Online Program

Return to main conference page

Thursday, January 11
Thu, Jan 11, 9:00 AM - 10:45 AM
Crystal Ballroom A
Novel Statistical Methods for Patient-Reported Outcome Measures

Using Structural Equation Modeling to Establish Measurement Invariance for Patient Reported Outcomes (304012)

*Bellinda L. King-Kallimanis, US Food and Drug Administration 

Assessing how patient reported outcomes (PROs) differ between groups or change over time is of particular interest to a large number of stakeholders. Before such substantive conclusions can be drawn, an important step is to establish that measurement invariance has not been violated. Measurement invariance is said to have been violated when the relationship between the measurement instrument and other variables (e.g. time, group membership) is not fully explained by their relationship with the attribute of interest (e.g. health related quality-of-life). As the attribute of interest is a latent variable, the use of structural equation modeling (SEM) is one analytic method for establishing measurement invariance. This is achieved by testing a set of increasingly constrained models and assessing whether there is significant deterioration of model fit. In this presentation, I will present a three step procedure for using SEM. Examples will include across country assessment for 12 European countries and across time evaluation in cancer patients who completed PROs at baseline and subsequent follow-up after the initiation of treatment.