Retrieving research studies: a comparison of bibliographic and full-text versions of the New England Journal of Medicine.

It has been established that subject searches of medical full-text databases obtain higher recall than subject searches in a bibliographic database. In this study we attempted to determine if the same rule might apply when searching for a non-subject parameter such as study design. A simultaneous search of bibliographic and full-text records from the New England Journal of Medicine provided data on the number of items retrieved by each kind of search. Filtering strategies were created for 5 different study types: randomized controlled trials, other clinical trials and prospective studies, cohort studies, longitudinal and follow-up studies, and multicenter studies. The point of the study was to compare the numbers of items retrieved from the bibliographic database, MEDLINE, and those retrieved from the full-text version of NEJM, and to examine the unique access points available in each file. For all the study types the full-text file retrieved a larger number of records than MEDLINE, most of which were retrieved because of methodology terms found in the text but not in the title or abstract. In MEDLINE, descriptors and publication types, two value-added fields supplied by indexers, retrieved 11-89% more than title and abstract alone.