As the financial, occupational, and social impacts of the COVID-19 pandemic became apparent, voices across clinical psychological science and medicine raised concerns regarding potential consequences for population-wide mental health (Fiorillo & Gorwood, Reference Fiorillo and Gorwood2020; Gruber et al., Reference Gruber, Prinstein, Clark, Rottenberg, Abramowitz, Albano and Weinstock2021; Pfefferbaum & North, Reference Pfefferbaum and North2020). Some early predictions anticipated a broad decline in mental health, including development of psychopathology in previously healthy individuals and exacerbated symptoms in clinical populations (Fiorillo & Gorwood, Reference Fiorillo and Gorwood2020; Gruber et al., Reference Gruber, Prinstein, Clark, Rottenberg, Abramowitz, Albano and Weinstock2021). Stress is a well-established risk factor for the development and worsening of psychopathology (Liu & Miller, Reference Liu and Miller2014; McLaughlin, Reference McLaughlin, Harkness and Hayden2020), particularly internalizing symptoms such as anxiety and depression (Faravelli et al., Reference Faravelli, Lo Sauro, Lelli, Pietrini, Lazzeretti, Godini and Ricca2012; Kendler & Gardner, Reference Kendler and Gardner2016), and these predictions largely followed from conceptualizations of the COVID-19 pandemic as both an acute and chronic stressor (Bridgland et al., Reference Bridgland, Moeck, Green, Swain, Nayda, Matson and Takarangi2021; Waugh, Leslie-Miller, & Cole, Reference Waugh, Leslie-Miller and Cole2022).
Other perspectives argued that mental health impacts of the COVID-19 pandemic would be more nuanced than population-level decline (Koushik, Reference Koushik2020; Mancini, Reference Mancini2020). For example, decreased work and social obligations might provide temporary relief from anxiety symptoms, while mandated social isolation might exacerbate depression (Koushik, Reference Koushik2020). An increased sense of togetherness, previously observed in communities experiencing mass trauma, could even benefit mental health (Lau et al., Reference Lau, Chi, Cummins, Lee, Chou and Chung2008; Mancini, Reference Mancini2020). Inter-individual differences may also play a role, with vulnerable populations shouldering the burden of worsening symptoms (Mancini, Reference Mancini2020). Such predictions are in line with prospective studies of the psychological impacts of disaster, with only a minority of participants (typically 30% or less) developing severe, chronic psychopathology (Bonanno, Brewin, Kaniasty, & Greca, Reference Bonanno, Brewin, Kaniasty and Greca2010).
Early systematic reviews of general population mental health during the first few months of the COVID-19 pandemic reported startling rates of clinically significant anxiety (ranging from 26% to 38%), depression (26–34%), and psychological distress (30–38%; Deng et al. Reference Deng, Zhou, Hou, Silver, Wong, Chang and Huang2021; Krishnamoorthy, Nagarajan, Saya, & Menon, Reference Krishnamoorthy, Nagarajan, Saya and Menon2020; Luo, Guo, Yu, Jiang, & Wang, Reference Luo, Guo, Yu, Jiang and Wang2020; Necho, Tsehay, Birkie, Biset, & Tadesse, Reference Necho, Tsehay, Birkie, Biset and Tadesse2021; Salari et al. Reference Salari, Hosseinian-Far, Jalali, Vaisi-Raygani, Rasoulpoor, Mohammadi and Khaledi-Paveh2020). However, the studies covered by these reviews are largely epidemiological and cross-sectional, often with single item or unvalidated assessments. The absence of pre-pandemic timepoints in particular limits inferences about change over time that is potentially attributable to the pandemic. To our knowledge, only one meta-analysis has synthesized studies with a pre-pandemic timepoint (Robinson, Sutin, Daly, & Jones, Reference Robinson, Sutin, Daly and Jones2022). Across 65 longitudinal cohort studies, there was a small overall increase in symptoms of psychopathology, with the largest symptom increases observed in studies that sampled participants early in the pandemic (March–April 2020). Although this meta-analysis focused on longitudinal cohort studies, samples were largely European and North American, and did not include symptom clusters beyond the broad categories of anxiety, depression, and psychological distress.
Important questions also remain regarding variability in trajectories across symptom types and populations. For example, OCD symptoms related to contamination and health behaviors (e.g. handwashing, disinfecting surfaces) may have been impacted differently than other kinds of internalizing symptoms (Guzick et al., Reference Guzick, Candelari, Wiese, Schneider, Goodman and Storch2021). PTSD symptoms might emerge in populations such as healthcare workers (Bridgland et al., Reference Bridgland, Moeck, Green, Swain, Nayda, Matson and Takarangi2021). Finally, theoretical and empirical frameworks distinguish between acute autonomic reactions to imminent threat (fear) v. prolonged apprehension involving chronic or distal threat (anxiety; Kotov et al., Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby and Zimmerman2017; Öhman, Reference Öhman, Lewis, Haviland-Jones and Barrett2008). Whereas fear might be expected to show a sharp increase relative to pre-pandemic followed by a steady decline, the evolving and prolonged nature of the pandemic suggests a potentially different course for anxiety. The present systematic review gives in-depth consideration to these and other internalizing-related symptom clusters. We adopt a global, lifespan, and transdiagnostic perspective, with findings reported separately for unselected samples, samples with diagnosed psychopathology, and selected samples such as patients with preexisting medical conditions or other special characteristics (e.g. veterans).
Method
Search strategy
A systematic review was conducted in accordance with PRISMA guidelines (Moher, Liberati, Tetzlaff, & Altman, Reference Moher, Liberati, Tetzlaff and Altman2009) and preregistered on PROSPERO (#CRD42021255042). We searched PubMed and PsycINFO for studies that assessed mental health symptoms both before and during the COVID-19 pandemic. Keywords included (COVID-19 OR coronavirus OR COVID OR pandemic OR SARS-CoV-2 OR social distancing OR quarantine OR lockdown) AND (anxi* OR obsess* OR trauma* OR fear OR panic OR agoraphobi* OR social anxiety OR separation anxiety OR acute stress* OR depress*). A database search was conducted from 29 January 2020 (the day before the World Health Organization designated the COVID-19 pandemic a Public Health Emergency of International Concern) to 24 June 2021. Complete search strategies for each database and deviations from the preregistration can be found in online Supplementary Materials.
Study selection
Titles and abstracts were screened according to eligibility criteria (outlined below). Full-text screening was conducted by four authors (M. B., T. I. E., I. C. O., and L. S. H.). Where eligibility was uncertain, a consensus decision was made following discussion. Study selection process and reasons for exclusion are described in Fig. 1.
PRISMA diagram showing study selection process.
Inclusion criteria. Studies were included if they met all of the following criteria: (a) published or in press on or after 29 January 2020; (b) published in English; (c) participants and/or studies based in a country or region with confirmed COVID-19 cases, lockdown measures, and/or quarantine; and (d) assessed symptoms of mental health both prior to (no later than 29 January 2020) and during the COVID-19 pandemic, either in the same sample at both timepoints or in two nationally or regionally representative samples.
Exclusion criteria. Studies were excluded if they met any of the following criteria: (a) cross-sectional design, or repeated measures in different, non-representative samples; (b) administered mental health treatment between the pre-pandemic and peri-pandemic assessments (healthy and no-treatment control groups remained eligible); (c) did not present original empirical work (e.g. commentaries, reviews).
Methodological quality. Included studies were assessed for methodological quality by the first author across four domains: sample size, sampling methodology, response rate, and measurement (see online Supplementary Materials).
Results
The search terms yielded 981 results from PsycINFO and 8666 results from PubMed, for a total of 9647 potentially eligible items. Following the removal of 672 duplicate items, 8646 items were excluded during title and abstract screening, and 232 items were excluded during full-text screening, yielding 97 studies which were included in the final review (see Figs 2–3 and Table 1 for included studies). Symptom changes described below were statistically significant unless otherwise specified.
Distribution of studies in each symptom cluster reporting either an increase, decrease, or no change in symptoms. Colors reflect the reported statistical significance of each symptom change.
Spread of effect sizes for studies where Cohen's d was reported or could be calculated (k = 55). Effect sizes <0 indicate symptom decrease; effect sizes >0 indicate symptom increase. See Supplementary Table S4 for studies corresponding to point labels.
Summary of included studies, sample characteristics, and symptom change results
OCD, obsessive-compulsive disorder; MDD, major depressive disorder; HIV, human immunodeficiency virus; Y-BOCS, Yale-Brown Obsessive Compulsive Scale; CIS-R, Clinical Interview Schedule-Revised; CBCL, Child Behaviour Check List; OCI-R, Obsessive-Compulsive Inventory-Revised; PCL-5, PTSD Checklist for DSM-5; SCL-90-R, Symptom Checklist 90 Revised; CY-BOCS, Children's Yale-Brown Obsessive Compulsive Scale; DASS, Depression Anxiety Stress Scale (42 or 21 item version specified); BAI, Beck Anxiety Subscale; PHQ, Patient Health Questionnaire (2, 4, 8, or 9-item version specified); PSWQ, Penn State Worry Questionnaire; RCADS, Revised Child Anxiety and Depression Scales; HADS, Hospital Anxiety and Depression Scales; GAD, Generalized Anxiety Disorder (2 or 7-item version specified); MASC, Multidimensional Anxiety Inventory for Children; PSS, Perceived Stress Scale (4-item version specified); GAI, Geriatric Anxiety Inventory (Short Form [SF] specified); HAM-A, Hamilton Anxiety Scale; PROMIS, Patient-Reported Outcomes Measurement Information System; SCA-S, Spence Children's Anxiety Scale; MINI, Mini International Neuropsychiatric Interview; IDS-30, Inventory of Depressive Symptomology 30; CIDI, Composite International Diagnostic Interview; CDI, Children's Depression Inventory (Short Form [S] specified); CES-D, Center for Epidemiologic Studies Depression Scale; EPDS, Edinburgh Postnatal Depression Scale; HAM-D, Hamilton Depression Scale; SMFQ-C, Short Mood and Feelings Questionnaire – Child Version; GDS, Geriatric Depression Scale; QIDS, Quick Inventory of Depressive Symptoms; NVDDI-E, Neurological Disorders Depression Inventory for Epilepsy; DRS, Depression Rating Scale; BSI, Brief Symptom Inventory; SDQ, Strengths and Difficulties Questionnaire; IDAS-II, Inventory of Depression and Anxiety Symptoms – II; STAI, State-Trait Anxiety Inventory; GHQ-12, General Health Questionnaire; K10 or K6, Kessler Psychological Distress Scale (10 or 6-item version); PSI-SF, Parenting Stress Index – Short Form.
Note. In many cases, authors only reported T1 data (e.g. sample size) for participants for whom T2 data were also available. An equivalent N at T1 and T2 may be indicative of participants with available data at both timepoints, rather than the total sample being the same size at both timpoints. Response rate is reported as the percentage of participants with available data at T1 who responded to an invitation to participate in T2 follow-up data collection. Many studies invited only a subset of participants from T1 to participate at T2; hence, response rate is not always obtained by diving N at T2 by N at T1.
Obsessive-compulsive disorder (k = 11)
Three of four studies in unselected (including undergraduate student) samples reported increases in overall OCD symptoms, particularly washing and contamination symptoms (Cox & Olatunji, Reference Cox and Olatunji2021; Jelinek, Göritz, Miegel, Moritz, & Kriston, Reference Jelinek, Göritz, Miegel, Moritz and Kriston2021; Knowles & Olatunji, Reference Knowles and Olatunji2021). The fourth, a longitudinal cohort study of 2117 Brazilian university employees, found no change in the prevalence of OCD diagnoses from pre- to peri-pandemic (Brunoni et al., Reference Brunoni, Suen, Bacchi, Razza, Klein, Dos Santos and Benseñor2021). Of five studies including participants with pre-existing OCD diagnoses (N = 60 to N = 270), four found that OCD severity increased. In one study of an outpatient clinic sample, pre-existing contamination symptoms predicted increases in overall OCD severity (Davide et al., Reference Davide, Andrea, Martina, Andrea, Davide and Mario2020). A study of children and adolescents with OCD treated at a university psychiatry department only found changes in washing and contamination symptoms (Tanir et al., Reference Tanir, Karayagmurlu, Kaya, Kaynar, Türkmen, Dambasan and Coşkun2020). In a Spanish sample of 127 clinic outpatients with OCD, ~31% showed at least a moderate increase in severity (over 25%), with the remaining changes ranging from a small increase to a small decrease (Alonso et al., Reference Alonso, Bertolín, Segalàs, Tubío, Real, Mar-Barrutia and Menchón2021). Similarly, about half of a sample of 84 Indian adults with OCD showed no change, while the other half mostly showed increases in severity of less than 5% (Chakraborty & Karmakar, Reference Chakraborty and Karmakar2020). In samples with psychiatric diagnoses other than OCD, a study of 80 children and adolescents with various neurologic and psychiatric disorders found an increase in OCD symptoms (Conti et al., Reference Conti, Sgandurra, De Nicola, Biagioni, Boldrini, Bonaventura and Battini2020), while a study of 35 Catalán adults with autism found no changes (Lugo-Marín et al., Reference Lugo-Marín, Gisbert-Gustemps, Setien-Ramos, Español-Martín, Ibañez-Jimenez, Forner-Puntonet and Ramos-Quiroga2021).
Across samples, OCD symptoms tended to increase, with few exceptions. Calculable effect sizes (k = 4) were small to moderate (median Cohen's d = 0.59, range = 0.11–0.79). Consistent with predictions, symptom increase was more pronounced for washing-checking domains and in diagnosed samples.
Post-traumatic stress disorder (k = 5)
Findings for PTSD were inconclusive, driven in part by the small total number of studies. The largest study available, in a nationally representative panel survey of 3078 American veterans, did not find significant changes in PTSD probable diagnosis prevalence (Hill et al., Reference Hill, Nichter, Na, Norman, Morland, Krystal and Pietrzak2021). In a sample of 473 Canadian adults, no changes in PTSD symptoms were identified (Minhas et al., Reference Minhas, Belisario, González-Roz, Halladay, Murphy and MacKillop2021). An increase in PTSD symptoms was observed in 80 Italian children and adolescents with a pre-existing neuro-developmental disorder (Conti et al., Reference Conti, Sgandurra, De Nicola, Biagioni, Boldrini, Bonaventura and Battini2020), as well as in 85 German adults with and without psychiatric disorders (Seitz, Bertsch, & Herpertz, Reference Seitz, Bertsch and Herpertz2021). By contrast, there was a decrease in PTSD severity among 76 older adults with a pre-existing PTSD diagnosis compared to trauma-exposed controls in the USA (Rutherford et al., Reference Rutherford, Choi, Chrisanthopolous, Salzman, Zhu, Montes-Garcia and Roose2021).
Fear (k = 10)
Six studies assessed fear/autonomic anxiety in unselected samples, with variable results. A study of 217 Indian undergraduates found a small increase in fear (Saraswathi et al., Reference Saraswathi, Saikarthik, Senthil Kumar, Madhan Srinivasan, Ardhanaari and Gunapriya2020), as did two smaller studies [99 adult women in Poland (Ilgen, Kurt, Aydin, Bilen, & Kula, Reference Ilgen, Kurt, Aydin, Bilen and Kula2021) and 68 Italian undergraduates (Bussone, Pesca, Tambelli, & Carola, Reference Bussone, Pesca, Tambelli and Carola2020)]. By contrast, a study of 2364 Chinese undergraduates found a slight decrease in fear (Yang, Ji, et al., Reference Yang, Ji, Yang, Wang, Zhu and Cai2021), as did a study of 66 Brazilian pharmacy students (Campos, Campos, Bueno, & Martins, Reference Campos, Campos, Bueno and Martins2021). The remaining study, in a crowdsourced sample of 146 American adults, found no change (Haliwa, Wilson, Lee, & Shook, Reference Haliwa, Wilson, Lee and Shook2021).
Both studies including samples with pre-existing psychiatric diagnoses found no change in fear [275 American adults with autism spectrum disorder drawn from an existing research registry (Adams, Zheng, Taylor, & Bishop, Reference Adams, Zheng, Taylor and Bishop2021); 1181 adults with internalizing disorders drawn from a prospective longitudinal study in the Netherlands (Pan et al., Reference Pan, Kok, Eikelenboom, Horsfall, Jörg, Luteijn and Penninx2021)]. The latter study did report a small increase in fear among 336 healthy control participants (recruited through primary care settings; Pan et al., Reference Pan, Kok, Eikelenboom, Horsfall, Jörg, Luteijn and Penninx2021). Two studies of selected samples showed stronger effects, including a clinically significant increase in fear in 595 Turkish cancer patients (Yildirim, Poyraz, & Erdur, Reference Yildirim, Poyraz and Erdur2021), and an increase in the proportion of 63 pregnant participants meeting the threshold for moderate or severe fear (Ayaz et al., Reference Ayaz, Hocaoğlu, Günay, Yardımcı, Turgut and Karateke2020).
Taken together, findings suggest that medically selected samples experienced a clinically significant increase in fear, but this pattern did not apply to unselected samples or those with pre-existing psychiatric vulnerabilities, who tended to experience at most small increases. Although several studies reporting null results had small sample sizes, a few larger studies also showed no change in fear severity, suggesting that null results are unlikely to be due to low statistical power.
Anxiety (k = 47)
Of 21 studies prospectively examining anxiety in unselected samples, 11 found an increase in anxiety and eight found no change, while two studies reported a decrease. One of the largest studies, a randomly sampled cohort study of 113 928 German adults ages 20–74, reported an increase in anxiety for participants under 60 years of age only, with women ages 20–39 showing the largest increase (Peters, Rospleszcz, Greiser, Dallavalle, & Berger, Reference Peters, Rospleszcz, Greiser, Dallavalle and Berger2020). In a prospective cohort study of 1237 French adults (Ramiz et al., Reference Ramiz, Contrand, Rojas Castro, Dupuy, Lu, Sztal-Kutas and Lagarde2021), the proportion of participants with ‘possible anxiety’ (GAD-7 score >4) increased over time, with the most marked increases occurring for women ages 23–49 and ages 70 or above. A study comparing two randomly sampled, nationally representative samples of Czech adults (Winkler et al., Reference Winkler, Mohrova, Mlada, Kuklova, Kagstrom, Mohr and Formanek2021) similarly found that the prevalence of anxiety disorders increased from 8% in November 2017 to 13% in May 2020. Younger adults, women, those who struggled to retain employment, and those without a high school diploma had the highest absolute rates of anxiety disorder during the second wave of the pandemic.
Well-powered studies that did not find changes in anxiety include a representative sample of 1041 Irish adults (Hyland et al., Reference Hyland, Shevlin, Murphy, McBride, Fox, Bondjers and Vallières2021) and a nationally representative study of 944 604 American adults compared to a 2019 propensity-matched sample (Jacobs & Burch, Reference Jacobs and Burch2021). However, interpretability of these large studies may be offset by other methodological constraints (use of a cut-off score and single-item assessment, respectively).
The only two studies reporting a decrease in anxiety were a study of 2364 Chinese undergraduates (assessed October 2019, February 2020, and May 2020; Yang, Ji., et al., Reference Yang, Ji, Yang, Wang, Zhu and Cai2021) and 2117 Brazilian adults (May–July 2020 compared to 2008–2010, with no change compared to 2016–2018; Brunoni et al., Reference Brunoni, Suen, Bacchi, Razza, Klein, Dos Santos and Benseñor2021).
Five studies examined anxiety in psychiatric samples. One study (Pan et al., Reference Pan, Kok, Eikelenboom, Horsfall, Jörg, Luteijn and Penninx2021) found an increase in worry across 1181 adults with psychiatric illness and 336 healthy controls from three longitudinal cohort studies in the Netherlands. Another longitudinal cohort study found an increase in 147 healthy controls, but not 345 adults with bipolar disorder (Yocum, Zhai, McInnis, & Han, Reference Yocum, Zhai, McInnis and Han2021). Two smaller studies also found an increase in anxiety; in 76 Chinese participants receiving methadone maintenance treatment for substance use disorder (Liu et al., Reference Liu, Jin, Zhang, Zhang, Li and Ma2021); and in 46 American older adults with PTSD, but not 30 trauma-exposed controls (Rutherford et al., Reference Rutherford, Choi, Chrisanthopolous, Salzman, Zhu, Montes-Garcia and Roose2021). A third study found no change in a sample of 35 Catalán adults with autism (Lugo-Marín et al., Reference Lugo-Marín, Gisbert-Gustemps, Setien-Ramos, Español-Martín, Ibañez-Jimenez, Forner-Puntonet and Ramos-Quiroga2021).
Seven studies prospectively assessed changes in anxiety in children and adolescents with remarkably consistent findings; all but one found at least a small increase in anxiety coinciding with the onset of the pandemic (Breaux et al., Reference Breaux, Dvorsky, Marsh, Green, Cash, Shroff and Becker2021; Conti et al., Reference Conti, Sgandurra, De Nicola, Biagioni, Boldrini, Bonaventura and Battini2020; Magson et al., Reference Magson, Freeman, Rapee, Richardson, Oar and Fardouly2021; Rogers, Ha, & Ockey, Reference Rogers, Ha and Ockey2021). Studies reporting an increase in anxiety include a sample of 775 children and adolescents assessed via an experience-sampling application in Australia (Arjmand, Seabrook, Bakker, & Rickard, Reference Arjmand, Seabrook, Bakker and Rickard2021); a representative sample of 844 Dutch children and adolescents (compared to a representative sample from 2018; Luijten et al., Reference Luijten, van Muilekom, Teela, Polderman, Terwee, Zijlmans and Haverman2021); and 136 Canadian children and adolescents (De France, Hancock, Stack, Serbin, & Hollenstein, Reference De France, Hancock, Stack, Serbin and Hollenstein2021), who demonstrated a marginal increase in anxiety that was driven by an increase in girls only.
Ten studies prospectively assessed changes in anxiety in older adults and other medically vulnerable samples. Relatively larger studies tended to find an increase in anxiety, including a multinational sample of 435 adults with systemic sclerosis recruited during medical visits (Thombs et al., Reference Thombs, Kwakkenbos, Henry, Carrier, Patten, Harb and Benedetti2020); 538 Chinese older adults with two or more chronic health conditions (Wong et al., Reference Wong, Zhang, Sit, Yip, Chung, Wong and Mercer2020); 721 Chilean older adults from a random community sample (Herrera et al., Reference Herrera, Elgueta, Fernández, Giacoman, Leal, Marshall and Bustamante2021); and 133 American adults with HIV (but not 54 healthy controls; Cooley, Nelson, Doyle, Rosenow, & Ances, Reference Cooley, Nelson, Doyle, Rosenow and Ances2021). Null findings were observed in a prospective observational study of 1051 patients with remitted breast cancer (Mink van der Molen et al., Reference Mink van der Molen, Bargon, Batenburg, Gal, Young-Afat, van Stam and Verkooijen2021); 450 Australian adults with type 2 diabetes (Sacre et al., Reference Sacre, Holmes-Truscott, Salim, Anstey, Drummond, Huxley and Shaw2021); and 411 Chinese older adults (Siew, Mahendran, & Yu, Reference Siew, Mahendran and Yu2021).
Studies of other selected samples also tended to find increases in anxiety. A longitudinal cohort study of 2288 American sexual and gender minority (SGM) adults found an increase in anxiety severity during the early stages of the pandemic (Flentje et al., Reference Flentje, Obedin-Maliver, Lubensky, Dastur, Neilands and Lunn2020). However, this increase was driven by participants who were relatively lower in anxiety, with those who screened positive for GAD prior to the pandemic showing no change. A study of 1028 recent mothers in a hospital-based birth cohort in Brazil found a twofold increase in GAD prevalence from 2019 (during pregnancy) to 2020 (Loret de Mola et al., Reference Loret de Mola, Martins-Silva, Carpena, Del-Ponte, Blumenberg, Martins and Cesar2021). A nationally representative study of 3078 predominantly male American veterans found an increase in the prevalence of positive GAD screenings from 7% to 9%, driven by a marked increase in anxiety severity among middle-aged veterans (Hill et al., Reference Hill, Nichter, Na, Norman, Morland, Krystal and Pietrzak2021).
Collectively, most studies found an increase in anxiety associated with the onset of the pandemic, although calculable effect sizes (k = 14) tended to be small (median Cohen's d = 0.16, range = −0.35 to 0.54). Findings were especially pronounced and consistent in child and adolescent samples and in medically vulnerable participants. Psychiatric samples also showed some vulnerability to an increase in anxiety, but this vulnerability was not more pronounced than that observed in unselected samples.
Depression (k = 74)
Twenty-seven studies prospectively assessed changes in incidence or severity of depression in unselected samples. Only two studies found a decrease in severity, one in 1020 Irish adults (Hyland et al., Reference Hyland, Shevlin, Murphy, McBride, Fox, Bondjers and Vallières2021) and one in 2364 Chinese undergraduate students (Yang, Ji, et al., Reference Yang, Ji, Yang, Wang, Zhu and Cai2021). The remaining studies found either an increase (k = 17) or no change (k = 8) in severity. In the largest study (N = 113 928 German adults; Peters et al., Reference Peters, Rospleszcz, Greiser, Dallavalle and Berger2020), incidence of moderate-to-severe depression symptoms increased from 6.4% 1–5 years before the pandemic to 8.8% in May 2020. Incidence of depression increased by 7.8% (from 4% in November 2017 to 11.8% in May 2020) across two nationally representative samples of over 3000 Czech adults (Winkler et al., Reference Winkler, Mohrova, Mlada, Kuklova, Kagstrom, Mohr and Formanek2021). Incidence doubled across two representative random samples of Chinese adults (from 6.3% of 4054 adults in 2017 to 14.8% of 1501 adults in April 2020; Zhao et al., Reference Zhao, Wong, Luk, Wai, Lam and Wang2020); and increased more than fivefold in a representative random sample of 715 Czech adults (Novotný et al., Reference Novotný, Gonzalez-Rivas, Kunzová, Skladaná, Pospíšilová, Polcrová and Stokin2020).
Studies comparing demographic groups found that younger adults (e.g. under age 60; Peters et al., Reference Peters, Rospleszcz, Greiser, Dallavalle and Berger2020) and women (Fruehwirth, Biswas, & Perreira, Reference Fruehwirth, Biswas and Perreira2021; Minhas et al. Reference Minhas, Belisario, González-Roz, Halladay, Murphy and MacKillop2021; Peters et al. Reference Peters, Rospleszcz, Greiser, Dallavalle and Berger2020) were especially vulnerable to increased depression incidence or severity. Studies of undergraduates had proportionately more non-significant results, which may be due in part to smaller sample sizes (although see findings from Yang, Ji, et al., Reference Yang, Ji, Yang, Wang, Zhu and Cai2021, above).
Of seven studies including psychiatric samples, findings were mixed, with increases in depression severity reported in a sample of 1181 Dutch participants with internalizing disorders and in 336 healthy controls (Pan et al., Reference Pan, Kok, Eikelenboom, Horsfall, Jörg, Luteijn and Penninx2021); 52 German adults with eating disorders (Giel, Schurr, Zipfel, Junne, & Schag, Reference Giel, Schurr, Zipfel, Junne and Schag2021); and 76 Chinese adults with substance use disorder (Liu et al., Reference Liu, Jin, Zhang, Zhang, Li and Ma2021). The remaining studies reported no change in depression in 275 American adults with autism (Adams et al., Reference Adams, Zheng, Taylor and Bishop2021); a decrease in 35 Catalán adults with autism (Lugo-Marín et al., Reference Lugo-Marín, Gisbert-Gustemps, Setien-Ramos, Español-Martín, Ibañez-Jimenez, Forner-Puntonet and Ramos-Quiroga2021); or non-significant changes (Rutherford et al., Reference Rutherford, Choi, Chrisanthopolous, Salzman, Zhu, Montes-Garcia and Roose2021; Yocum et al., Reference Yocum, Zhai, McInnis and Han2021).
Of nine studies with child and adolescent samples, all but two reported an increase in depression severity, with some studies reporting small effect sizes (e.g. Magson et al., Reference Magson, Freeman, Rapee, Richardson, Oar and Fardouly2021; Rogers et al., Reference Rogers, Ha and Ockey2021), and others reporting moderate-to-large effects (e.g. Breaux et al., Reference Breaux, Dvorsky, Marsh, Green, Cash, Shroff and Becker2021; Thorisdottir et al., Reference Thorisdottir, Asgeirsdottir, Kristjansson, Valdimarsdottir, Jonsdottir Tolgyes, Sigfusson and Halldorsdottir2021). Only one large study did not find changes (1778 Chinese children and adolescents recruited through school-based cluster sampling; Teng, Pontes, Nie, Griffiths, & Guo, Reference Teng, Pontes, Nie, Griffiths and Guo2021). Studies that examined demographic predictors tended to find worse outcomes for girls compared to boys (e.g. De France et al., Reference De France, Hancock, Stack, Serbin and Hollenstein2021; Thorisdottir et al., Reference Thorisdottir, Asgeirsdottir, Kristjansson, Valdimarsdottir, Jonsdottir Tolgyes, Sigfusson and Halldorsdottir2021) and for children whose parents experienced employment difficulties (e.g. Luijten et al., Reference Luijten, van Muilekom, Teela, Polderman, Terwee, Zijlmans and Haverman2021).
Of seven studies examining depression in older adults, five reported an increase in incidence or severity. In the largest study, a national opt-in panel survey of 16 644 older American adults (Barcellos, Jacobson, & Stone, Reference Barcellos, Jacobson and Stone2021), this increase was driven by an increase in women only. Of 13 studies examining changes in depression in other medically vulnerable individuals, seven reported increases in severity, though only two studies with sample sizes >100 found this increase. The remainder of larger studies reported no change (k = 4) or a decrease (k = 1). This pattern is potentially suggestive of a true null effect.
Seven studies assessed other selected samples, with all but the smallest reporting an increase in incidence or severity of depression, typically with moderate or large effect sizes. A study of 2288 American SGM individuals found a small increase in self-reported depression severity (Flentje et al., Reference Flentje, Obedin-Maliver, Lubensky, Dastur, Neilands and Lunn2020). In a sample of 419 American undergraduates (Fruehwirth et al., Reference Fruehwirth, Biswas and Perreira2021), increased depression was observed in both SGM and non-SGM students. Four studies assessed participants who were pregnant or newly post-partum, with Ns ranging from 50 (Lorentz et al., Reference Lorentz, Chagas, Perez, da Silva Cassol, Vettorazzi and Lubianca2021) to 1042 (Loret de Mola et al., Reference Loret de Mola, Martins-Silva, Carpena, Del-Ponte, Blumenberg, Martins and Cesar2021).
Across studies, findings for depression did not follow a clear pattern. While the majority of studies reported increased depression incidence or severity, other studies in similar populations found no change or a decrease. Effect sizes (k = 24) tended to be small (median Cohen's d = 0.22, range = −0.2 to 1.4).
General distress (k = 23)
Of 10 studies investigating general psychological distress in unselected samples, seven reported pandemic-related increases. These included the three largest studies, including two nationally representative samples (2032 American adults, Twenge & Joiner, Reference Twenge and Joiner2020; 13 754 adults in the UK, Chandola, Kumari, Booker, & Benzeval, Reference Chandola, Kumari, Booker and Benzeval2020). The magnitude of these increases varied, but tended to be large. Only two studies found a decrease in general distress, in 555 Chinese undergraduate students (Li, Cao, Leung, & Mak, Reference Li, Cao, Leung and Mak2020); and 71 Brazilian undergraduates (da Freitas, de Medeiros, & de Lopes, Reference da Freitas, de Medeiros and de Lopes2021). In both cases, absolute severity was low at both occasions.
Of four studies examining general distress in psychiatric samples, increased severity was observed in 76 Chinese adults undergoing methadone maintenance treatment (Liu et al., Reference Liu, Jin, Zhang, Zhang, Li and Ma2021); 32 children with dyslexia and their mothers (Soriano-Ferrer, Morte-Soriano, Begeny, & Piedra-Martínez, Reference Soriano-Ferrer, Morte-Soriano, Begeny and Piedra-Martínez2021); and 66 adults with psychiatric diagnoses and 22 healthy controls (Seitz et al., Reference Seitz, Bertsch and Herpertz2021). Only one study, of 37 children and 35 adults on the autism spectrum (Lugo-Marín et al., Reference Lugo-Marín, Gisbert-Gustemps, Setien-Ramos, Español-Martín, Ibañez-Jimenez, Forner-Puntonet and Ramos-Quiroga2021), found no change in psychiatric distress.
Results for children and adolescents without psychiatric diagnoses were variable. A study of 1778 Chinese youth who play video games (Teng et al., Reference Teng, Pontes, Nie, Griffiths and Guo2021) found a small increase in distress. However, a study of 127 Canadian youth found a decrease (Gagné, Piché, Clément, & Villatte, Reference Gagné, Piché, Clément and Villatte2021), while a study of 203 Dutch youth found no change (Achterberg, Dobbelaar, Boer, & Crone, Reference Achterberg, Dobbelaar, Boer and Crone2021).
Of six studies in medically vulnerable samples, four found an increase in distress (including both studies in pregnant women; Ayaz et al., Reference Ayaz, Hocaoğlu, Günay, Yardımcı, Turgut and Karateke2020; Perzow et al., Reference Perzow, Hennessey, Hoffman, Grote, Davis and Hankin2021), while the remainder found no change. Small sample size constrains interpretability for these studies, as the largest (177 dialysis patients; Bonenkamp et al., Reference Bonenkamp, Druiventak, van Eck van der Sluijs, van Ittersum, van Jaarsveld and Abrahams2021) found no change in distress, while the next largest (135 pregnant participants; Perzow et al., Reference Perzow, Hennessey, Hoffman, Grote, Davis and Hankin2021) found an increase. The remaining studies had fewer than 70 participants each and were likely underpowered to detect increases of small magnitude. Finally, a longitudinal cohort study of 208 American transgender and gender non-binary individuals found an increase in distress (Kidd et al., Reference Kidd, Jackman, Barucco, Dworkin, Dolezal, Navalta and Bockting2021).
There was a stable tendency for distress to increase in adult samples. Calculable effect sizes (k = 10) tended to be small (median Cohen's d = 0.29; range = −0.24 to 3.8). Undergraduate students and children demonstrated the most consistent exception to this pattern, often showing a decrease in general distress.
General discussion
The present systematic review found that changes in psychopathology from pre- to peri-pandemic varied as a function of symptom cluster and sample characteristics. Contrary to expectations (e.g. Gruber et al., Reference Gruber, Prinstein, Clark, Rottenberg, Abramowitz, Albano and Weinstock2021), adults with pre-existing mental health conditions were not disproportionately affected, excepting adults with pre-existing OCD, whose symptoms tended to worsen. Age also showed unexpected effects. Several large studies (Peters et al., Reference Peters, Rospleszcz, Greiser, Dallavalle and Berger2020; Ramiz et al., Reference Ramiz, Contrand, Rojas Castro, Dupuy, Lu, Sztal-Kutas and Lagarde2021; Winkler et al., Reference Winkler, Mohrova, Mlada, Kuklova, Kagstrom, Mohr and Formanek2021) found more striking increases in anxiety symptoms in children and relatively younger adults, despite those being among the demographics least susceptible to serious COVID-19 infection. Studies of older adults and medically vulnerable individuals tended to have relatively smaller samples and more mixed results, though studies with larger samples found increases in anxiety and fear. Symptom trajectories were similarly variable; OCD and distress-related psychopathology (anxiety; depression) tended to increase, while PTSD and fear-related psychopathology failed to show a consistent pattern.
These patterns are likely multidetermined, but some candidate explanations can be offered. Different trajectories for anxiety (generally increased) v. fear (generally remained stable) may be attributable in part to the time course of the pandemic. Because acute fear reactions unfold on a much shorter timeline compared to anxiety, assessments weeks or months into the pandemic may have captured increased anxiety, but missed an initial uptick in fear. That women in their 20s–40s showed an especially prominent increase in anxiety and depression stands in contrast to their lower medical risk, and aligns with empirical findings that women took on more caregiving work than men when schools and childcare facilities closed during lockdown (OECD, 2021), which may have been especially stressful in light of reduced social support. Samples with pre-existing psychopathology tended to show increases in OCD symptoms, anxiety, and general distress, perhaps due to heightened vulnerability to contamination fears and prolonged uncertainty associated with a viral pandemic. Student samples, individuals with autism, and medical samples were the most likely to demonstrate stable or decreasing symptoms, suggesting that pandemic-related reductions in academic and social demands may have actually reduced overall stress for these populations.
The conclusions of the present review should be interpreted in light of its relative strengths and limitations. The reviewed studies varied in quality, with tradeoffs evident. The largest and most representative studies tended to use briefer assessments, and countries were not evenly represented. Methodological quality coding indicated that most studies used validated measures and reported average to good response rates, but often relied on smaller convenience samples. This suggests that while samples were characterized accurately, they may have failed to include those less likely to participate in voluntary mental health research, such as older individuals or those with existing psychopathology. However, many of the reviewed studies targeted samples with vulnerabilities related to age and psychiatric or medical characteristics, potentially mitigating bias introduced by non-random sampling methods. Although formal meta-analysis or modeling symptom trajectories was not possible due to considerable variability in utilized measures and reporting standards, we placed greater interpretative weight on studies with larger samples and well-validated assessments (and see Table 1, which includes sample nationality and months elapsed between assessments).
Taken together, these findings underscore the importance of specificity in investigating and responding to pandemic-related changes in mental health. Findings are more consistent with theoretical conceptualizations of the pandemic as a chronic stressor, v. an acute trauma. Increased mental health symptoms may have reflected contextual adaptations to a high-risk environment (i.e. ‘true alarms’). Future studies of mental health in disaster contexts should consider the functional context (e.g. potential adaptive value) and impairment associated with symptom changes, while carefully weighing psychometric considerations, such as assuming measurement invariance. Demographic factors such as age, gender, socioeconomic status, and marginalized identity status should also be assessed as possible hidden moderators of disaster impact on mental health. In the present review, although lifespan risk factors such as pregnancy and older age were associated with increased internalizing symptoms, medically vulnerable and most psychiatric populations showed unexpected resilience, which suggests the potential value of a strengths-based perspective.
Anxiety and other forms of internalizing psychopathology have long been conceptualized as evolutionary adaptations that operate in excess in modern, generally safe contexts (Öhman & Mineka, Reference Öhman and Mineka2001). Chronically anxious individuals may have experienced a sense of validation from the societal consensus that the environment was unsafe, or may have been more experienced in navigating day-to-day life while anxious. This resilience was not universal, however. Individuals with contamination-related OCD in particular experienced a worsening of symptoms, perhaps due in part to public health messaging around risks of the virus and responsibility for preventing harm. From a clinical and public health perspective, children and relatively younger adults, and particularly younger women, appear to be shouldering most of the mental health burden. Additional research is needed to identify the major psychological determinants of these vulnerabilities (e.g. caregiving responsibilities; social isolation) to best inform the development of public policy and interventions.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291723002295.
Acknowledgements
The authors thank Paige DeGennaro for assistance with data checking.
Open access
