Which depression gets counted? Measurement approach, region, and the making of prevalence estimates: a systematic review and meta-analysis

Aims: Prevalence estimates of depression vary widely across studies. We examined how much of this variation is associated with the measurement approach used (screening tools, rating scales, or diagnostic interviews) and with WHO region, treating measurement instruments as classifying technologies that partly constitute the populations they describe. Methods: We systematically searched EBSCOhost, Web of Science, and Scopus for studies published from 2013 onward that reported point prevalence of depression among adults in the general population. Twenty-five studies contributing 35 prevalence estimates were included. We fitted multilevel random-effects meta-analytic models with measurement approach and WHO region as moderators. Results: The pooled point prevalence was 5.8% (95% CI 4.5%–7.5%), with considerable heterogeneity (I² = 99.7%; 95% prediction interval 1.3%–22.7%). Prevalence differed by measurement approach: 6.9% for screening tools, 4.4% for rating scales, and 2.5% for diagnostic interviews, and measurement approach accounted for 16.5% of the estimated heterogeneity. WHO region was not significantly associated with prevalence and explained no additional variation once measurement approach was included. A complete region-by-approach interaction could not be estimated because several combinations contained no studies. Conclusions: Depression prevalence is partly constituted by the measurement apparatus rather than simply detected by it. Prevalence should be understood in relation to the instruments and institutional contexts that produce them, especially when they inform mental health policy in times of uncertainty.