A systematic approach to choosing outcome measures for meta-analyses of healthcare interventions

Tags: Poster
Ipser J, Stein D

Background: One of the difficulties in synthesising data in meta-analyses of healthcare interventions is choosing between the multitude of outcome scales which are used in clinical trials. Traditional approaches to estimating treatment effects are not ideal, involving as they do either the ad hoc selection of a commonly employed outcome measure or the averaging of results from multiple scales of varying quality.

Objectives: To create a database of measures of treatment efficacy employed in randomized controlled trials of the anxiety disorders. This will aid in the objective identification of suitable measures for the extraction of summary statistics of treatment effect. Methods: A search of The Cochrane Library (2006, Issue 1) was conducted to identify reviews of the pharmacotherapy of anxiety disorders. The psychometric properties of the disorder-specific outcome measures employed by the short-term placebo-controlled trials in these reviews were retrieved from articles identified through searching PubMed and PsycINFO, and by consulting references lists. Data on the frequency with which the outcome measures have been used in the reviews, as well as their published construct validity and reliability scores will be made accessible through the internet (http://workhorse.pry.uct.ac.za/anxiety). Scales with no psychometric data will be ranked by a weighed count of the number of times agreement was reached with the validated measures.

Results: Reviews of pharmacotherapy for generalized anxiety disorder, post-traumatic stress disorder and social phobia were retrieved. A total of 61 short-term placebo-controlled trials yielded 29 outcome measures for inclusion in the pilot version of the database. A description of the database, as well as the correspondence between the validity of the scales and their frequency of use will be reported at the conference.

Conclusions: A database of systematically collected psychometric and utilisation data is one way of providing an objective basis for the selection of outcome measures in systematic reviews.