Mizspellin and Medline

Joel G Ray, graduate student Marian J Vermeulen, graduate student

Literature searches, whether conducted for patient care or for construction of a systematic overview, depend on at least two factors to be comprehensive. These are, firstly, use of an inclusive set of search strategies,(1) and, secondly, correct entry and referencing of published material within the database.(2) There is limited evidence on the accuracy of information within electronic databases. We assessed the accuracy of entries in Medline by searching for misspelt textwords.

Methods and results
We conducted a literature search of 10 commonly used medical terms selected from the subject index of the ACP Journal Club (May/June 1996 issue).(3) We intentionally misspelt each term by altering one or two letters within the word (for example, myocardial infarction became myocardial infraction). We searched for the terms as textwords in Medline from 1966 to November 1996 but did not use medical subject headings (MeSH). We analysed the number of times a misspelt term occurred within an article's title, abstract, or both and the proportion of misspelt citations that might be missed if a search was conducted using only a textword search with the correctly spelt term.

Results
Table 1 summarises the results of our misspelt searches. A total of 200 citations were retrieved from the 10 selected search terms. Most misspelt textwords occurred within the abstract only (141/200; 71%). Surprisingly, 98 of the 200 articles (49%) with misspelt textwords might be missed if you conducted a Medline search using the correctly spelt word alone without the MeSH heading.

Table 1-Results of Medline search using 10 misspelt textwords
Misspelt term (correct spelling)	No of misspelt citations retrieved	No (%) with misspelling in title alone	No (%) with misspelling in abstract alone	No (%) of potentially missed citations using textword search alone
Angima (angina)	3	2 (67)	1 (33)	1 (33)
Antibotics (antibiotics)	15	6 (40)	9 (60)	8 (53)
Asprin (aspirin)	13	5 (38)	8 (62)	7 (54)
Canzer (cancer)	0	0	0	0
Dopler (Doppler)	8	3 (38)	5 (63)	5 (63)
Cholestrol (Cholesterol)	47	9 (19)	38 (81)	25 (53)
Hamorrhage (haemorrhage)	4	1 (25)	3 (75)	2 (50)
Myocardial infraction (myocardial infarction)	96	22 (23)	71 (74)*	41 (43)
Spetic (septic)	10	5 (50)	5 (50)	7 (70)
Thrombolism (thromboembolism)	4	3 (75)	1 (25)	2 (50)
Total	200	56 (28)	141 (71)	98 (49)
*Three articles contained misspelling within both the title and abstract.

Comment

Although we did not evaluate the impact of adding proper MeSH headings to the above searches (assignment of MeSH headings is automated and thus they are never misspelt), we feel that a substantial proportion of minor articles on these subjects would be missed in a detailed systematic literature search. Indeed, given the number of ways that words can be misspelt our findings clearly underestimate the magnitude of the problem. The topics that we chose for our misspelt searches are common to clinical practice and have been used within systematic literature reviews for various medical specialties, thereby making our findings relevant to such uses of Medline.

Researchers need to beware of misspelt textwords within Medline and to realise the importance of using both MeSH headings and textwords in any systematic literature search.(1) (4)

Special thanks to Dr R B Haynes for his help with the preparation of this manuscript.

References:
1 Haynes RB, Wilczynski N, McKibbon KA, Walker CJ, Sinclair JC. Developing optimal search strategies for detecting clinically sound studies in MEDLINE. Journal of the American Medical Informatics Association 1994;1:447-58.

2 Funk ME, Reid CA, McGoogan LS. Indexing consistency in MEDLINE. Bulletin of the Medical Library Association 1983;71:176-83.

3 Index. ACP Journal Club 1996;124:88-96.

4 Lowe HJ, Barnett GO. Understanding and using the medical subject headings (MeSH) vocabulary to perform literature searches. JAMA