Mental disorder symptoms during the COVID-19 pandemic in Latin America - a systematic review and meta-analysis.

AIMS: There is a lack of evidence related to the prevalence of mental health symptoms as well as their heterogeneities during the coronavirus disease 2019 (COVID-19) pandemic in Latin America, a large area spanning the equator. The current study aims to provide meta-analytical evidence on mental health symptoms during COVID-19 among frontline healthcare workers, general healthcare workers, the general population and university students in Latin America.
METHODS: Bibliographical databases, such as PubMed, Embase, Web of Science, PsycINFO and medRxiv, were systematically searched to identify pertinent studies up to August 13, 2021. Two coders performed the screening using predefined eligibility criteria. Studies were assigned quality scores using the Mixed Methods Appraisal Tool. The double data extraction method was used to minimise data entry errors.
RESULTS: A total of 62 studies with 196 950 participants in Latin America were identified. The pooled prevalence of anxiety, depression, distress and insomnia was 35%, 35%, 32% and 35%, respectively. There was a higher prevalence of mental health symptoms in South America compared to Central America (36% v. 28%, p < 0.001), in countries speaking Portuguese (40%) v. Spanish (30%). The pooled prevalence of mental health symptoms in the general population, general healthcare workers, frontline healthcare workers and students in Latin America was 37%, 34%, 33% and 45%, respectively.
CONCLUSIONS: The high yet heterogenous level of prevalence of mental health symptoms emphasises the need for appropriate identification of psychological interventions in Latin America.

Introduction

Latin America, consisting of 33 countries or territories, has had the second-highest amount of coronavirus disease 2019 (COVID-19) cases and deaths per capita (Burki, 2020; World Health Organization, 2020; Ríos, 2021). Latin America is vulnerable to the destructive outbreak for several reasons including long-standing structural and socioeconomic inequities (Carvalho et al., 2015; Dávila-Cervantes and Agudelo-Botero, 2019; Burki, 2020) over 20% of the population in poverty, lack of healthcare access, underfunded healthcare systems, poor governance or political dynamics, a high burden of chronic and metabolic health conditions and lack of preparedness to fight the pandemic (Malta et al., 2020). Reportedly, there is a considerable increase in psychological morbidities among several demographic groups, including healthcare workers, the general population and students (Campos et al., 2021). Latin America is a vast area where tropical regions span across almost all countries and regional disparities on mental health have been reported (Malta et al., 2020), but we still lack evidence on the prevalence of mental health symptoms as well as their heterogeneities during the COVID-19 pandemic.

Recently, meta-analyses have provided early global evidence on the prevalence of mental health symptoms across groups, including healthcare workers, the general population and students (Batra et al., 2020; Luo et al., 2020; Pappa et al., 2020). These reports included very few studies based on Latin American samples. With emerging studies on mental health in Latin America, it is critical to synthesise meta-analytical evidence to provide integrated data on mental health among key demographic groups in Latin America during the COVID-19 pandemic. Therefore, this meta-analysis aims to investigate the pooled prevalence of mental health symptoms during the COVID-19 pandemic among frontline healthcare workers, general healthcare workers, the general population and university students in Latin America. We first perform subgroup analysis for Latin America based on South America (a majority but not all countries are in tropical regions) and Central America (all countries are entirely tropical).

Methods

Protocol registration

We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement 2020 (Liberati et al., 2009) to guide our meta-analysis and registered it with the International Prospective Register of Systematic Reviews (PROSPERO: CRD42020224458).

Eligibility criteria

The search targeted observational studies that assessed the prevalence of psycho-morbid symptoms of anxiety, depression, distress and insomnia among frontline healthcare workers, general healthcare workers, the general population aged 18 years or above and university students in Latin America. A priori inclusion criteria were established to identify eligible studies that used established psychometric survey tools, used the English language, and were available as full-texts. Studies that targeted other populations, including children, adolescents and certain subgroups (e.g., pregnant women), were excluded. Other study designs, such as reviews and meta-analyses, qualitative, mixed methods, case reports, studies published only as abstracts, biochemical and experimental studies, or articles lacking the use of robust psychometric instruments or with an ambiguous methodology to identify prevalence were also excluded. Studies based on non-Latin American countries were excluded. Studies with unclear methodology and results were reviewed carefully, and a researcher (WX) attempted to contact authors to seek the information in several instances: (1) if the study reported estimates for both targeted and excluded populations, posing challenges for us to delineate the prevalence rate for the population of interest to our study; (2) if the study did not report the prevalence as proportions; (3) if the study did not specify cut-off scores for levels of severity; or (4) if the study was missing crucial information such as response rate, duration of data collection and gender distribution.

Data sources and search strategy

This meta-analysis is part of a large project on meta-analysis of mental health symptoms during COVID-19. Bibliographic databases, such as PubMed, Embase, PsycINFO and Web of Science, were searched on 13 August 2021. medRxiv was also searched for preprints. Search algorithms specific to each database were used to yield a comprehensive pool of literature. A detailed search strategy appears in online Supplementary Table S1.

Phases of screening

A researcher (JC) exported the search results from various databases into Endnote to remove duplicates and then imported them into Rayyan for subsequent screening. Two reviewers (AD & BZC) independently screened the titles and abstracts of all papers in accordance with the prespecified eligibility criteria. The eligible abstracts proceeded to full-text screening for possible inclusion. Any conflicts between reviewers were resolved by a third reviewer (RKD).

Data extraction

A codebook was developed for standardisation and consistency. The final studies included from the screening process were sent to three groups (two reviewers in each group, WX & AY, BZC & AD, RZC & SM) for thorough investigation and extraction of relevant data elements into a coding book. Standardised codes were used to record pertinent variables, including author, title, country, duration of data collection, study design, population, sample size, response rate, female proportion, mean age, psychological outcome, severity level of outcome, type of survey instruments with cut-off scores and prevalence of psycho-morbid events. The severity of psychological outcomes of interest was coded as above mild, moderate above and severe levels (if available). The studies that reported only mild, moderate, and severe prevalence data were recoded into mild above, moderate above and severe prevalence for consistency purposes. The severity levels in studies that only reported the overall prevalence were determined based on cut-off scores (if available). After finishing independent coding, all the extracted data elements were subject to a second round of review by the coders to identify any discrepancies. In case of disagreements, a third reviewer (WX or TL) helped to achieve consensus through re-verification and discussion.

Risk of bias (RoB) assessment

The Mixed Methods Appraisal Tool (MMAT) with seven questions was used as a quality assessment tool (Hong et al., 2018; Pablo et al., 2020; Usher et al., 2020). Two reviewers independently assessed and assigned scores to the studies using the tool dictionary and guidelines. Disagreements were resolved through discussion with the lead reviewer (RKD). The quality scores ranged from 0 to 7 (highest quality). Studies were categorised as high, medium, or low quality if they attained the score of 6, 5 to 6, or <5, respectively.

Effect measure and data analysis

Using Version 16.1 of Stata (metaprop package), a random-effects model was used to compute the pooled estimates of outcome prevalence between populations by assuming that these studies are randomly selected from their targeted populations in Latin America to generalise our results to comparable studies in the region (Borenstein et al., 2021). We computed prediction intervals to show the range of the effect sizes across studies (Borenstein et al., 2017). The I2 statistic was used to calculate variance difference from effect sizes in order to quantify heterogeneity rather than sampling error (Higgins et al., 2019). Visual inspection of the Doi plot and the Luis Furuya–Kanamori (Furuya-Kanamori et al., 2018) index were used to assess publication bias (Kounou et al., 2020; Yitayih et al., 2020). The event ratio was used as the primary effect measure for the pooled estimates.

Results

Screening of studies

A total of 446 records were identified through searching bibliographical databases and other sources (Fig. 1). After removing 114 duplicates, a total of 332 records advanced to the screening phase. After excluding 225 records that did not pass the title and abstract screening, 107 records were identified as eligible for full-text screening. Among them, 40 papers were excluded for different reasons. For example, we excluded seven papers in Spanish and one paper in Portuguese. We sent emails to the authors of eight studies, to request missing critical information; three studies provided new prevalence data and were included in the final pool. Therefore, 62 studies, focused on populations in Latin America, were used in the final data extraction and analysis (online Supplementary Table S2).

Fig. 1.

PRISMA flow diagram.

Study characteristics

A total of 65 unique samples from 62 studies involving 196 950 participants from Latin America were included in this meta-analysis (Badellino et al., 2020, 2022; Campos et al., 2020, 2021, 2021; Chen et al., 2021, 2021; Civantos et al., 2020; Cortés-Álvarez et al., 2020; Dal'Bosco et al., 2020; De Boni et al., 2020; Fernández et al., 2020; Giardino et al., 2020; Guiroy et al., 2020; Malgor et al., 2020; Martinez et al., 2020; Medeiros et al., 2020; Mier-Bolio et al., 2020; Monterrosa-Castro et al., 2020; Mora-Magaña et al., 2020; Passos et al., 2020; Paz et al., 2020; Samaniego et al., 2020; Schuch et al., 2020; Yáñez et al., 2020; Antiporta et al., 2021; Boluarte-Carbajal et al., 2021; Brito-Marques et al., 2021; Cayo-Rojas et al., 2021; Cénat et al., 2021; Dantas et al., 2021; de Oliveira Andrade et al., 2021; Espinosa-Guerra et al., 2021; Esteves et al., 2021; Fernandez et al., 2021; Ferreira et al., 2021; Feter et al., 2021; Flores-Torres et al., 2021; García-Espinosa et al., 2021; Goularte et al., 2021; Landaeta-Díaz et al., 2021; Loret de Mola et al., 2021; Mautong et al., 2021; Mendonca et al., 2021; Mota et al., 2021; Nayak et al., 2021; Puccinelli et al., 2021; Ribeiro et al., 2021; Schmitt Jr et al., 2021; Scotta et al., 2021; Serafim et al., 2021; Souza et al., 2021; Torrente et al., 2021, 2021; Villela et al., 2021; Vitorino et al., 2021; Werneck et al., 2021; Zhang et al., 2021, 2021; da Silva Júnior et al., 2021; Robles et al., 2021) (Table 1 and online Supplementary Table S2). Some studies include multiple independent samples. For example, one study examined the prevalence of both general healthcare workers and frontline healthcare workers. Among them, 35 samples (53.85%) were of general populations (Passos et al., 2020; Antiporta et al., 2021; Boluarte-Carbajal et al., 2021; de Oliveira Andrade et al., 2021; Espinosa-Guerra et al., 2021; Ferreira et al., 2021; Landaeta-Díaz et al., 2021; Mautong et al., 2021; Ribeiro et al., 2021; Schmitt Jr et al., 2021; Souza et al., 2021; Torrente et al., 2021; Vitorino et al., 2021; Badellino et al., 2022), two samples (3.08%) were of frontline healthcare workers (Dal'Bosco et al., 2020; Robles et al., 2021), 19 samples (29.22%) were from general healthcare workers (Chen et al., 2020; Civantos et al., 2020; Giardino et al., 2020; Guiroy et al., 2020; Malgor et al., 2020; Monterrosa-Castro et al., 2020; Mora-Magaña et al., 2020; Samaniego et al., 2020; Yáñez et al., 2020; Zhang et al., 2021; Villela et al., 2021; Nayak et al., 2021; Dantas et al., 2021; Mota et al., 2021; Brito-Marques et al., 2021; Campos et al., 2021; Mier-Bolio et al., 2020; Robles et al., 2021) and nine samples (13.85%) were based on university students (Medeiros et al., 2020; Campos et al., 2021; Cayo-Rojas et al., 2021; Esteves et al., 2021; Fernandez et al., 2021; García-Espinosa et al., 2021; Mendonca et al., 2021; Scotta et al., 2021; da Silva Júnior et al., 2021). Of the 62 studies, 32 were from Brazil (49.22%) (Table 1). Except for three (4.84%) longitudinal cohort studies (Feter et al., 2021; Flores-Torres et al., 2021; Loret de Mola et al., 2021), the majority of the studies were cross-sectional (95.16%). The sample size varied from 62 to 196 950 participants. The participation rates varied from 11.4% to 100.0% with a median value of 72.25%. The female proportions among the 65 samples varied from 3.4% to 89.8% with a median of 72.25%.

Table 1.

Characteristics of the studies on mental health in Latin America during the COVID-19 pandemic

Characteristics	Total number of studies/samplesa	Percent	Level of analysis
Overall	62/65	100
Outcomeb		-	Prevalence
Anxiety	95	42.79
Depression	87	39.19
Distress	21	9.46
Insomnia	19	8.56
Severityb			Prevalence
Above mild	77	34.68
Above moderate	87	39.19
Above severe	52	23.42
Overall	6	2.71
Population			Sample
Frontline HCWs	2	3.08
General HCWs	19	29.22
General population	35	53.85
Students	9	13.85
Sampling country			Sample
Argentina	8	12.31
Bolivia	1	1.54
Brazil	32	49.22
Chile	1	1.54
Colombia	1	1.54
Ecuador	3	4.62
Haiti	1	1.54
Mexico	8	12.30
Panama	1	1.54
Paraguay	1	1.54
Peru	6	9.23
Trinidad and Tobago	1	1.54
Mixed	1	1.54
Quality			Study
High	30	48.39
Medium	32	51.61
Design			Study
Cohort	3	4.84
Cross-sectional	59	95.16
Publication			Study
Preprint	4	6.45
Published	58	93.55
Overall	Mean (median)	Range
Number of participants	3030 (671)	31–57 446	Sample
Female proportion	67.9% (72.25%)	3.4–89.8%	Sample
Response rate	66.0% (73.7%)	11.4–100%	Sample

Some studies include multiple independent samples. For example, one study47examined the prevalence of both general healthcare workers and frontline healthcare workers.

One independent sample in a study may report anxiety, depression and insomnia at the levels of mild above, moderate above and severe. Therefore, the total number of prevalence is larger than the total number of independent samples.

Estimates of pooled prevalence of psychological morbidity symptoms

In Latin America, 56 samples from 54 studies reported the prevalence of anxiety symptoms among 128 060 participants (Badellino et al., 2020; Campos et al., 2020, 2021, 2021; Chen et al., 2020; Civantos et al., 2020; Cortés-Álvarez et al., 2020; Dal'Bosco et al., 2020; De Boni et al., 2020; Fernández et al., 2020; Malgor et al., 2020; Martinez et al., 2020; Medeiros et al., 2020; Mier-Bolio et al., 2020; Monterrosa-Castro et al., 2020; Mora-Magaña et al., 2020; Passos et al., 2020; Paz et al., 2020; Samaniego et al., 2020; Schuch et al., 2020; Yáñez et al., 2020; Boluarte-Carbajal et al., 2021; Cayo-Rojas et al., 2021; Cénat et al., 2021; Dantas et al., 2021; de Oliveira Andrade et al., 2021; Espinosa-Guerra et al., 2021; Fernandez et al., 2021; Ferreira et al., 2021; Feter et al., 2021; Flores-Torres et al., 2021; García-Espinosa et al., 2021; Giardino et al., 2020; Goularte et al., 2021; Landaeta-Díaz et al., 2021; Loret de Mola et al., 2021; Mautong et al., 2021; Mendonca et al., 2021; Nayak et al., 2021; Puccinelli et al., 2021; Ribeiro et al., 2021; Serafim et al., 2021; Souza et al., 2021; Torrente et al., 2021, 2021; Vitorino et al., 2021; Werneck et al., 2021; Zhang et al., 2021, 2021; Caycho-Rodriguez et al., 2022;; da Silva Júnior et al., 2021; Robles et al., 2021). Among all the anxiety survey tools used, the Generalised Anxiety Symptoms 7-items scale (GAD-7) was the most common (51.85%), followed by the Depression, Anxiety and Stress Scale – 21 Items (DASS-21) (18.52%), the Hospital Anxiety and Depression Scale (HADS) (9.26%), Beck Anxiety Inventory (BAI) (3.70%) and nine others (each 1.85%). The cut-off values to determine the overall prevalence as well as severe anxiety varied across studies. In the random-effects model, the pooled prevalence of anxiety was 35% (95% CI: 31–38%) in the 54 studies (Fig. 2a). This finding suggests that, on average, 35% of the adults in Latin America had anxiety symptoms during COVID-19. Based on a normal distribution, its prediction internal is 5−75%, and the prevalence of anxiety symptoms in any comparable study will fall in this range.

Fig. 2.

The square markers indicate the prevalence of insomnia symptoms among population groups of interest. The diamonds represent the pooled estimates. (a) Forest plot indicating the pooled prevalence of anxiety among included studies. (b) Forest plot indicating the pooled prevalence of depression among included studies. (c) Forest plot indicating the pooled prevalence of distress among included studies. (d) Forest plot indicating the pooled prevalence of insomnia among included studies.

A total of 49 samples from 46 studies reported the prevalence of depression among 139 559 respondents (Badellino et al., 2020, 2022; Campos et al., 2020; Civantos et al., 2020; Cortés-Álvarez et al., 2020; Dal'Bosco et al., 2020; De Boni et al., 2020; Fernández et al., 2020; Giardino et al., 2020; Guiroy et al., 2020; Martinez et al., 2020; Medeiros et al., 2020; Mora-Magaña et al., 2020; Passos et al., 2020; Paz et al., 2020; Samaniego et al., 2020; Schuch et al., 2020; Antiporta et al., 2021; Boluarte-Carbajal et al., 2021; de Oliveira Andrade et al., 2021; Espinosa-Guerra et al., 2021; Ferreira et al., 2021; Feter et al., 2021; García-Espinosa et al., 2021; Goularte et al., 2021; Loret de Mola et al., 2021; Mautong et al., 2021; Mendonca et al., 2021; Nayak et al., 2021; Puccinelli et al., 2021; Ribeiro et al., 2021; Schmitt Jr et al., 2021; Serafim et al., 2021; Souza et al., 2021; Torrente et al., 2021, 2021; Villela et al., 2021; Zhang et al., 2021, 2021; Caycho-Rodriguez et al., 2022; Robles et al., 2021). Among all the depression survey tools, the Patient Health Questionnaire (PHQ)-9 was the most frequently used (50%), followed by DASS-21 (21.74), HADS (10.87%), the Centre for Epidemiologic Studies Depression Scale (CESD) (4.35%) and six others (each 2.17%). Analysing the random-effects model, the pooled prevalence of depression was 35% (95% CI: 31−39%) among the 46 studies (Fig. 2b). This finding suggests that, on average, 35% of the adults in Latin America had depression symptoms during COVID-19. Its prediction internal is 7−71%.

Thirteen studies studied mental distress among 10 335 participants (Chen et al., 2020; Civantos et al., 2020; Cortés-Álvarez et al., 2020; Fernández et al., 2020; Reidy, 2020; Samaniego et al., 2020; Yáñez et al., 2020; Boluarte-Carbajal et al., 2021; Espinosa-Guerra et al., 2021; Ferreira et al., 2021; Zhang et al., 2021). Among all the distress survey tools, DASS-21 was the most frequently used (30.77%), followed by COVID-19 Peritraumatic Distress Index (CPDI), Impact of Event Scale – Revised (IES) and K6 (15.38% each) and three others (7.69% each). In the random-effects model, the pooled prevalence of distress was 32% (95% CI: 25–40%) (Fig. 2c). This finding suggests that, on average, 32% of the adults in Latin America had distress symptoms during COVID-19. Its prediction interval is 1–79%.

Nine samples from seven studies (Giardino et al., 2020; Samaniego et al., 2020; Brito-Marques et al., 2021; Goularte et al., 2021; Mota et al., 2021; Scotta et al., 2021; Robles et al., 2021) studied insomnia among 12 134 respondents. The Insomnia Severity Index (ISI) (71.43%) was used most often, followed by Diagnostic and Statistical Manual (DSM) (28.57). In the random-effects model, the pooled prevalence of insomnia was 35% (95% CI: 25–46%) (Fig. 2d). Its prediction interval is 1–86%. The finding suggests that, on average, 35% of the adults in Latin America had insomnia symptoms during COVID-19 and the prevalence of insomnia symptoms in any comparable study will fall in this range.

The overall prevalence of mental health symptoms in frontline healthcare workers, general healthcare workers, the general population and students in Latin America was 37%, 34%, 33% and 45%, respectively. The overall prevalence rates of mental health symptoms that exceeded the cut-off values of mild, moderate and severe symptoms were 54%, 32% and 14%, respectively (Table 2). The pooled prevalence rates of mental health symptoms in South America, Central America, countries speaking Spanish and countries speaking Portuguese were 36%, 28%, 30% and 40%, respectively (Table 2). Subgroup analyses results on the anxiety, depression, distress and insomnia by population, severity, region and instrument are reported in Table 3.

Table 2.

Pooled prevalence estimates of mental health symptoms by outcome, population, severity and region subgroups during the COVID-19 pandemic

First-level subgroup	Second-level subgroup	Prevalence (%)	95% CI (%)	p Value
Aggregated		35	32–37	<0.001
Outcome	Anxiety	35	31–38	<0.001
	Depression	35	31–39	<0.001
	Distress	32	25–40	<0.001
	Insomnia	35	25–46	<0.001
Population	Frontline HCWs	37	24–51	<0.001
	General HCWs	34	29–39	<0.001
	General population	33	30–37	<0.001
	Students	45	37–53	<0.001
Severity	Above mild	54	50–58	<0.001
	Above moderate	32	30–35	<0.001
	Above severe	14	12–17	<0.001
	Overall	32	22–44	<0.001
Region	South America	36	33–38	<0.001
	Central America	28	24–33	<0.001
	Countries speaking Spanish	30	27–33	<0.001
	Countries speaking Portuguese	40	36–43	<0.001
Quality	Studies with high quality	42	38–45	<0.001
	Studies with medium quality	31	28–34	<0.001

CI, confidence interval.

Table 3.

Subgroup analyses of the prevalence of anxiety, depression and insomnia symptoms

Groups	Subgroups	Anxiety	Depression	Distress	Insomnia
Number of studies		54	46	13	7
Number of samples		56	49	13	9
Number of prevalence		95	87	21	19
Number of participants		128 060	139 559	10 335	12 134
Aggregated		35%, 95% CI: 31–39%	35%, 95% CI: 31–39%	32%, 95% CI: 25–40%	35%, 95% CI: 25–46%
Population	Frontline HCWs	23%, 95% CI: 21–26%	37%, 95% CI: 34–39%	NA	NA
	General HCWs	34%, 95% CI: 26–42%	34%, 95% CI: 25–44%	30%, 95% CI: 19–43%	34%, 95% CI: 21–47%
	General population	34%, 95% CI: 29–40%	33%, 95% CI: 28–37%	32%, 95% CI: 23–43%	NA
	Students	43%, 95% CI: 33–53%	54%, 95% CI: 42–65%	NA	31%, 95% CI: 2–75%
Severity	Above mild	55%, 95% CI: 48–61%	51%, 95% CI: 45–56%	55%, 95% CI: 42–67%	61%, 95% CI: 52–69%
	Above moderate	32%, 95% CI: 29–36%	32%, 95% CI: 28–37%	32%, 95% CI: 19–47%	32%, 95% CI: 27–37%
	Severe	14%, 95% CI: 11–17%	17%, 95% CI: 12–23%	17%, 95% CI: 13–22%	5%, 95% CI: 3–6%
Region	South America	37%, 95% CI: 32–41%	36%, 95% CI: 32–40%	33%, 95% CI: 25–41%	33%, 95% CI: 20–49%
	Central America	27%, 95% CI: 21–32%	27%, 95% CI: 20–34%	NA	45%, 95% CI: 37–53%
	Countries speaking Spanish	29%, 95% CI: 23–35%	29%, 95% CI: 25–34%	32%, 95% CI: 22–42%	34%, 95% CI: 21–48%
	Countries speaking Portuguese	40%, 95% CI: 35–45%	41%, 95% CI: 35–47%	33%, 95% CI: 21–46%	37%, 95% CI: 18–59%
Instrument		GAD: 32%, 95% CI: 26–38%	PHQ: 37%, 95% CI: 30–45%	CPDI: 35%, 95% CI: 10–65%	ISI: 32%, 95% CI: 17–49%
		DASS-21: 35%, 95% CI: 29–40%	DASS-21: 34%, 95% CI: 21–47%, I2: 99.9%	IES: 18%, 95% CI: 12–24%	DSM: 48%, 95% CI: 40–55%

CI, confidence interval.

Quality of the studies

Of all studies, 30 studies (48.39%) were of high quality, and 32 studies (51.61%) were of medium quality (Table 1). The subgroup analysis suggests the high-quality studies reported a higher prevalence of mental health symptoms in Latin America (42%) than those of medium quality (31%) (Table 2).

Detection of publication bias

The Doi plot and Luis Furuya–Kanamori index were used to quantify publication bias rather than the funnel plot and Egger's statistics (Furuya-Kanamori et al., 2018; Kounou et al., 2020). The symmetrical, hill-shaped Doi plot and a Luis Furuya–Kanamori (LFK) index of −0.81 indicated ‘no asymmetry’ and a lower likelihood of publication bias (Fig. 3).

Fig. 3.

The Doi plot and the Luis Furuya–Kanamori (LFK) index for publication bias. ES, effect size.

Discussion

The analysis of 62 studies with 196 950 participants from Latin America generated pooled prevalence of anxiety, depression, distress and insomnia of 35%, 35%, 32% and 35%, respectively. Notably, this meta-analysis is the first to investigate the prevalence of mental health symptoms during the COVID-19 crisis in Latin America. The anxiety levels in Latin America were significantly higher than other regions, such as China (25%; p < 0.001 (We compared the prevalence between two regions using t-test https://www.medcalc.org/calc/comparison_of_proportions.php)) (Ren et al., 2020) and Spain (20%; p < 0.001) (Chen et al., 2021). Latin America has a long-standing history of scarce resources to deal with mental health symptoms (Alarcón, 2003), which could explain the higher prevalence of mental health symptoms among Latin Americans as revealed by this meta-analysis. Notably, the pooled prevalence of mental health symptoms was lower in Latin America than in Africa and South Asia, as reported by other meta-analyses (Hossain et al., 2020; Chen et al., 2021). These cross-region differences may be due to multiple reasons, including heterogeneity in COVID-19 infection rate and mortality rate, variations in and timing of containment strategies adopted by countries across regions (Middelburg and Rosendaal, 2020), and the varying degrees of resources available, including personal protective equipment (PPE), to address mental health symptoms (Batra et al., 2020).

The prevalence of mental health symptoms was higher in South America than Central America (36% v. 28%; p < 0.001). This difference might be attributed to variations across these countries in the evolution of the pandemic (e.g. some countries such as Peru and Brazil started out well but deteriorated rapidly) (We appreciate a reviewer raising this point of discussion.), the provision and availability of PPE, healthcare facilities and capacities, the stringency of the COVID-19 responses and the political climate (Garcia et al., 2020). Previous research noted that South America generally has a high degree of political polarisation, which resulted in conflicting information being conveyed to the public that could increase the burden of COVID-19 and its associated psychological corollaries (Garcia et al., 2020). In addition, public health actions or decisions were made mostly at municipal and state levels rather than at central government levels, and the lack of central coordination posed several challenges in the control of the pandemic, contributing to an increased psychological burden (Garcia et al., 2020).

Based on the evidence of individual studies, our study found a higher prevalence of mental health symptoms among frontline HCWs (37%, p < 0.001) and university students (45%, p < 0.001) than the general population and general HCWs (Batra et al., 2020; Luo et al., 2020; Pappa et al., 2020). The vulnerabilities of frontline healthcare workers are often attributed to a higher risk of infection, burnout, the more direct exposure to suffering or dying patients, fear of COVID-19 transmission to their family members and job loss (Xiang et al., 2020; Bhandari et al., 2021). The greater prevalence of mental health symptoms among university students can be explained by the uncertainties surrounding the course of the pandemic and the sudden transition to online education (Adedoyin and Soykan, 2020; Batra et al., 2021). Moreover, many businesses scaled down their recruitment efforts, leading to limited employment opportunities for students and more competition in the graduate labour market (Reidy, 2020). These challenges added to the mental health burden among university students.

Study limitations

There are a few limitations that merit discussion. First our analysis reveals substantial heterogeneities across studies in the type of survey instruments used and the cut-off scores, both of which may affect the interpretation of the findings. Second, not all Latin American countries have been well-studied, therefore our results may have limited generalisability for the less studied nations. Third, a majority of the included studies were cross-sectional, which provides no information on the prevalence over time during the pandemic. In addition, studies included in this meta-analysis relied on self-reported data of psychological symptoms by the participants and hence do not constitute mental health diagnosis from clinicians. Fourth, other outcomes, such as post traumatic stress disorder (PTSD), suicidal ideation and burnout, were not studied in this meta-analysis, leaving opportunities for prospective studies. Last, a language bias is expected because of the language restriction (only English) applied in this study. The systematic search uncovered eight papers (7.5%) that were not included for language reasons out of 107 eligible papers.

Practical implications

First, our systematic review and meta-analysis support evidence-based medicine by revealing a high proportion of mental health symptoms among the general population and healthcare workers during the COVID-19 pandemic in Latin America. However, our systematic review also reveals there is a lack of evidence in many Latin American countries to guide the relevant practice of evidence-based medicine on this topic. Only 12 of the 33 Latin American countries have been studied, leaving 21 countries without any studies to assist the practice of evidence-based healthcare. For instance, no relevant research has been done in Venezuela, the fifth-biggest South American country with a population of 28 million, in Chile, the sixth biggest South American country with a population of 18 million, nor in Guatemala (18 million population), Cuba (11 million population) and the Dominican Republic (11 million population), respectively the second, fourth and fifth most populous countries in Central America. In practice, healthcare organisations in those unstudied countries may use our results in the same region as approximate evidence before direct evidence in those countries emerges.

Our findings that the prevalence of mental health symptoms was higher in South America than Central America (36% v. 28%; p < 0.001) provide evidence for international healthcare organisations, such as the World Psychiatric Association, on their assistance and resource allocation efforts. Our findings of a higher prevalence of mental health symptoms among frontline healthcare workers (37%, p < 0.001) and university students (45%, p < 0.001) than the general population (33%) and general healthcare workers (34%) suggest psychiatric and healthcare organisations should prioritise frontline healthcare workers and university students in Latin America.

Conclusions

This meta-analysis, to our knowledge, provides the first pooled estimates of mental health symptoms among key demographic groups during the COVID-19 crisis in Latin America. The meta-analytical findings of this study underscore the high prevalence of mental health symptoms in Latin Americans during the COVID-19 crisis. Hence, we call for more research to identify people vulnerable to mental health symptoms to enable evidence-based medicine during the pandemic.