Strelitz and colleagues provide a timely synthesis of randomised evidence on interventions to improve school attendance.
With heterogeneity of this magnitude, the CI around the pooled estimate quantifies uncertainty in the average effect but does not indicate the range of effects that might plausibly occur when the intervention is implemented in a new school or population. A 95% prediction interval, interpreted cautiously in light of the number and size of contributing trials, could be particularly informative because it incorporates between-study heterogeneity and addresses a question closer to that faced by decision-makers: what effect might reasonably be expected in another setting?
Interpretation also depends on how attendance is measured. An absolute difference…