Stefan Elbers

49 Evaluating IMPT programmes over time Effect of Time For all outcomes, the median pre-post effect sizes show a favorable trend, indicating a positive change in health from pre to post treatment (range = .38 - 1.94). The post to final- follow up effect sizes vary from -.49 to .15, indicating different trends. The median effect sizes from pre to final follow-up show an overall favorable change in health outcomes during the course of the study (range = .32 to .85). Table 3 shows the median effect size, range and the amount of statistical heterogeneity per contrast. The table also includes an overview of the number of cohorts that follow a particular pattern of effect over time, symbolized by different plotlines. For example, a statistically significant favorable pre-post effect that is followed by no effect from post to follow-up is represented by a positive slope, that flattens halfway. The general trend across all outcomes indicates a statistically significant favorable effect of time in 85% of the pre-post effect sizes. This is reflected in a positive median effect size (median SMC = .63, range = -.21 to 4.93). Fifteen percent of the effect sizes shows no pre-post effects and there were no statistically significant unfavorable effects. For all cohorts that included a measurement at pre, post and follow-up time points, a pattern with a significant pre-post effect that is maintained at follow-up was found in 72 (49%) of the cases. Twenty-three patterns (16%) indicated a favorable pre-post effect that further improved at follow-up. A triangular relapse pattern was found in 29 (20%) of the calculated effect sizes. Two patterns (1%) showed no effect from pre to post, but a positive effect from post to follow-up and 17 outcomes (12%) did not show any effect from pre to post or from post to follow-up. Three outcomes (2%) showed no pre-post effect, but an unfavorable effect from post to follow-up. Finally, the dataset did not contain any pattern with statistically significant unfavorable pre-post results. Not all studies included a post- treatment measure, which explains why the pre to final follow-up evaluations include more effect sizes than in the pattern analysis. The effect of time from pre to final follow-up was favorable for 168 (77%) of the effect sizes. Forty-seven effect sizes (21%) did not indicate an effect and four effect sizes (2%) showed an unfavorable effect over time. In the multimedia appendix, we provided time-series plots where we standardized each outcome measure on a scale from 0 to 100 (percentage of maximum score of the measurement instrument) and plotted the development of each cohort over time using all available data points.