Testing the best performing methodological search filters to retrieve health economic evaluations in Embase: A filter validation study

Abstract

This master’s thesis consist of an introductory part describing the background for preforming the validation study, the methods used, and a discussion of the methods and results. The introductory part describes the filter validation study in more detail than possible in the article manuscript. Background: Health technology assessments (HTAs) are increasingly used by Norwegian health authorities as the evidence base when prioritizing which health care services to offer. HTAs typically consist of a systematic review of the effects and safety of two or more health care interventions, and an economic evaluation of the interventions, based on systematic literature searches in bibliographic databases. Objective: To identify the best performing of seven search filters to retrieve health economic evaluations used to inform health technology assessments (HTAs), by comparing the costeffectiveness analysis (CEA) filter to six published filters in Ovid Embase, and achieve a sensitivity of at least 0.90 with a precision of 0.10, and specificity of at least 0.95. Methods: In this filter validation study, the included filters’ performances were compared against a gold standard of economic evaluations published in 2008-2013 (n=2,248) from the National Health Service Economic Evaluation Database (NHS EED), and the corresponding records (n=2,198) in the current version of Ovid Embase Results: The CEA filter had a sensitivity of 0.899 and precision of 0.029. One filter had a sensitivity of 0.880 and a precision of 0.075, which was closest to the objective. The filter with lowest sensitivity (0.702) had a precision of 0.141. Conclusion: Developing search filters for identifying health economic evaluations, with a good balance between sensitivity and precision, is possible but challenging. Researchers should agree on acceptable levels of performance before concluding on which search filter to use.