Fig. 3: Comparison of observed values in RCTs (x-axis) and modelled estimates (y-axis), considering validation points only.

The black dashed line indicates the Y = X line, i.e. perfect agreement. A Observed and modelled prevalence, with associated uncertainty intervals. On the y-axis, point estimates represent the median model prediction across EHT trials (7 trials in East Africa, 3 in West Africa), transmission intensity calibration and model stochasticity (10 stochastic replicates). Error lines represent the minimum and maximum over these uncertainty ranges (see ‘Methods’ section for methodological details). On the x-axis, the point estimate correspond to the raw proportions reported in each trial (all computed over more than 800 individuals) and error lines represent 2 empirical standard deviations from the point estimate. B Observed and modelled effect sizes (relying on the median model prediction across EHT trials, transmission intensity calibration and model stochasticity). Effect size is defined as \(100\,{{{\rm{x}}}}\,\frac{{{{\mathrm{PR}}}}_{0}-{{{\mathrm{PR}}}}_{{{{\rm{i}}}}}}{{{{\mathrm{PR}}}}_{0}}\), where \({{{\mathrm{PR}}}}_{0}\) is the prevalence in the control arm and \({{{\mathrm{PR}}}}_{{{{\rm{i}}}}}\) is the prevalence in the intervention arm. C Observed and modelled relative change compared to baseline (relying on the median model prediction across EHT trials, transmission intensity calibration and model stochasticity). Relative change compared to baseline is defined as \(100\,{{{\rm{x}}}}\,\frac{{{{\mathrm{PR}}}}_{{{{\rm{t}}}}=0}-{{{\mathrm{PR}}}}_{{{{\rm{t}}}}}}{{{{\mathrm{PR}}}}_{{{{\rm{t}}}}=0}}\), where \({{{\mathrm{PR}}}}_{{{{\rm{t}}}}=0}\) is the prevalence at baseline (pre-intervention) and \({{{\mathrm{PR}}}}_{{{{\rm{t}}}}}\) is the prevalence post-intervention.