We used the estimates of cannabis use by age (in single years), sex, country and year described above, assuming that the prevalence of regular cannabis use was the same among individuals with and without schizophrenia. Estimates of the number of incident cases and the corresponding duration of schizophrenia by age, sex, country and year were based on the DisMod-MR model for schizophrenia. The value of one minus the ratio of the counterfactual to the observed duration is an estimate of the population attributable fraction of schizophrenia due to the effect of regular cannabis use on age of onset.
