Article type
Year
Abstract
Background: Searching and identifying primary care relevant studies in the literature is challenging. Although specialist journals exist, relevant research is often published in general medical journals. To our knowledge, search strategies for finding studies relevant to primary care have yet to be developed and validated.
Objectives: To develop and validate a search filter for identifying studies of relevance to primary care in Medline.
Methods: We conducted a Medline search of articles published in five core medical journals at five yearly intervals (1998, 2003 and 2008). We excluded comments, editorials, letters and news publication types. All articles were randomly allocated to one of two sets: Reviewer A screened the titles and abstracts of the first set, while Reviewer B screened the second set. In cases of uncertainty the full-text was evaluated. The gold standard dataset was randomly divided into a development set and validation set. A textual analysis was conducted on the titles, abstracts and MeSH terms of the development set to extract frequently occurring words. Test search strategies were then developed and sensitivity/specificity data were derived using the validation set. The search strategies were further tested in Medline.
Results: A total of 12,045 articles were retrieved in the search. After exclusion of specified publication types the title/abstract of 9,028 records were screened and a further 495 records required full-text evaluation. The gold standard dataset included 371 articles (Figure 1). The sensitivity, specificity and precision of different search strategies will be reported.
Conclusions: Using textual analysis we have developed a range of Medline search strategies to identify articles relevant to primary care. However, inadequate description of the clinical research setting, inconsistent use of terminology to describe 'primary care’ and the overlap between primary care and the public health or outpatient setting make the construction of a highly efficient search filter highly challenging.
Objectives: To develop and validate a search filter for identifying studies of relevance to primary care in Medline.
Methods: We conducted a Medline search of articles published in five core medical journals at five yearly intervals (1998, 2003 and 2008). We excluded comments, editorials, letters and news publication types. All articles were randomly allocated to one of two sets: Reviewer A screened the titles and abstracts of the first set, while Reviewer B screened the second set. In cases of uncertainty the full-text was evaluated. The gold standard dataset was randomly divided into a development set and validation set. A textual analysis was conducted on the titles, abstracts and MeSH terms of the development set to extract frequently occurring words. Test search strategies were then developed and sensitivity/specificity data were derived using the validation set. The search strategies were further tested in Medline.
Results: A total of 12,045 articles were retrieved in the search. After exclusion of specified publication types the title/abstract of 9,028 records were screened and a further 495 records required full-text evaluation. The gold standard dataset included 371 articles (Figure 1). The sensitivity, specificity and precision of different search strategies will be reported.
Conclusions: Using textual analysis we have developed a range of Medline search strategies to identify articles relevant to primary care. However, inadequate description of the clinical research setting, inconsistent use of terminology to describe 'primary care’ and the overlap between primary care and the public health or outpatient setting make the construction of a highly efficient search filter highly challenging.
Images
