Global prevalence and burden of depressive and anxiety disorders in 204 countries and territories in 2020 due to the COVID-19 pandemic

Global prevalence and burden of depressive and anxiety disorders in 204 countries and territories in 2020 due to the COVID-19 pandemic Damian Santomauro Ana Maria Mantilla Herrera Jamileh Shadid Peng Zheng Charlie Ashbaugh David M. Pigott Cristiana Abbafati Christopher Adolph Joanne O Amlag Aleksandr Y. Aravkin Bree Bang-Jensen Gregory J Bertolacci Sabina Bloom Rachel Castellano Emma Castro Suman Chakrabarti Jhilik Chattopadhyay Rebecca M Cogen James K. Collins Xiaochen Dai William James Dangel Carolyn Dapper Amanda Deen Megan Erickson Samuel B Ewald Abraham D. Flaxman Joseph Frostad Nancy Fullman J Giles Ababi Zergaw Giref Gaorui Guo Jiawei He Monika Helak Erin Hulland Bulat Idrisov Akiaja Lindstrom Emily Linebarger Paulo A. Lotufo Rafael Lozano Beatrice Magistro Déborah Carvalho Malta Johan Månsson Fatima Marinho Ali H. Mokdad Lorenzo Monasta Paulami Naik Shuhei Nomura James Kevin O'Halloran Samuel M Ostroff Maja Pasovic Louise Penberthy Robert C. Reiner Grace Reinke Antônio Luiz Pinho Ribeiro Aleksei Sholokhov Reed J D Sorensen Elena A. Varavikova Anh Vo Rebecca Walcott Stefanie Watson Charles Shey Wiysonge Bethany Zigler Simon I Hay Theo Vos Christopher Murray Harvey Whiteford Alize J Ferrari

Tag layer of the DHSS Meta-Tagging Project. The article text is reproduced from the staged copy named in sourceDesc and is not altered; the tags are held separately in standOff so that overlapping and nested spans are preserved.

Global prevalence and burden of depressive and anxiety disorders in 204 countries and territories in 2020 due to the COVID-19 pandemic 2021 The Lancet 10.1016/s0140-6736(21)02143-7 santomauro2021-global-prevalence-and-burden-of-de santomauro2021-global-prevalence-and-burden-of-de.txt, the project's staged copy cba88e8581cd124f15840c0f1a0ef4da0fe6a0e7 cc-by health-psychology

the paper's own words

the tag layer; a sentence written by the tagger about the paper, always anchored to a quotation but not itself a quotation

0 spans located by searching the stored text for the quote's exact words, allowing any whitespace between them the change in prevalence of major depressive disorder and anxiety disorders in 2020 across 204 countries and territories, estimated for locations with and without data Quotation withheld because this record is not cleared for redistribution. how much did the COVID-19 pandemic raise the global prevalence and burden of depression and anxiety, by age, sex and location Quotation withheld because this record is not cleared for redistribution. the first global estimate of the pandemic's effect on these two disorders, and a data standard the authors offer to the field for measuring the next population shock Quotation withheld because this record is not cleared for redistribution. existence Quotation withheld because this record is not cleared for redistribution. a population shock raising the prevalence of common mental disorders - the same pattern seen after the 2008 and 2009 financial crises, now measured globally Quotation withheld because this record is not cleared for redistribution. meta-analysis meta-analysis Quotation withheld because this record is not cleared for redistribution. why women were affected more: carer and household load from school closures, lower salaries and less secure employment, and a measured rise in domestic violence during lockdowns - social channels rather than biological ones unequal exposure to consequences Quotation withheld because this record is not cleared for redistribution. why the young were affected more: 1.6 billion learners out of school in over 190 countries, no physical spaces for peer interaction, and a higher chance of unemployment during and after an economic crisis disrupted development Quotation withheld because this record is not cleared for redistribution. the general category the paper places COVID-19 in - an unexpected event disrupting a population's environmental, health, economic or social circumstances, with the 2008-09 financial crises as its worked precedents population shock Quotation withheld because this record is not cleared for redistribution. distancing and travel restriction made medication and in-person care harder to get, services were interrupted or redirected, and people avoided seeking care for fear of infection - demand up and supply down together care interrupted at the moment of need Quotation withheld because this record is not cleared for redistribution. the two indicators are unlikely to have captured the full impact of the pandemic, and infection rates are not consistently measured or reported everywhere Quotation withheld because this record is not cleared for redistribution. mobile-phone mobility tracking is inaccurate where phone use is low, and people who cannot reduce their mobility because of their job may have their mobility drop overestimated - a bias with a direction, stated Quotation withheld because this record is not cleared for redistribution. VERY FEW QUALIFYING SURVEYS FROM LOW- AND MIDDLE-INCOME COUNTRIES, and large increases were nonetheless estimated for Latin America and the Caribbean, north Africa and the Middle East, and south Asia - super-regions from which NO study met the inclusion criteria at all Quotation withheld because this record is not cleared for redistribution. so the paper asks readers not to extrapolate rankings between countries, and its own leave-one-country-out cross-validation showed the rankings move with data availability Quotation withheld because this record is not cleared for redistribution. most surveys used symptom scales estimating PROBABLE cases; the assumption that screening thresholds kept their predictive validity through the pandemic could bias the estimates, and a high anxiety score in 2020 might be a natural reaction to a real threat Quotation withheld because this record is not cleared for redistribution. only three diagnostic surveys existed to test that assumption against; two showed increases and the third used very small samples and changed from face-to-face to telephone administration at the pandemic's onset Quotation withheld because this record is not cleared for redistribution. the severity distribution had to be assumed unchanged - existing cases may have worsened and new cases may be milder, and the burden estimate cannot tell Quotation withheld because this record is not cleared for redistribution. studies were excluded for convenience sampling, non-standard case definitions, and instruments with no comparable pre-pandemic estimate Quotation withheld because this record is not cleared for redistribution. other disorders may also have been affected - eating disorders are named - and were not estimated because the data have not been assessed Quotation withheld because this record is not cleared for redistribution. whether the increased prevalence came with increased suicide is left open: a 21-country time-series found no significant rise, and the authors say they have insufficient data to conclude either way Quotation withheld because this record is not cleared for redistribution. DALYs here are entirely years lived with disability - despite potential excess deaths, neither disorder is counted as an underlying cause of death, which is an ICD convention and not a finding about lethality Quotation withheld because this record is not cleared for redistribution. the geographical spread of the evidence: 22 studies from western Europe, 14 from high-income North America, 5 Australasia, 5 high-income Asia Pacific, 2 east Asia, 1 central Europe Quotation withheld because this record is not cleared for redistribution. an eligibility rule requiring a comparable PRE-PANDEMIC BASELINE, applied before any modelling: 5,683 sources found, 1,674 past title and abstract, 48 admitted eliminates: a during-pandemic prevalence being read as a CHANGE in prevalence. Without a baseline a survey measures a level, and the level was already high - this paper's own introduction notes no reduction since 1990 · outcome: held Quotation withheld because this record is not cleared for redistribution. GBD case definitions tied to DSM-IV-TR and ICD-10 rather than to whatever each survey called a case eliminates: the pooled estimate being an average over incompatible definitions of depression and anxiety · outcome: held Quotation withheld because this record is not cleared for redistribution. market-research quota sampling admitted but carried as a bias covariate rather than pooled with random samples eliminates: a non-probabilistic sampling frame inflating the estimate silently - AND THE COVARIATE EARNED ITS PLACE: cross-sectional quota studies significantly over-estimated the increase, which the model could then correct for · outcome: held Quotation withheld because this record is not cleared for redistribution. a separate bias covariate for instruments measuring depression and anxiety symptoms together, such as the K-6 eliminates: a combined-symptom instrument being counted as evidence about either disorder alone - and again it was measured, not assumed: those instruments overestimated the anxiety increase but not the depression increase · outcome: held Quotation withheld because this record is not cleared for redistribution. leave-one-country-out cross-validation of the final model eliminates: the estimates for locations with no data being untested - PARTIAL, and the authors report what it found against them: estimates stayed within their uncertainty bounds but the RANKINGS between locations moved with data availability · outcome: partial Quotation withheld because this record is not cleared for redistribution. backward elimination on the Akaike information criterion, with each unique sample given a random intercept and random effects on the impact indicators eliminates: covariates surviving in the model on the analyst's preference rather than on fit · outcome: held Quotation withheld because this record is not cleared for redistribution. PRISMA reporting, a protocol registered with PROSPERO before the review, and GATHER compliance with all analysis code published eliminates: the search and the inclusion rule being decided after the data were seen - the protocol is registered and the code is inspectable · outcome: held Quotation withheld because this record is not cleared for redistribution. no language restriction on the literature search eliminates: an English-only search compounding the geographical gap the authors already report - though it did not close it: no qualifying study came from a low- or middle-income country · outcome: partial Quotation withheld because this record is not cleared for redistribution. confirming that no 2020 epidemiological data entered the DisMod-MR model, so its 2020 output is a genuine counterfactual eliminates: the pandemic's effect being partly inside the baseline it is measured against - the pre-pandemic prevalence for 2020 is uncontaminated by 2020 · outcome: held Quotation withheld because this record is not cleared for redistribution. a statement that the funders had no role in design, collection, analysis or interpretation eliminates: a funder's interest shaping a global prevalence estimate that policy is set from · outcome: held Quotation withheld because this record is not cleared for redistribution. major depressive disorder major depressive disorder Quotation withheld because this record is not cleared for redistribution. anxiety disorders anxiety disorders Quotation withheld because this record is not cleared for redistribution. prevalence prevalence Quotation withheld because this record is not cleared for redistribution. pre-pandemic baseline pre-pandemic baseline Quotation withheld because this record is not cleared for redistribution. human mobility human mobility Quotation withheld because this record is not cleared for redistribution. daily SARS-CoV-2 infection rate daily SARS-CoV-2 infection rate Quotation withheld because this record is not cleared for redistribution. daily excess mortality rate daily excess mortality rate Quotation withheld because this record is not cleared for redistribution. COVID-19 pandemic COVID-19 pandemic Quotation withheld because this record is not cleared for redistribution. risk factors risk factors Quotation withheld because this record is not cleared for redistribution. proxies proxies Quotation withheld because this record is not cleared for redistribution. DisMod-MR 2.1 DisMod-MR 2.1 Quotation withheld because this record is not cleared for redistribution. DALYs DALYs Quotation withheld because this record is not cleared for redistribution. YLDs YLDs Quotation withheld because this record is not cleared for redistribution. market research quota sampling market research quota sampling Quotation withheld because this record is not cleared for redistribution. population shocks population shocks Quotation withheld because this record is not cleared for redistribution. screening measures screening measures Quotation withheld because this record is not cleared for redistribution. daily excess mortality rate → daily SARS-CoV-2 infection rate COLLINEAR AT r=0.8, WHICH IS WHY ONE OF THEM IS NOT IN THE MODEL. Excess mortality showed no independent association with the prevalence change once the other two indicators were controlled for, and the paper attributes that to collinearity rather than to absence of effect. None of the eight types says 'correlates with', and none of them can say 'dropped from the model because it duplicates another variable' - which is the actual relation and the reason the coefficient is missing Quotation withheld because this record is not cleared for redistribution. pre-pandemic baseline → prevalence A CRITERION THAT DECIDES WHAT COUNTS AS EVIDENCE. Requiring a comparable pre-pandemic measurement took 5,683 sources to 48 - it is the single most consequential step in the analysis and it is not a relation between two concepts in any of the eight senses. 'part-of' is false, 'causes' is false, and 'operationalised-as' inverts it: the baseline is not how prevalence is measured, it is what makes a measurement admissible Quotation withheld because this record is not cleared for redistribution. COVID-19 impact indicator - THE PAPER PRE-EMPTS ITS OWN MISREADING AND SAYS SO OUTRIGHT: human mobility and infection rate `should not be interpreted as risk factors`. They are PROXIES for the pandemic, which is the risk factor. A reader who takes the coefficient on human mobility as an effect of mobility has read the model backwards, and the authors put the warning in the discussion rather than leaving it to be inferred Quotation withheld because this record is not cleared for redistribution. prevalence, in this paper - A MODELLED QUANTITY, NOT A MEASURED ONE, and for most of the 204 locations there is no survey behind it. DisMod-MR estimates prevalence for locations missing raw data by cascading priors down a five-level geographical hierarchy, so a country's figure can be entirely inherited from its region Quotation withheld because this record is not cleared for redistribution. case, as counted here - A PROBABLE CASE ON A SCREENING SCALE, not a diagnosis. Only three diagnostic surveys existed to check against. The authors assume the predictive validity of screening thresholds did not change during the pandemic, and then give the counter-example themselves: a high anxiety-scale score in 2020 might be a natural reaction to a real threat rather than a disorder. SAME THRESHOLD PROBLEM AS lai2020, one level up Quotation withheld because this record is not cleared for redistribution. pre-pandemic baseline - THE ELIGIBILITY CRITERION, AND THE ONE THAT DID THE WORK. A prevalence survey run during the pandemic was not admissible without comparable data from before it, however large or well conducted. 5,683 sources became 48. Both COVID-19 mental health papers in this corpus - wang2020 and lai2020 - are cross-sectional with no pre-pandemic measurement and would fail it Quotation withheld because this record is not cleared for redistribution. not associated, said of daily excess mortality - A COLLINEARITY RESULT, NOT A NULL EFFECT. Excess mortality correlates at r=0.8 with both surviving indicators, so it drops out of the model after they are controlled for. The paper says the reason. Reading it as `deaths did not affect mental health` is the opposite of what the analysis supports Quotation withheld because this record is not cleared for redistribution. claim_type: existence a second reading reaching the same value. Recorded: existence — note: A BLIND second reading by the project on 2026-08-25: the record's own claim type/method/phenomenon were not looked at until every value here was written and substring-verified. Not applied - a person decides. Sampling frame: the 504 staged records carrying no alternatives block, so this never re-measures the 202. All 11 records with 130+ tags were taken as a census and 19 more spread one per discipline across 22 disciplines - the dense non-neuroscience cell is the one the 202 do not contain (inside them, 94-100% of the densest bands are neuroscience; among these 11, one is)., Quotation withheld because this record is not cleared for redistribution. method: systematic-review-and-meta-regression a second reading DIFFERING from the recorded value. Recorded: meta-analysis — note: A BLIND second reading by the project on 2026-08-25: the record's own claim type/method/phenomenon were not looked at until every value here was written and substring-verified. Not applied - a person decides. Sampling frame: the 504 staged records carrying no alternatives block, so this never re-measures the 202. All 11 records with 130+ tags were taken as a census and 19 more spread one per discipline across 22 disciplines - the dense non-neuroscience cell is the one the 202 do not contain (inside them, 94-100% of the densest bands are neuroscience; among these 11, one is)., Quotation withheld because this record is not cleared for redistribution. phenomenon: the global prevalence and burden of major depressive and anxiety disorders in 2020 attributable to the COVID-19 pandemic a second reading DIFFERING from the recorded value. Recorded: a population shock raising the prevalence of common mental disorders - the same pattern seen after the 2008 and 2009 fin — note: A BLIND second reading by the project on 2026-08-25: the record's own claim type/method/phenomenon were not looked at until every value here was written and substring-verified. Not applied - a person decides. Sampling frame: the 504 staged records carrying no alternatives block, so this never re-measures the 202. All 11 records with 130+ tags were taken as a census and 19 more spread one per discipline across 22 disciplines - the dense non-neuroscience cell is the one the 202 do not contain (inside them, 94-100% of the densest bands are neuroscience; among these 11, one is)., Quotation withheld because this record is not cleared for redistribution. critiques → the body of single-location snapshot surveys of depression and anxiety during COVID-19, and the qualitative syntheses and pooled estimates built on them note: the criticism is not of any one survey but of what the set of them cannot do: no study had synthesised a global prevalence and burden for 2020, and the pooled estimates on offer did not adjust for the baseline problem Quotation withheld because this record is not cleared for redistribution. distinguishes-from → surveys of prevalence during the pandemic that lack a comparable pre-pandemic baseline - which is the design of both wang2020 and lai2020 in this corpus note: THE PAPER DOES NOT NAME THOSE TWO. What it states is the criterion, and the criterion excludes their design - both are cross-sectional with no pre-pandemic measurement, and lai2020 says so itself. The relation is to the DESIGN CLASS, not to the papers Quotation withheld because this record is not cleared for redistribution. qualifies → Pirkis and colleagues' 21-country time-series finding no significant increase in suicide rates between April and July 2020 note: and the qualification runs BOTH ways - the authors say they have insufficient data to conclude anything, and offer that the window may have been too early rather than claiming the finding is wrong Quotation withheld because this record is not cleared for redistribution. builds-on → Frasquilho and colleagues' review of mental health outcomes after the 2009 economic crisis, and the Greek and Hong Kong prevalence series that followed the 2008-09 crashes Quotation withheld because this record is not cleared for redistribution. extends → GBD 2019's estimates of depressive and anxiety disorder burden, which this study adjusts rather than replaces Quotation withheld because this record is not cleared for redistribution. predicts → future mental health surveys responding to COVID-19 or another population shock, which the authors ask to align their sampling and instruments with pre-pandemic baselines note: an unusual relation to record and worth having: the paper is addressing studies that do not exist yet, and the data standard is offered as its second contribution Quotation withheld because this record is not cleared for redistribution. COVID-19 pandemic —causes→ major depressive disorder COVID-19 pandemic major depressive disorder direction on the measure: increases · strength of the relation: 0.9 · note: the paper is explicit that the PANDEMIC is the risk factor and the indicators are proxies for it, so this edge and not an indicator edge is the one it licenses Quotation withheld because this record is not cleared for redistribution. COVID-19 pandemic —causes→ anxiety disorders COVID-19 pandemic anxiety disorders direction on the measure: increases · strength of the relation: 0.9 Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but endpoint by anaphora: 'We also estimated' continues the previous sentence's estimate 'due to the COVID-19 pandemic'. Quotation withheld because this record is not cleared for redistribution. risk factors —distinct-from→ proxies risk factors proxies strength of the relation: 0.95 · note: the distinction the paper draws about its own model, in its own discussion, before anyone could get it wrong Quotation withheld because this record is not cleared for redistribution. prevalence —operationalised-as→ DisMod-MR 2.1 prevalence DisMod-MR 2.1 strength of the relation: 0.85 · note: and for locations with no raw data the operationalisation is the whole of it - the figure is a cascaded prior, not a measurement Audit 2026-09-28: DOUBTFUL - read against its span and kept, but prevalence is estimated by DisMod-MR 2.1; operationalised-as is loose. Quotation withheld because this record is not cleared for redistribution. YLDs —part-of→ DALYs YLDs DALYs strength of the relation: 1 · note: AUDIT REVERSED THIS. Years lived with disability are a COMPONENT of disability-adjusted life-years, not the container. I had written it backwards and then written a note explaining why backwards was acceptable. For these two disorders the component is the whole - DALYs are composed entirely of YLDs, because neither is counted as an underlying cause of death - and that is what makes the containment total here rather than partial Quotation withheld because this record is not cleared for redistribution. market research quota sampling —causes→ prevalence market research quota sampling prevalence direction on the measure: increases · strength of the relation: 0.85 · note: a bias with a measured direction, and the qualifier is load-bearing: cross-sectional quota studies overestimated, longitudinal ones did not Audit 2026-09-28: RESTORED AND DOUBTFUL - the full span was re-read; kept, but the full span says studies using market research quota sampling 'significantly over-estimated the increase in disorder prevalence' - the relation is stated, but 'causes prevalence' drops that it is an OVER-ESTIMATE of the increase; the withdrawal said the span only listed the sampling as a data source. The earlier withdrawal is kept in the record. Quotation withheld because this record is not cleared for redistribution. population shocks —causes→ major depressive disorder population shocks major depressive disorder direction on the measure: increases · strength of the relation: 0.85 Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but the span says these disorders are historically more SENSITIVE to population shocks; causes is loose. Quotation withheld because this record is not cleared for redistribution. an additional 53.2 million cases of major depressive disorder globally in 2020, an increase of 27.6% Quotation withheld because this record is not cleared for redistribution. an additional 76.2 million cases of anxiety disorders, an increase of 25.6% Quotation withheld because this record is not cleared for redistribution. 49.4 million DALYs from major depressive disorder and 44.5 million from anxiety disorders in 2020 Quotation withheld because this record is not cleared for redistribution. females were affected more than males and younger age groups more than older ones Quotation withheld because this record is not cleared for redistribution. the locations hit hardest by the pandemic had the greatest increases Quotation withheld because this record is not cleared for redistribution. 48 of 5,683 identified sources met the inclusion criteria - 46 for depression, 27 for anxiety Quotation withheld because this record is not cleared for redistribution. AND THE INCREASE WAS ESTIMATED FOR SUPER-REGIONS WITH NO QUALIFYING SURVEY AT ALL - Latin America and the Caribbean, north Africa and the Middle East, south Asia Quotation withheld because this record is not cleared for redistribution. no reduction in global prevalence or burden of either disorder had been detected since 1990, before the pandemic began Quotation withheld because this record is not cleared for redistribution. eligible study = reported prevalence during the pandemic AND had a comparable pre-pandemic baseline; without one it could not be included eligible studies Quotation withheld because this record is not cleared for redistribution. COVID-19 impact indicator = something that captures an effect of COVID-19, is consistently measured across locations, and is measured with sufficient granularity over time COVID-19 impact indicator Quotation withheld because this record is not cleared for redistribution. population shock = an unexpected or unpredictable event disrupting the environmental, health, economic or social circumstances of a population population shocks Quotation withheld because this record is not cleared for redistribution. case definitions taken from the GBD framework, adhering to DSM-IV-TR and ICD-10 case definitions Quotation withheld because this record is not cleared for redistribution. excess mortality correlates at r=0.8 with both surviving indicators and at 0.5 between them, which is why it drops from the model rather than because it has no effect Quotation withheld because this record is not cleared for redistribution. cross-sectional market-research quota studies significantly over-estimated the prevalence increase; longitudinal ones using the same sampling did not Quotation withheld because this record is not cleared for redistribution. instruments combining depression and anxiety symptoms overestimated the anxiety increase but not the depression increase Quotation withheld because this record is not cleared for redistribution. 95% uncertainty intervals are the 25th and 975th of 1,000 posterior samples Quotation withheld because this record is not cleared for redistribution. after the 2009 Greek crisis, point prevalence of major depressive episodes went from 3.3% in 2008 to 6.8% in 2009 to 8.2% in 2011 Quotation withheld because this record is not cleared for redistribution. in Japan suicide fell 14% over the first five pandemic months and then rose 16%, more among females and the young Quotation withheld because this record is not cleared for redistribution. outcome: change in logit prevalence of each disorder, pre-pandemic versus mid-pandemic change in prevalence Quotation withheld because this record is not cleared for redistribution. predictors: a single composite COVID-19 impact indicator, age, sex, and bias covariates for sampling strategy and instrument type bias covariates Quotation withheld because this record is not cleared for redistribution.

The full text of this article is not redistributable, so it is not included here. The evidence sentences are withheld with it; the tag layer and its grounding status remain. The article itself is at https://doi.org/10.1016/s0140-6736(21)02143-7.