How to optimize measurement protocols: An example of assessing measurement reliability using generalizability theory

Abstract

Purpose: This article identifies how to assess multiple sources of measurement error and identify optimal measurement strategies for obtaining clinical out-comes. Method: Obtaining, interpreting, and using information gained from measurements is instrumental in physiotherapy. To be useful, measurements must have a sufficiently small measurement error. Traditional expressions of reliability include relative reliability in the form of an intra-class correlation coefficient and absolute reliability in the form of the standard error of measurement. Traditional metrics are limited to assessing one source of error; how-ever, real-world measurements consist of many sources of error. The measurement framework generalizability theory (GT) allows researchers to partition measurement errors into multiple sources. GT further allows them to calculate the relative and absolute reliability of any measurement strategy, thereby allowing them to identify the optimal strategy. We provide a brief comparison of classical test theory and GT, followed by an overview of the terminology and methodology used in GT, and then an example showing how GT can be used to minimize error associated with measuring knee extension power. Conclusion: The methodology described provides tools for researchers and clinicians that enable detailed interpretation and understanding of the error associated with their measurements.