Exploring the variation in muscle response testing accuracy through repeatability and reproducibility

Research Objectives To explore the variation in mean muscle response testing (MRT) accuracy and whether this variation can be attributable to participant characteristics. Methods A prospective study of diagnostic test accuracy was carried out in a round-robin format. Sixteen practitioners tested each of 7 test patients (TPs) using 20 MRTs broken into 2 blocks of 10 which alternated with 2 blocks of 10 intuitive guessing. Mean MRT accuracies (as overall percent correct) were calculated for each unique pair. Reproducibility and repeatability were assessed using analyses of variance (ANOVA) and scatter and Bland-Altman plots. Results The mean MRT accuracy (95% CI) was 0.616 (0.578–0.654), which was significantly different from both the mean intuitive guessing accuracy, 0.507 (95% CI 0.484–0.530; p<0.01) and chance (p<0.01). Visual inspection of scatterplots of mean MRT accuracies by practitioner and by TP suggest large variances among both subsets, and regression analysis revealed that MRT accuracy could not be predicted by TP (r = ‒0.14; p = 0.19), nor by Practitioner (r = 0.01; p = 0.90). A significant effect imposed by both practitioners and TPs individually and together was found at the p<0.05 level; however, together they account for only 57.0% of the variance, with 43.0% of the variance unexplained by this model. From a statistical perspective, Bland-Altman Plots of mean MRT accuracy by practitioner do show adequate repeatability since all scores fell within 2 SDs of the mean; however, the wide range of scores also suggests insufficient repeatability from a clinical perspective. Finally, ANOVA demonstrated that an insignificant amount of variance could be explained by block [F(1,21) = 0.02, p=0.90].