Skip to main content

Finding the largest pool of relevant citations in MEDLINE using the Pearl Harvesting Information Retrieval Theory

Date and Location




Sunday 22 September 2013 - 10:30 - 12:00


Presenting author and contact person

Presenting author

Robert Sandieson

Contact person

Robert Sandieson
Nazi Torabi
Abstract text
Background: Locating as many topic related research studies as possible is critical for meta-analyses, systematic reviews and scoping studies. However, performing a comprehensive literature search can be complicated using current protocols, and evidence indicates difficulties even for experienced researchers (e.g., Valentine et al., 2010). MEDLINE is important for health sciences so developing a comprehensive, verified system for information retrieval here is essential. Objectives: The Pearl Harvesting Information Retrieval Theory is premised on finding and using the largest possible set of topic related search terms (i.e., a synonym cluster) to locate a comprehensive set of relevant studies. The present investigation tested this theory in MEDLINE using search terms representing the topic of autism Methods: A synonym cluster of search terms representing autism was produced according to the theory. Testing was done comparing this list with terms from MeSH searches and text searches used in 18 meta-analyses on autism in the Cochrane library. Results: Twenty-three potential autism search terms were found, 9 of which were verified as essential. Two terms were not used in any of the Cochrane reviews, but they added only a small number of relevant citations. However, many terms used in the Cochrane reviews produced thousands of non-relevant citations, thereby making them unnecessary.Conclusions: Locating the maximum number of relevant studies is critical for acquiring knowledge leading to evidence informed decision-making. As found here, also knowing what not to search is equally important in terms of managing time and costs of searching. The Pearl Harvesting Theory utilizes a unique, yet simple approach relying on a verified list of topic related search terms. The effectiveness of this approach for the topic of autism was indicated here in MEDLINE. Once located, synonym clusters can be shared with all those interested in specific topics; saving time and effort.