Reducing outcome measures in mental health: a systematic review of the methods.

Abstract

BACKGROUND: Traditionally, classical test theory (CTT) has been used for instrument development and various methods have since been proposed for reducing outcome measures to shorter versions. These reduction methods have not previously been compared in mental or physical health. AIM: To identify and compare the various methods used to develop brief versions of outcome measures from existing measures in mental health. METHOD: A systematic review of the literature in Embase, Medline, PsychInfo and from a grey literature was done. Search strategies were developed in each database to identify all relevant literature based on the inclusion criteria. Each paper identified was briefly described and then assessed using a bespoke assessment checklist developed by the authors. Methods for reducing outcome measures found across all studies were compared. RESULTS: Ten papers were identified. Five methods were used for scale reduction: Rasch analysis (RA), exploratory factor analysis (EFA), graded response models (GRMs), all-subset regression, and regression. RA was the most widely used process. CONCLUSION: The Rasch model (RM) is the only model where "'specific objectivity"' is a defining property of the model. This property is necessary for constructing scales in line with the fundamental principles of measurement.