MODELING LONGITUDINAL CHANGE IN PATIENT REPORTED OUTCOMES USING LATENT CURVE MODELS

• Published in: Value in health Vol. 20; no. 5; p. A329
• Main Authors: McGinley, JS, Wirth, R
• Format: Journal Article
• Language: English
• Published: Lawrenceville Elsevier Science Ltd 01.05.2017
• Subjects: Averages; Changes; Clinical outcomes; Fatigue; Functional impairment; Individual differences; Physical ability; Variability
• ISSN: 1098-3015; 1524-4733
• DOI: 10.1016/j.jval.2017.05.005

OBJECTIVES: Latent curve models (LCMs) offer a flexible method for analyzing change over time in patient reported outcomes (PROs). In this presentation, we demonstrate how LCMs can be used to understand within- and between-patient variability and to test a wide array of PRO-focused research hypotheses. METHODS: The current study used a simulated data example consistent with data obtained in PRO research to illustrate how a series of LCMs can be applied and interpreted in practice. The outcomes were based on PRO domains of fatigue and physical functioning. We begin by describing independent univariate LCMs for fatigue and physical functioning and then expand these models to the bivariate case (i.e., modeling the longitudinal relationship between fatigue and physical functioning simultaneously). RESULTS: The univariate fatigue and physical functioning LCMs showed that patients varied in their baseline levels and rates of change over time in both PRO domains. On average, fatigue increased through the course of the study whereas physical functioning decreased over time. The bivariate LCM, which jointly modeled fatigue and physical functioning, showed that patients with more fatigue at baseline tended to have decreased physical functioning at baseline and patients with greater increases in fatigue over time had faster declines in physical functioning. Further, there was evidence of a negative within-person autoregressive effect of fatigue on physical functioning. More precisely, when a patient was more fatigue than usual on a given day, they had worse physical functioning than usual on the following assessment day. CONCLUSIONS: Researchers are often interested in assessing hypotheses concerning longitudinal change in PROs over time. While many traditional statistical methods for repeated measures have limitations (e.g., highly restrictive, no insights on individual differences, disconnect from theoretical models), LCMs offer a dynamic framework for testing many hypotheses relating to between- and within-person change in one or more constructs over time.