Bradford’s law to predict the size of the literature in cochrane reviews

Article type
Authors
Kruesi L1, Nash-Stewart C2, Bernard-Del-Mar C3
1UQ Library, The University of Queensland, St Lucia, Queensland, Australia
2School of Medicine, The University of Queensland, Brisbane, Queensland, Australia
3Faculty of Health Sciences and Medicine, Bond University, Gold Coast, Queensland, Australia
Abstract
Background: Bradford’s Law is a mathematical formula that shows few journals account for the many articles on a subject. If journals are grouped into zones of roughly equal numbers of articles, Bradford’s Law predicts the number of journals in each zone. This enables a prediction of how much relevant information will be missed if a search is incomplete. Objectives: Following examination of Bradford’s Law upon Cochrane reviews on acute otitis media and pneumonia it was found the Law, which is based on diminishing return in searching exhaustively, did not conform. A previous analysis of all Randomised Controlled Trials (RCTs) in the MEDLINE database found that the distribution showed variations from the standard Bradford Law. Whilst there are numerous papers reporting the productivity of cited journals in health, the impact of Bradford’s Law upon a systematic reviewers need to search beyond core journals has not been studied. Methods: Titles, abstracts or keywords were searched in the Cochrane Library for reviews of treatments of acute otitis media and pneumonia. References to the conditions were sorted to generate a cumulative distribution for the subject. Results: For a cute otitis media, 71% of the core journal articles were included in the reviews, versus 35% of non-core journal articles (Figure 1). For pneumonia the numbers were 43.5% and 30%, a similar trend but with much reduced magnitude of difference. Less than half of all included articles came from core journals (otitis media 47% and pneumonia 38%). From the tail zone for pneumonia, the number of studies exceeded those from the core. Eighty-six per cent of all studies and 92% of included studies came from MEDLINE indexed journals. Conclusions: The study found that less than half of all included articles came from core journals. This confirms that searching non-core journals is necessary to capture the relevant literature and to avoid bias.