03/01/2024
By Hsien-Yuan Hsu
Time: Noon to 1 p.m.
Location: Coburn Hall 275
Speaker: Aaron Kaat (in-person presentation; Department of Medical Social Science, Northwestern University)
Title: Modeling decisions for patient reported outcomes: PROMIS Pain Interference and Upper Extremity after distal radius fractures.
Brief Bio:
Aaron Kaat Ph.D. is an Associate Professor in the Department of Medical Social Science, Northwestern University Feinberg School of Medicine and in the Roxelyn and Richard Pepper Department of Communication Sciences and Disorders, Northwestern University School of Communication. He obtained his PhD in Psychology—Intellectual and Developmental Disabilities from the Ohio State University, with a secondary concentration in Quantitative Psychology. He is a measurement scientist with expertise in measure development, adaptation, and validation in special populations using latent variable modeling, including item response theory. Substantive areas of interest include genetic conditions associated with neurodevelopment (GCAND), orthopaedic outcomes, and cognitive assessment. Recently he has advocated for the use of person-ability scores from IRT models for clinical trials and longitudinal study designs.
Talk Abstract:
Historically measure development and study methodology have been separate endeavors. Measurement scientists may create high-quality patient-, caregiver-, or clinician-reported outcome measures or direct assessments, using appropriate psychometric methods—but then they “toss” the assessments to biostatisticians to use in clinical trials and other research designs as if they were any other outcome. Modern measurement theory, however, provides additional methodological options to build upon traditional biostatistical methods and improve model precision.
Methodological choices, however, are just that—a choice. In an ongoing observational study, the AO BERT Consortium is monitoring recovery after distal radius fractures over the course of 12 months. Using interim data up to 6 months, this talk will consider some of the modeling choices for patient-reported outcomes, using the PROMIS Upper Extremity and Pain Interference domains as exemplars. Naïve mixed effects models will be compared to alternate models at increasing levels of complexity that leverage various aspects of modern measurement theory, from simple plausible value imputation to factor mixture modeling.