A validation study revealed differences in design and performance of MEDLINE search filters for qualitative research

Abstract

ObjectivesSeveral search filters exist to identify qualitative research, but so far none of them has been validated with an independent set of relevant references irrespective of a medical topic. The objective of this study was to provide a comparative overview of validation results for various MEDLINE search filters. Study Design and SettingSearch filters were tested for plausibility. A relative recall approach was used to generate a gold standard based on an overview of systematic reviews of qualitative studies. For each review, the included qualitative studies were collected and checked for MEDLINE-indexing. The body of indexed articles yielded the gold standard. Validation tests were conducted to determine whether the references of the gold standard could be identified with the respective search filters. ResultsThirteen search filters were validated in MEDLINE. One search filter by Wong et al. (2004) was found to be the most sensitive (93.63%). While medical subject heading “qualitative research” achieved the best precision (2.15%), sensitivity was the lowest (22.56%). University of Texas provided the best balanced search filter with a sensitivity of 81.96% and a precision of 0.80%. ConclusionSearch filters to identify qualitative research in MEDLINE differ greatly in design and performance. The selection of the appropriate search filter depends on project-specific demands and resources.