Enhancing prevalence estimation by addressing inconsistent reporting: insights from a cross-national, data-adjusted...

Evidence on adolescent alcohol use in low- and middle-income countries (LMICs) is scarce and, like research in high-income countries, generally overlooks inconsistent reporting. Based on a data-adjusted replication of a study in 57 LMICs, this work assessed the effect of inconsistent reporting on prevalence estimates.

To standardize the operationalization of inconsistent reporting, I examined the 37 datasets that contained the exact same set of items. Inconsistencies were identified based on logical contradictions. Prevalence estimates of past 30-day alcohol use and sources of alcohol acquisition were computed using unadjusted data (i.e., including all reporters) and adjusted data (i.e., excluding inconsistent reporters).

Inconsistent reporting was common (M = 9.6%; range: 3.7%-23.6%). Excluding inconsistent reporters had a modest average effect on past 30-day alcohol use (mean relative change: -8.9%), but substantial heterogeneity was observed. In eight datasets, prevalence decreased by at least 15%, reaching -29.4% in Ghana. Effects were, on average, larger in younger students and males. Excluding inconsistent reporters occasionally strengthened or reversed age- and sex-related patterns. Results pertaining to alcohol acquisition were similarly heterogeneous, with multiple changes <-30% or >30%, sporadic elimination of a specific acquisition source, and occasional reordering of source hierarchies.

As in high-income countries, inconsistent reporting of alcohol use is common among adolescents in LMICs and can substantially affect prevalence estimates. Effects are heterogeneous and hardly predictable, emphasizing the importance of incorporating consistency checks to improve the validity of research guiding public health decisions.