Measuring Change

Abstract

We present mixed-effects regression analysis as a way to examine change over time in a descriptive and illustrative way without equations. Beginning with concepts from ordinary regression, we discuss intention to treat and review some antiquated methods as principles for examining repeated measures. We introduce mixed-effects regression as a method for analyzing repeated measures data from randomized clinical trials and data from longitudinal observational studies. Using several examples, we illustrate mixed-effects regression as a flexible technique for incorporating time-invariant and time-varying measures in analyses using centered variables for ease of the interpretation.We use publicly available epidemiological data to illustrate change in mean arterial pressure (MAP) in adults. Among examples provided, we illustrate ways to interpret change in multivariable analyses of MAP, differences in changes in MAP, differences in rates of change in MAP, and change in a dichotomous outcome. These examples are intended to guide analysts in understanding and interpreting repeated measures analyses.