OBJECTIVES: In multi-centre Time and Motion (T&M) studies, time endpoints can be highly variable due to differences in centre practices. Our aim was to assess the impact  of the type of  analysis employed on the results of a  T&M study. METHODS: Data from 13 centres were analyzed in relation to each of the following: drug preparation, distribution, and injection, using three methods.  Base case methodology included a random intercept generalized linear mixed effect model assuming gamma distribution with log link function to account for potential centre clustering effect and non-normality of the outcome measure. The two alternative methods were: standard linear regression (assuming time data are normally distributed) and gamma regression with log link function (assuming time data are positively skewed), both of which do not account for centre clustering effect. Sample means and variability as measured by 95% confidence interval (CI)) were also compared. RESULTS: For the base case, mean time was 0.53 min (95% CI: 0.33-0.85) for “preparation”, 0.30 min (95% CI: 0.22-0.40) for “distribution”, and 0.81 min (95% CI: 0.59-1.11) for “injection”. Mean time resulting from the standard linear regression was markedly higher for “preparation”: 0.66 (95% CI: 0.59-0.73), and similar for “distribution” and “injection”: 0.34 (95% CI: 0.30-0.37) and 0.84 minutes (95% CI: 0.79-0.88), respectively. Using the gamma regression yielded similar results to standard linear regression: 0.65 (95% CI: 0.59-0.71), 0.31 (95% CI: 0.29-0.34), and 0.83 minutes (95% CI: 0.79-0.88), respectively. The base case scenario detected a “centre-clustering” effect, hence producing substantially wider CIs compared to both alternative methods which ignore dependence in the data. CONCLUSIONS: Although mean task times remained relatively stable across the various methods, 95% CIs were substantially wider for random intercept model. If “centre-clustering” is detected, random effects regression models must be employed to produce valid confidence intervals around point estimates.