A LATENT CHANGE SCORE FRAMEWORK FOR OPERATIONALLY DEFINING AND ESTIMATING MEANINGFUL SCORE DIFFERENCES

Researchers have long grappled with how to evaluate the meaningfulness of a treatment effect. Recently, understanding what constitutes a meaningful change has become a key research area in the clinical outcomes assessment (COA) field, where the patient perspective is strongly considered. A latent change score (LCS) modeling framework for operationally defining and estimating meaningful score differences (MSDs) in COA scores is detailed. It is shown that the most commonly used method for estimating meaningful change in COA scores—the transition rating average change (TRAC) method—can be recast as an LCS model and several alternative LCS models for estimating meaningful change are described. The MSD parameter recovery performance of two newly developed methods for operationally defining and estimating MSDs was evaluated in two simulation studies. Results indicate that MSD_(TRAC_θ ) and MSD_LRV consistently exhibit lower relative bias under the studied conditions than does the commonly used MSD_TRAC. When estimated with maximum likelihood in particular, MSD_(TRAC_θ ) exhibited trivial bias in the majority of conditions studied. Additionally, MSD_LRV exhibited trivial bias when transition rating item thresholds were truly invariant or when there was only one non-invariant threshold. The precision of MSD_(TRAC_θ ) and MSD_LRV was also more stable across the studied conditions than that of MSD_TRAC.