Figure 4

Years of education and age conjointly moderate the effect of lesion size on stroke-induced impairment (NIHSS) and disability (mRS). Three-way interactions of age, years of education, and lesion size were consistently observed for prediction of NIHSS in the acute (A) and the chronic stroke phase (B) as well as for prediction of mRS in the acute stroke phase (C) and a trend thereof in the chronic stroke phase (D). The left panels illustrate the model fit by plotting observed against predicted class assignments. The right panels illustrate the predicted odds of the probabilities for class 1 (favorable outcome) against class 2 (poor outcome) as a function of lesion size, age, and years of education. An odds of 1 would hence denote equal probabilities for better and poorer outcome, whereas an odds of 2 would reflect that the probability of a favorable outcome (class 1) is twice the size of the probability of a poor outcome (class 2). In turn, an odds of 0.5 would reflect that the probability of a favorable outcome (class 1) is half the size of the probability of a poor outcome (class 2). Slopes are plotted at centerings for age, education, and lesion size at their respective 25th, 50th, and 75th percentiles. Please note that predictor values corresponding to these percentiles vary slightly across panels due to slightly different sample sizes for clinical outcome scores (see “Results” section for details). The figure was created using Matlab software, version 2018a https://www.mathworks.com/products/matlab.html and Inkscape, version 1.0 https://inkscape.org/.