Three ways to quantify uncertainty in individually applied “minimally important change” values

Abstract

For application in clinical research and practice, MIC values are always considered at the individual level, but determined in groups of patients. The interpretation comes with different forms of uncertainty. To appreciate the uncertainty, knowledge of the underlying distributions of change scores is indispensable.

Minimal Clinically Important Difference of the Disabilities of the Arm, Shoulder and Hand Outcome Measure (DASH) and Its Shortened Version (QuickDASH)

“…9,27 After triangulation of all our results, for the DASH, a change of 10.83 points was defined as the most acceptable MCID for moderate improvement, with good sensitivity (82%), specificity (74%), and classification accuracy (79%). 10 This MCID was inside the 95% CI for our ROC cutoff values, slightly superior to both our MDC 90 (10.8 points) and the MDC 95 (10.7 points) reported by Beaton et al, 3 and in line with the MCID (10 points; 95% CI: 5, 15) calculated in a sample of patients undergoing nonoperative treatment for forearm, wrist, and Various authors have suggested that it would be better to define a range of MCID values rather than a fixed value, 10,11 and there are reasons to be skeptical about claims of a single MCID value. 24 Overall, due to our methodological procedure, with its main focus on the ROCcurve approach and an MCID value higher than MDC 90 and not MDC 95 , our threshold of 10.83 points could represent the lower boundary for a small range of reasonable MCIDs, in which the upper boundary could be represented by the 15 points proposed by the DASH website, 20 according to Beaton et al, 2 who just considered the AUC in ROC curves for score changes of -1, -5, -7, -10, -15, and -20.…”

“…The major disadvantage of distribution-based approaches is that they do not provide a good indication of the importance of the observed change and thus cannot give the MCID. 10 Their main role lies in identifying the minimum detectable change (MDC), that is, the smallest change in score that can be detected beyond random error. 41 On the T T STUDY DESIGN: Prospective, single-group observational design.…”

“…40 For the distribution-based methods, we calculated the standard error of measurement (SEM), which links the reliability of a measurement tool to the standard deviation of the population. 10 The SEM was calculated from the analysis of variance used to produce the test-retest ICC. 38 Starting from the SEM, we calculated the MDC, which represents the smallest change in score likely to reflect true change, free of measurement error.…”

“…32,40,42 Furthermore, our calculation of the 95% CIs gave a useful indication of the sampling variation. 10 In addition, patients who reached (or surpassed) MCID thresholds were considered positive responders. Then, we calculated how many positive responders showed a final score of 16 points or less, which was lower than both the fifth percentile of our dysfunctional population at baseline (16.42 Splitting data according to the presence of a moderate versus large GRCS improvement (GRCS less than +2 versus GRCS of +2 or greater), the AUC of the DASH was 0.87 (95% CI: 0.84, 0.91), and that of the QuickDASH was 0.86 (95% CI: 0.82, 0.89) (FIGURE 3).…”

Improving the reporting and interpretation of clinical trial outcomes

“…At the very least, greater care needs to be taken when designing trials and when interpreting outcomes that use judgment thresholds, as inadequate distinctions between individuals and populations may be being made. 6 But we suggest that further work is needed to critically review the methods suggested for defining thresholds of population level importance.…”