“Developing a geographic search filter to identify randomized controlled trials in Africa: finding the optimal balance between sensitivity and precision” This is the title of research done in South Africa and the UK.
They found that geographic location is not consistently reported in the bibliographic records of randomized controlled trials (RCTs). The location may not be evident from the title or abstract; and thesaurus headings may not be consistently applied. Sensitivity can be improved by expanding the text word search to include the author affiliation field. Since contact data is not standardized many variants of country names are needed; inclusion of capital and major cities might improve retrieval. Address data alone will miss those trials which are conducted by researchers from outside Africa.
The African search filter for Medline and Embase contained many text words as well as thesaurus/ MeSH headings. I counted 85 OR terms with 3 ‘NOT’ terms – e.g. NOT “guinea pig”. When the author affiliation field was included, the precision of the strategy was improved to 87% in Medline/ PubMed and to 89% in Embase.
This article and the search filter is available through HINARI.
Title: – [see above], Authors: Elizabeth Pienaar and four others.
Health Information and Libraries Journal. 2011; vol. 28 pp.210-215.
Jean Shaw, Partnerships in Health Information.
