In the current study, we aimed to determine if the variability of the incidence rates of psychosis found was associated with methodo-logical aspect of the studies such as: study type, case ascertainment, diagnosis, diagnostic instrument to confirm cases, diagnostic classification system, duration of the case ascertainment, lower and upper age included in the studies and quality of the study.A meta-analysis of these factors was performed for all incidence rates as a whole. For the analysis of subgroups we used a mixed effects model. For each subgroup, heterogeneity was calculated using the Q sta-
tistic and its corresponding p-value. In addition, a multivariate regression was performed. We analysed 92 estimates of incidence of psychosis corresponding to 30 citations from our previous meta-analysis. The pooled incidence rates of psychosis (mean effect: 18.30 per 100,000 [14.19–22.42]) revealed that heterogeneity between studies was very high. The final model explains 63.53% of the variability in the incidence rates reported by the different studies. Based on the subgroup analysis, we found statistically significant differences in incidence rates associated with diagnosis, diagnostic instruments used, duration of case ascertainment period and ages of both lower and upper cutoff. Finally, study type and quality of the studies had a weak association.Regarding meta-regression, studies that collected patients in the first contact with mental health units presented higher incidences rated, but with a weak association, more cases are detected in the studies in which the cases are included at the first contact, since there is not always an admission in the first episode. Studies that recruited patients in mental health units and primary care services, or in both and in social services had significantly higher incidence rates than only in primary care.
