Abstract
The fast-tracked publication of coronavirus disease 2019 (COVID-19)-related meta-analytic evidence has undeniably facilitated rapid public health policymaking; however, concerns are mounting that this publication policy may compromise research quality and scientific integrity. To investigate this, we conducted a meta-research study systematically evaluating risk of bias (ROB), transparency, and reproducibility in pandemic-era meta-analyses synthesizing COVID-19-derived mental health problem epidemics. From 98 identified studies-including data from 18.6 million individuals across 94 countries-we observed significant ROBs in publication, with one new meta-analysis published approximately every 5 days at peak output. Despite apparent sample diversity, nearly half of participants were from China, and only 8.9% originated from less economically developed countries. Of these meta-analyses, a substantial proportion (70.6%) showed discrepancies between Preferred Reporting Items for Systematic Reviews and Meta-Analyses-guided reporting and actual research conducts, while 57.1% exhibited high methodological ROBs due to insufficient data sources and lack of sensitivity analysis. Alarmingly, none achieved full computational reproducibility, and fewer than one-fifth were fully replicable. Furthermore, neither publication in high-impact journals, citation performance, nor fast-track publication mode correlated with lower ROBs that we identified above. To address these limitations, we re-estimated global COVID-19-derived mental health epidemics using their individual participant data after minimizing identified ROBs. Our recalibrated meta-analytic findings provide more reliable benchmarks for understanding the pandemic's mental health impact. This study demonstrated that rigorous methodology and scientific integrity must remain central priorities-even under urgent, crisis-driven conditions-establishing a foundation for transparent, reproducible, and unbiased global mental health surveillance during public health emergencies.