references than expected are retrieved revise the search strategy to increase sensitivity.
What to search: choosing the best resource for your needs
It must be appreciated that no single resource can be relied upon to provide all the evidence. Where thoroughness and search quality are important, searching a range of resources should be considered. By no means all healthcare literature is indexed in MEDLINE: non-English language references are under-represented and only published articles are included. This, in itself, may introduce publication bias if studies with positive results are selectively published.6
Evidence-based dentistry research
Most questions associated with applying evidence-based dentistry to clinical dentistry will revolve around the effectiveness of competing interventions that are provided for patients. To answer these questions, up-to-date systematic reviews of randomized controlled trials (RCTs) are generally accepted as being the most reliable source of evidence. The best place to search for systematic reviews of RCTs is the Cochrane Database of Systematic Reviews (CDSR), one of the electronic databases in The Cochrane Library (see below). If there are no relevant, up-to-date systematic reviews in a particular area of interest, consider evidence contained in the individual reports of RCTs. The Cochrane Controlled Trials Register (CCTR), also available in The Cochrane Library, is recognized as being the best single source of such reports.7 CCTR is the result of a number of on-going projects within the Cochrane Collaboration worldwide8 and brings together, in one bibliographic database, citations to controlled clinical trials from across the world (currently, in Issue 4, 2002, there are 345,378 citations). Updated quarterly, it is sourced from searching a wide range of electronic databases, conference proceedings, and trials found only through the Cochrane Collaboration’s world-wide journal hand-searching programme.
The Cochrane Library is available through many university libraries and postgraduate medical centres. All residents in England, Wales and the island of Ireland can access The Cochrane Library via the internet (www.cochrane.org) free of charge thanks to government funding. Other countries that have free access are listed on the website. The Cochrane Library is also available on personal subscription on CD or via the Internet [details available from Update Software Ltd (Summertown Pavilion, Middle Way, Summertown, Oxford OX2 7LG, UK) or Update Software Inc. (1070 South Santa Fe Ave., Suite 21, Vista, CA 92084, USA) or via websites www.cochranelibrary.com and www.update-software.com].
The Cochrane Oral Health Group (www.cochrane-oral.man.ac.uk) has developed and maintains a dedicated oral health register currently holding over 15,000 citations to reports of controlled clinical trials. This is a most valuable resource for those undertaking Cochrane Systematic Reviews and reviewers can arrange searches of the database through the Group’s Trials Search Co-ordinator, who also offers support to reviewers in developing sensitive search strategies.
Hand-searching Journals
It can be seen from the foregoing that where thoroughness of literature searching is paramount, the serious researcher will need to broaden the search to databases beyond MEDLINE. It should also be appreciated that, even then, electronic searching has its limitations and ‘hand-searching‘, which involves searching journals page by page, may need to be undertaken in order to trace as many relevant articles as possible.
The Cochrane Collaboration has recognized the importance of the need to identify randomized controlled trials for systematic reviews, and has set up a worldwide journals hand-searching programme to identify RCTs and CCTs. This is a highly organized programme, co-ordinated by the New England Cochrane Center, Providence Office, geared to avoiding duplication of effort by registering searches and making the results accessible to all through The Cochrane Library’s CENTRAL/CCTR database. The Master List of journals being searched can be accessed online via www.cochrane.us/cochranemainpage.asp. The Cochrane Oral Health Group contributes to this programme through its responsibility for the registration and overseeing of hand-searching of the oral health literature. The Group is always pleased to hear from anyone who can spare even a small amount of time to contribute to this important programme, which ultimately benefits researchers in oral health throughout the world. For an information sheet on the oral health journal hand-searching programme please contact Sylvia Bickley, Trials Search Co-ordinator, Cochrane Oral Health Group, MANDEC, University Dental Hospital of Manchester, Higher Cambridge Street, Manchester M15 6FH, UK (email: sylvia.r.bickley{at}man.ac.uk).
Limitations of electronic searching
To demonstrate the limitation of electronic searching and emphasize the importance of hand-searching journals, we undertook an exercise to examine and compare the results of searching for randomized controlled trials and controlled clinical trials on MEDLINE relying on searching the indexing field ‘Publication Type’ (PT). Four orthodontic journals for the publication period of 1991–2000 inclusive, which had also been hand-searched as part of the Cochrane Oral Health Group’s journal hand-search programme, made up the test-set for the exercise. (NB: the British Journal of Orthodontics changed its title to Journal of Orthodontics in year 2000 and these two titles were included in the MEDLINE search.)
SRB searched MEDLINE via OVID, using the following search strategy:
The combined results of the electronic MEDLINE search and manual hand-search identified a total of 304 citations as either RCTs or CCTs. The MEDLINE search identified 143 citations and hand-searching of the journals identified 266 citations. The MEDLINE search identified 105 of the 266 (39.5%) citations that had been found by hand-searching journals together with an additional 38 records. The 38 unmatched MEDLINE citations (12.5%) were examined and 32 (84.2%) of these were found not to be controlled clinical trials. These included 17 in vitro, five retrospective, and four matched control studies, cross-sectional, and non-clinical studies. The remaining six citations (15.8%) were found to be controlled clinical trials and these had been missed by hand-searchers (Table 1).
|
In presenting the results of this exercise, we would like to emphasize that there is no intention to denigrate the unquestionable value of MEDLINE. It is a resource that allows us to search and download swathes of valuable literature without moving from our desks. Nor would we wish this report to be seen as a criticism of the National Library of Medicine’s professional indexers because there are a number of things, outside their control, which may lead to incomplete or inappropriate indexing of publications, including unclear descriptors of study design and/or methodology. Several of these problems could be addressed by journals using structured abstracts.
This exercise serves to emphasize that when thoroughness of searching is paramount, for example for systematic reviewing, there are several points that you need to be aware of: These include the:
In conclusion, in this article we have endeavoured to give an insight into approaching searching the literature systematically, whilst cautioning against too much reliance on electronic searches. It has been shown that effective searching requires a knowledge and understanding of the many facets and anomalies of searching, concentrated effort, and practice. Where there is limited experience in searching and where accuracy of searching is paramount, we would advise seeking the guidance of a medical librarian or information specialist.
References
1 Dickersin K, Scherer R, Lefebvre C, Identifying relevant studies for systematic reviews. Br Med J 1994; 309: 1286–1291.
2 Harrison JE, Ashby D, Lennon MA. An analysis of papers published in the British and European Journals of Orthodontics. Br J Orthod 1996; 23: 203–209.[Abstract]
3 Haynes, RB, Mulrow, CD, Huith, EJ, Altman, DG, Gardner, MJ. More informative abstracts revisited. Ann Intern Med 1990; 113: 69–76.
4 Ad Hoc Working Group for Critical Appraisal of the Medical Literature. A proposal for more informative abstracts of clinical articles. Ann Intern Med 1987; 106: 598–604.
5 Begg C, Cho M, Eastwood S, Horton R, Moher D, Olkin I, Pitkin R, et al. Improving the quality of reporting of randomized controlled trials. The CONSORT statement. J Am Med Ass 1996; 276:637–639.[CrossRef][Medline]
6 Clarke M, Oxman AD, (Eds). Locating and selecting studies. Cochrane Reviewers’ Handbook 4.1.5, [updated April 2002]; Section 5. In: The Cochrane Library, Issue 4, 2002. Oxford Update Software. Updated quarterly.
7 Egger M, Davey-Smith G. Meta-analysis bias in location and selection of studies. Br Med J 1998; 316: 61–66.
8 Lefebvre C, Clarke MJ. Identifying randomized trials. In: Egger M, Davey-Smith G, Altman DA (Eds) Systematic Reviews in Health Care: meta-analysis in context, 2nd edn. London: BMJ Publishing Group; 2001.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
