Fig. 3

The probability of impairment given an observation, and true change given an initial score. a The cumulative sum of simulated scores (solid black curve), and a confidence interval (CI) of ± 0.95, as simulated for this parameter. The plot is divided into 3 regions based on the likelihood that a score is actually impaired (i.e. is really ≥ 1.64) given an observed value of 1.64. A score X < 0.69 (1.64–0.95) is statistically unimpaired, i.e. the score is too low for there to be a reasonable probability that the true performance is impaired. A score 0.69 (1.64–0.95) ≤ X < 2.59 (1.64 + 0.95) is possibly impaired. A score X ≥ 2.59 encompasses likely impairment. b) The concept of a) can be generalized to detect a change in a follow-up assessment score X2 given an initial assessment score X1, using significant change. The plot can be divided again into 3 regions. A score X2 < (X1 − 1.34), i.e. the second score is less than the first score minus the significant change threshold for this parameter, is statistically improved from the first assessment. A score (X1–1.34) ≤ X2 < (X1 + 1.34) indicates possibly different performance at followup. Finally, a score X2 > (X1 + 1.34) is statistically poorer