The impact of simplifying statistical modelling assumptions when invesigating physical functional disability in psoriatic arthritis patients Seminar

Abstract
Patient reported outcomes are important for assessing the quality of life of patients with Psoriatic Arthritis (PsA). The Health Assessment Questionnaire (HAQ) is used in rheumatology for evaluating and monitoring physical functional status, and is the measure of choice in cost-effectiveness studies in this area. In this talk, we discuss two approaches for modelling physical functional disability (via HAQ) over time. The approaches are based on fitting two-part mixed models to accommodate the preponderance of zeroes and on multi-state models. The choice depends on whether interest lies in the actual level of HAQ or transitions between various states of disability. The impact on inference of seemingly "innocuous" assumptions regarding independence of random effects in one case and piecewise constancy of a potentially rapidly fluctuating time-dependent variable in the other is investigated. We motivate this research using data collected on a prospective cohort of psoriatic arthritis patients from the Toronto PsA Clinic.