Abstract
Aim
Evidence indicates most people were resilient to the impact of the COVID-19 pandemic on mental health. However, evidence also suggests the pandemic effect on mental health may be heterogeneous. Therefore, we aimed to identify groups of trajectories of common mental disorders’ (CMD) symptoms assessed before (2017–19) and during the COVID-19 pandemic (2020–2021), and to investigate predictors of trajectories.
Methods
We assessed 2,705 participants of the ELSA-Brasil COVID-19 Mental Health Cohort study who reported Clinical Interview Scheduled-Revised (CIS-R) data in 2017–19 and Depression Anxiety Stress Scale-21 (DASS-21) data in May–July 2020, July–September 2020, October–December 2020, and April–June 2021. We used an equi-percentile approach to link the CIS-R total score in 2017–19 with the DASS-21 total score. Group-based trajectory modeling was used to identify CMD trajectories and adjusted multinomial logistic regression was used to investigate predictors of trajectories.
Results
Six groups of CMD symptoms trajectories were identified: low symptoms (17.6%), low-decreasing symptoms (13.7%), low-increasing symptoms (23.9%), moderate-decreasing symptoms (16.8%), low-increasing symptoms (23.3%), severe-decreasing symptoms (4.7%). The severe-decreasing trajectory was characterized by age < 60 years, female sex, low family income, sedentary behavior, previous mental disorders, and the experience of adverse events in life.
Limitations
Pre-pandemic characteristics were associated with lack of response to assessments. Our occupational cohort sample is not representative.
Conclusion
More than half of the sample presented low levels of CMD symptoms. Predictors of trajectories could be used to detect individuals at-risk for presenting CMD symptoms in the context of global adverse events.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
Background
The COVID-19 pandemic brought concerns about an increase in common mental disorders (CMD) among populations worldwide [1, 2]. Studies conducted in the early stages of the pandemic showed rates of CMD (depression and anxiety disorders) above 30% [3], higher than pooled global estimates from multinational studies conducted in the last decades [4]. However, initial studies performed during the pandemic did not use longitudinal data, therefore, hindering the possibility of detecting change in prevalence of CMD during the COVID-19 pandemic. They also relied mostly on sampling methods (e.g., snowball sampling) prone to selection bias. Later, longitudinal studies comparing pre-pandemic with pandemic data conducted in 2020 showed an increase in mental symptoms in the beginning of the pandemic and a later decrease to similar levels prior to 2020 [5]. Meta-analytic data have shown that the impact of the COVID-19 pandemic and quarantine measures implemented in countries on population mental health was heterogeneous and small in terms of effect size [6], indicating most people are resilient to the changes brought by the pandemic.
Even though a growing body of evidence derived from well-designed longitudinal studies indicates most people are resilient to the changes brought by the pandemic, some specific groups may have been impacted differently. For instance, a meta-analysis investigating the effects of the COVID-19 pandemic on various maternal outcomes reported that depression symptoms, when compared to pre-pandemic levels, were elevated during the pandemic [7]. Recent longitudinal studies have highlighted that the pandemic effect on mental health may be heterogeneous [8, 9]. Understanding different patterns of change in mental health before and during the pandemic and detecting vulnerable groups at risk for developing CMD, as well as risk and protective factors associated with poor trajectories are key to comprehending the impact of the pandemic on populations and to plan interventions.
In this context, cohort studies present the optimal design for leveraging longitudinal data to identify trajectories of CMD symptoms before and during the pandemic. For instance, in a study using data from a United Kingdom sample, three groups of depression symptom trajectories in the early period of the pandemic were identified: low, moderate, and severe [10]. Likewise, a study in Australia identified 4 trajectories of anxiety (resilience, improving, worsening, and sustained) and 3 trajectories of depression (low, moderate, and severe then declining) [11]. The only trajectory study with mental health data before and during the pandemic was conducted in the United Kingdom (UK) [12]. The UK Household Longitudinal Study used the 12-item General Health Questionnaire (GHQ-12) to assess general mental health in 2016–2019 and across six time-points in 2020, from April to October (N = 19,763). The authors reported finding five groups of trajectories: consistently poor, consistently good, consistently very good, deteriorating, and recovery [12]. Among trajectory studies conducted during the pandemic, predictors of poor trajectories included younger age, ethnicity, lower socioeconomic status, living in a city, experiences of physical or psychological abuse, preexisting mental disorder, chronic health problem, social support, empathy, loneliness, financial distress related to the pandemic, functional impairment, and intolerance of uncertainty [10,11,12,13].
The aforementioned CMD trajectory studies were conducted in high-income countries (HIC), albeit more than 80% of the world population lives in low- and middle-income countries (LMIC) [14]. HICs and LMICs are different not only in the epidemiology of mental disorders; but also in several social, economic, and health aspects that impact CMD [15,16,17]. Evaluating CMD trajectories during the pandemic in a LMIC can provide unique insights by assessing predictors, such as low educational level, poverty, and frequent adverse events, which could be generalizable to other LMICs. Such findings could lead to a better understanding of the heterogeneity of the impact of the pandemic on mental health and the development of interventions tailored to the specific needs of LMIC populations worldwide.
In this context, our study the Brazilian Longitudinal Study of Health (ELSA-Brasil) [18, 19] represents one of the few LMIC cohorts that is systematically collecting data on mental health since the pandemic inception, while providing data of several predictors collected in pre-pandemic assessments. Here, our aims were twofold. First, to identify trajectory groups of CMD symptoms assessed at five time-points before and during the COVID-19 pandemic: 2017–19, May to July 2020, July to September 2020, October to December 2020, and April to June 2021. We hypothesized we would find at least three groups of trajectories characterized by low, moderate, and chronic high symptomatology, with at least one group distinguished by a decreasing pattern of symptoms. Our second aim was to investigate the role of sociodemographic characteristics, health-related characteristics, and adverse life events as predictors of trajectories. We hypothesized that age and family income would play a protective role, decreasing the odds of individuals presenting high symptomatology or increasing the odds of a recovery pattern, while previous mental disorders and adverse life events would increase the odds of individuals presenting chronic symptomatology.
Methods
Study design and participants
We are conducting the ELSA-Brasil COVID-19 Mental Health Cohort study, based on the larger ELSA-Brasil occupational cohort that, at baseline (2008–2010), recruited 35–74-year-old employees from six universities (N = 15,105) in major Brazilian cities (São Paulo, Rio de Janeiro, Salvador, Belo Horizonte, Vitoria, and Porto Alegre), with the objective of investigating determinants of mortality and chronic diseases. Since its inception, the ELSA-Brasil conducted three waves of assessment: 2008–2010 (w1), 2012–2014 (w2), and 2017–2019 (w3). Each wave consisted of comprehensive onsite assessments comprising clinical interviews, medical examinations, and laboratory tests, collecting information on sociodemographic variables, clinical history, family history of diseases, lifestyle factors, anthropometric measurements, and biomarkers. Additional details regarding the ELSA-Brasil study can be found elsewhere [18, 20, 21].
In 2020, all participants from the São Paulo research center (active or retired public servants from the University of São Paulo, USP; N = 4,712) were invited to respond to online assessments conducted at four time-points: May to July 2020 (COVID-19 wave 1, c1), July to September 2020 (c2), October to December 2020 (c3), and April to June 2021 (c4). Eligibility criteria included having access to the internet via smartphone, tablet, or personal computer. Research Electronic Data Capture (REDCap) [22] online electronic questionnaires were sent to participants by email or text message. Participants who did not reply to three emails sent at weekly intervals were contacted via three text messages or telephone calls also in a weekly interval. Participants with difficulties understanding, accessing or completing online questionnaires were interviewed via phone call. More details and initial findings from our study can be found in [19, 23].
Our study was approved by the Local Ethics Committee of the University Hospital, USP. All participants provided electronic informed consent. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines were used to report the present study [24].
Assessments and variables
We used two instruments to assess CMDs. In 2017–2019 (w3), CMD was assessed using the Clinical Interview Schedule-Revised (CIS-R) [25], a widely used structured instrument designed to measure non-psychotic mental disorders based on the International Classification of Disease 10th edition (ICD-10) criteria in population studies. The CIS-R has been adapted and validated in Brazil [26] and has been used in multiple studies [27,28,29]. The symptom domains assessed by the CIS-R are somatic complaints, fatigue, concentration and forgetfulness, sleep disturbance, irritability, worry about physical health, depression, depression ideas, worry, anxiety, phobias, panic, compulsions, and obsessions. Each domain has a score ranging from 0 to 4 (except depressive ideas that range from 0 to 5). These scores can be summed up to achieve a CMDs total score ranging from 0 to 57 [25].
During the COVID-19 pandemic, we used the Depression, Anxiety, and Stress Scale (DASS-21) [30] to assess CMD. The DASS-21 is a self-report instrument with a Likert four point scale (from 0 “strongly disagree'' to 3 “totally agree”) indicating the frequency or severity of symptoms. Each of its three domains are composed of 7 items. A total score summing items from the three domains was calculated, with higher scores indicating greater severity (ranging from 0 to 63). The DASS-21 was translated and validated to Brazilian Portuguese [31] and has been used extensively in Brazil [32, 33].
We identified potential predictors of trajectories in three domains: sociodemographic characteristics (age, sex, skin color, income, marital status), health-related characteristics (physical activity, obesity, active smoking, alcohol abuse, mental disorders), and experience of adverse life events (e.g., race discrimination, childhood adverse events). This selection was based on our previous works [19, 27, 34], as well as on reviews of mental symptoms’ trajectories studies during the COVID-19 pandemic [10, 11, 13, 35]. Predictors were derived from data collected in 2017–2019 and 2008–2010. Table 1 provides a description of all predictors. For additional details on how each variable was assessed and coded, see online supplementary material.
Analysis
Analyses were conducted using Stata 17 and R. Statistical significance was set under an alpha threshold of 0.005 [36,37,38]. Parameters are reported using a confidence interval of 99.5% (99.5% CI).
Since CMD was measured before the pandemic using the CIS-R and during the pandemic using the DASS-21, we used equi-percentile linking (R package equate) [39, 40], a statistical procedure designed to identify comparable scores on different scales measuring the same construct. This procedure is adequate when item endorsement may have a nonlinear relationship between scales [41]. First, we calculated a modified version of the CIS-R total score, excluding symptoms of compulsion, obsession, sleep disturbance, and phobias, leading to a range of 0 to 41, since there are no equivalent items in the DASS-21. Then, the modified CIS-R total score was linked to the DASS-21 total score at c1, generating a CIS-R total score equivalent to the DASS-21 total score with a range from 0 to 63.
We used group-based trajectory modeling (GBTM) [42, 43] to identify CMD symptoms trajectories over five time-points before and during the COVID-19 pandemic (w3, c1 to c4). GBTM is a finite mixture method designed to identify clusters of individuals with similar trajectories of an outcome of interest assessed over time. Selection of the best GBTM model was based on the following criteria: (1) Bayesian Information Criteria (BIC), (2) average posterior probabilities (APP), (3) domain knowledge (previous studies), and (4) interpretability. The largest BIC score (less negative) and APP above 0.7 were considered indicators of a good model fit [43]. The shape of trajectories was modeled using polynomial functions. The number of measurements minus 1 (time-1) defined the polynomial functions used [44]. We tested whether each CMD symptoms trajectory presented a constant, linear, quadratic, cubic, or quartic relationship. The p-values of trajectory shape parameters across different models were used to determine the best combination and order of polynomial functions. Different combinations of number of groups and trajectory shapes were tested.
Since the distribution of DASS-21 scores presented zero-inflation (Fig. S1), all tested models used zero-inflation Poisson (ZIP) as the probability distribution of the dependent variable (CMD symptoms). ZIP models presented better BIC scores when compared to models with Gaussian distribution. All participants with at least one DASS-21 report were included in the models. The selected model was plotted with parametric bootstrap confidence intervals.
Considering some participants had missing data on predictor variables (see Table S1 for distribution of missing values), we used multiple imputation by chained equations (MICE) [45,46,47]. MICE takes into account uncertainty in the imputation procedures, while the chained equations strategy can deal with different types of variables, making predictions more accurate. To determine the number of imputations, we ran a model including all variables of interest with five imputations to determine the fraction of missing information (FMI). The FMI indicates the proportion of the total sampling variance related to missing data. The largest FMI found was 0.54, indicating the necessity of at least 40 imputations to prevent loss of statistical power [48]. Our outcome of interest and all predictors were inserted in the MICE models, generating 40 imputations. Data were assumed to be missing at random [49]. We used multinomial logistic regression to investigate predictors of trajectories. Separate models were run for each predictor, adjusting for sex at birth, age, educational level, and skin color. The dependent variable was the identified group trajectories. We reported results of the pooled log odds ratio (lOR) considering imputations. lOR estimates were plotted by trajectory groups.
Results
Participants
Out of 4712 eligible participants, we obtained CMD data from 2,705 (57.4%). Non-inclusion reasons were unwillingness to participate, impossibility of making contact, and deaths (Fig. 1). We compared participants who provided CMD data in at least one time-point (included) vs. participants who did not provide CMD data (not included). The included sample had a significantly higher percentage of women (included = 57.2% vs. not included = 52.2%), younger participants (≥ 60 year: included = 40.3% vs. not included = 55.3%), higher educational level (college degree: included = 58.8% vs. not included = 28.0%), and presented higher score of CMD (mean CIS-R total score difference = 0.9) (Table S2). The mean age of the included sample was 58.3 (99.5% CI [58.0, 58.6]) years, while 57.2% were female (99.5% CI [55.3%, 59.0%]) and 33.5% were non-white (99.5% CI [31.7%, 35.3%]) (Table S3). The mean CMD total score at w3 was 8.0 (7.9); c1 was 8.1 (9.4); at c2 was 7.2 (8.7); at c3 was 6.9 (8.4); at c4 was 7.6 (9.1) (Table S3).
Trajectories of common mental disorders symptoms
Among simple models specifying linear shapes of trajectories, we selected the six-group model. Additional groups improved BICs slightly, but due to interpretability and domain knowledge (no previous study identified more than six groups), we chose the six-group model. Thus, we explored different combinations of polynomial functions among six-group models. The best fit presented a BIC of -30,973.36 with all groups APP above 0.8. Table S4 depicts the goodness of fit of tested models, while table S5 presents parameter estimates of the best fitting model.
The following groups (Fig. 2) were identified and labeled according to levels of CMD symptoms and shape of trajectories: Group 1 (17.6%) low symptoms; Group 2 (13.7%) low-decreasing symptoms; Group 3 (23.9%) low-increasing symptoms; Group 4 (16.8%) moderate-decreasing symptoms; Group 5 (23.3%) low-increasing symptoms; Group 6 (4.7%) severe-decreasing symptoms. Table S6 depicts group means of CMD by time-point.
Predictors of trajectories
Table S7 describes the distribution of predictors by CMD symptoms trajectory groups. Figure 3 and Table S8 depict results of the predictors analyses by trajectory group (multinomial logistic regressions). Below, we describe significant results from models. All 5 groups, when compared to Group 1, were negatively associated with female sex and previous mental disorders. Being older predicted decreased risk of membership in Groups 3, 5, and 6, while increased the risk of membership in Group 4. Participants who reported being non-white were at a decreased risk of membership in Groups 2 and 5. Having a college degree increased the risk of membership in Groups 2 and 5. Higher family income decreased the risk of membership in Group 6. Participants physically active were more likely to have severe-decreasing symptoms (Group 6). Having experienced at least one childhood adverse event increased the risk of membership in Groups 4, 5, and 6. Participants that reported adverse life events were more likely to present moderate-decreasing symptoms (Group 4), low-increasing symptoms (Group 5), and severe-decreasing symptoms (Group 6) trajectories. The following predictors were unrelated to membership in any of the Groups: being married, obesity, active smoking, alcohol abuse, and race discrimination.
Discussion
We sought to identify CMD symptom trajectories before (2017–2018) and during the COVID-19 pandemic (from May 2020 to June 2021) in a large Brazilian cohort. Our findings showed that 55% of the sample presented low levels of CMD symptoms (55%, Groups 1 to 3), while 4.7% of the sample was classified in the most severe trajectory. This finding is similar to previous pre-pandemic studies that investigated trajectories of depression and anxiety, showing the majority of the population do not develop high levels of symptoms [50, 51]. It is also in line with studies that used latent trajectories analysis during the pandemic [10,11,12,13].
In line with our hypothesis, we found groups showing decreasing levels of symptoms over time. The severe-decreasing group (Group 6, 4.7% of the sample) presented a clinically significant decline in CMD symptoms, from a mean of 35.4 to 22.8, resulting in a 35.6% change. Previous studies before the pandemic showed that trajectory patterns of decreasing symptoms over time are common [50]. In the context of the COVID-19 pandemic, a cohort study in Ireland found one trajectory group of a composite measure of depression, anxiety, and post-traumatic stress symptoms that improved over time (18%) [13]. In Australia, a cohort study assessed depression and anxiety symptoms online across seven time-points in 2020, finding a group with severe levels of depression symptoms declining over time (9%) and a group with moderate decreasing levels of anxiety (5%) [11]. The UK Household Longitudinal Study, a study with pre-pandemic and pandemic depression data also identified a group with decreasing symptoms (12%) [12]. One notable difference from our findings is that the decline in symptoms occurred in 2021. Before this time-point, the severe-decreasing group CMD means fluctuated around 31 to 35. However, our findings may not be directly comparable to the aforementioned studies, since our measure is a composite score of depression and anxiety disorders, while most studies assessed only depression.
Our data showed CMD trajectories were mostly stable across time. This could be because our sample was composed of older adults (mean age of 58.3 years). Trajectory studies of depression symptom trajectories in adolescence and early adulthood show less stability, with models revealing increasing and decreasing trajectories [52], compared to studies with older adults [50]. The stability of symptoms in our sample, especially during 2020, could also be explained by our time-point intervals, since assessments in 2020 occurred a few months apart. However, other studies had similar intervals but presented less stability [8, 12].
Furthermore, different from other studies [8, 12], we did not find a group characterized by severe-chronic symptoms. Diminishing rates of new cases and mortality by COVID-19 in Brazil, as well as the start of the vaccination campaign, may have buffered the negative impact of the pandemic on mental health, therefore explaining the decreasing pattern in the most severe group. Also, in the early stages of the pandemic, mental health awareness increased in Brazil due to campaigns and various institutions started offering free remote mental health care for the population [53]; thus, participants may have benefited from these measures by seeking and receiving treatment during the pandemic. Another explanation for this finding could be related to participants being retired or active public servants who had no changes to their income during the pandemic, while also having access to mental health care provided by the university.
Notably, apart from the severe-decreasing group, CMD symptom levels were below 20 points. Participants in this severity range were probably not facing impairment typical of full criteria depression and anxiety disorders. The moderate-decreasing group may have had subclinical CMD with some impairment, while the low-increasing group had a slight deterioration that could potentially lead to subclinical CMD. Even though these two groups have trajectories of low and moderate levels of symptoms, they may benefit from mental health care attention to prevent further deterioration. Considering a stepped care approach, individuals with low but increasing levels of symptoms could benefit from low intensity interventions, such as self-guided psychotherapy and psychoeducation [54]. Cases with further deterioration would then benefit directly from more intensive interventions (e.g., cognitive-behavioral therapy, pharmacological intervention, etc.). Future assessments of our sample will be needed to further understand post-pandemic trajectories.
Our second aim was to investigate predictors of CMD trajectories. Our findings show the most severe trajectory of CMD symptoms was characterized by age < 60 years, female sex, low family income, non-sedentary behavior, previous mental disorders, and the experience of adverse events in life. Pre-pandemic studies have shown older age is associated with a significant decline in the proportion of individuals with more severe trajectories. A systematic review of depression trajectories showed that 2–7% of older adults are classified in the chronic high levels of symptoms trajectory, compared to 14–32% in adolescents, and 2 to 28% in adults [50]. Previous trajectories studies conducted during the pandemic also found women are more likely to present depression and anxiety symptoms [11, 12]. A systematic review of studies conducted in the early stages of the pandemic found a consistent association between female sex and risk for mental disorders in general [35]. This is not surprising, since this is a well-known risk factor for CMD [55, 56]. It is worth noting a previous study analyzing our sample in the early stage of the pandemic showed increased age and male sex were protective factors for CMD [19].
Higher family income was a protective factor against the severe-decreasing trajectory of CMD symptoms. Socioeconomic status (SES) is a well-known factor associated with mental disorders. Robust evidence shows income has a dose–response relationship with depression [57]. Even though the mechanism behind this association is not yet fully understood, evidence suggests that low SES is commonly associated with exposure to several negative conditions, such as unhealthy work conditions, adverse living conditions, material deprivation, daily hassles, lack of cultural or leisure activities, among others [58, 59]. The persistent exposure to these conditions may be related to the onset and maintenance of CMD symptoms. Studies investigating the impact of the COVID-19 pandemic also revealed that income is negatively associated with mental health outcomes [35, 60].
Physically active participants presented a lower risk of pertaining to the most severe trajectory of CMD. Evidence shows there is a link between depression and sedentary behavior [61, 62]. People with depression on average spend less time doing physical activities and this pattern is stronger in older adults [63]. Even though the relationship between these variables is not totally understood, recent evidence suggests there is a causal link between being physically active and reduced risk for depression [64]. Interestingly, physical activity is a modifiable risk factor that can be the target of effective interventions to reduce depression symptoms [65]. Such interventions could be adapted and implemented in contexts similar to the pandemic.
In addition, preexisting mental disorders (measured in 2017–2019) were also a predictor of the more severe trajectories of mental symptoms, drawing attention to the need of offering or maintaining treatment for these individuals during the pandemic, as previously demonstrated by other studies [19, 66]. Most importantly, the group of individuals with mental health problems is at-risk for developing persistent CMD symptoms in the context of a global pandemic and should be prioritized to receive interventions.
Participants who experienced adverse life events in childhood and adulthood were at a higher risk of presenting the most severe trajectory of CMD symptoms. Adverse or stressful life events are of the most studied risk factors linked to the onset of CMDs. Multiple adverse events early in life (physical abuse, neglect, family conflict, household criminality, etc.) are negatively associated with anxiety and depression [67, 68]. These early adversities may alter physiological processes that can lead to impairments in the development of brain functions that are later in life associated with negative outcomes [69, 70]. In addition, adversities later in life may also negatively impact mental health [71, 72]. Meta-analytic evidence shows negative life events are also linked to depression in older people, such as in the case of our sample [73]. Screening of experiences of adverse life events could be used to detect individuals at risk for severe trajectories of CMD symptoms to plan and deliver interventions.
Interestingly, our data suggest not only that the impact of the pandemic on CMD trajectories is heterogeneous, but it is also characterized by different pre-pandemic predictors. For instance, the moderate-decreasing group was characterized by increased age, female sex, previous mental disorders, and the experience of adverse events in life. In this case, no predictor showed a protective effect. Also, predictors impacted specific groups differently. Older age, for example, acted as a protective factor against the severe-decreasing trajectory, but increased the odds of the moderate-decreasing trajectory. Noticeably, preexisting mental disorders presented a consistent pattern, increasing the risk of all trajectory groups, showing it may be a predictor not ideal to differentiate types of trajectories of CMD symptoms.
Our findings should be viewed in light of some limitations. First, we found that sex, age, educational level, and mental health were characteristics associated with not responding to assessments, although these differences were small to moderate. In the case of the CIS-R total score the differences were less than one unit. Additionally, approximately half of the eligible sample responded to assessments during the pandemic. Even though this is in line with other cohort studies [12, 66, 74], those who did not respond may be more likely to present a deteriorating trajectory of CMD symptoms or have more pandemic-related problems (e.g., financial issues, stress, etc.). Second, our findings may not be generalizable to other contexts, since our sample is not representative of the country's population. Our sample is composed of public servants of the University of Sao Paulo. Compared to national indicators in 2019, our sample has significant differences in the same year. While 58.8% of our sample had a college degree, only 17.4% of Brazilians had a college degree. In our sample, 57.3% reported non-white skin color, while in Brazil 33.5% of the population was non-white. Moreover, 51.8% of the Brazil population was female, while in our study 57.2% was female [75]. Also, the income of participants (working or retired) may have been mostly unaffected by the pandemic. Therefore, caution is needed when interpreting and generalizing our findings. However, our study was based on a well-defined 13-year large cohort of participants systematically assessed by trained professionals before the COVID-19 pandemic. We consider these aspects a strength compared to online surveys without pre-pandemic data and studies using snowball sampling. Also, our findings are relevant because they add to the growing body of evidence from developing countries, especially those that struggled to contain the COVID-19 pandemic, as it was the case of Brazil. Third, adverse life events and childhood adverse events were not derived from validated questionnaires, yet these were standardized questionnaires based on well-known measures. In addition, childhood adverse events were measured retrospectively when participants were adults, therefore this measure was prone to memory bias. Fourth, CMD symptoms before the pandemic were measured using a different scale when compared to pandemic data. To account for this discrepancy we used equipercentile linking, deriving a modified CIS-R total score equivalent to the DASS-21 total score. Our findings may have been different if the same measure was used across all time-points. However, equipercentile linking methods are known to be robust and have previously been effectively used in the field of psychiatry [76,77,78]. Fifth, our analytical approach combined depression and anxiety symptoms in a single domain. However, these symptoms may have distinct trajectories. Future studies could investigate these symptoms separately using a multi-trajectory strategy [79]. Sixth, our analysis did not include variables measured during the pandemic, such as COVID infection and economic impact. Previous studies showed CMD symptoms were greatly influenced by such variables [12, 19]. However, to establish a temporal link between CMD trajectories and predictors, we opted to include only variables measured before the pandemic. Seventh, we did not measure trauma symptoms during the pandemic. The pandemic was a major life-changing event that could be a potential trauma affecting general mental health or even causing post-traumatic stress disorder, as previous studies indicated [80, 81]. Eighth, we identified groups of trajectories using statistical models. Even though we followed guidelines and objective metrics to identify the best fitting model, we cannot be certain the same model would be identified in a different sample or context. Also, overall, models with more groups revealed larger BIC scores, although above six groups the increase was slight. So we based our decision on interpretability (more groups would be a challenge for the predictors analysis) and domain knowledge (no previous study identified more than six groups). Additionally, more groups can lead to smaller groups, resulting in model misspecification and problems to run multinomial regression models.
Conclusion
We longitudinally assessed symptoms of CMD in adults at five time points before and during the pandemic in São Paulo, Brazil. Six trajectories of CMD symptoms emerged from our data. More than half of the sample presented low levels of mental symptoms, meaning resilience to the challenges brought by the COVID-19 pandemic is a prevalent feature of our sample. However, approximately 5% of the sample presented severe symptomatology that attenuated over time. The severe-decreasing levels of CMD symptoms was characterized by increased age, female sex, low family income, previous mental disorders, and the experience of adverse events in life. These pre-pandemic risk and protective factors could be used to detect individuals at risk for presenting clinically significant CMD symptoms in the context of global adverse events such as the COVID-19 pandemic.
References
de Mari JJ, MOquendo MA (2020) Mental health consequences of COVID-19: the next global pandemic. Trends Psychiatry Psychother 42:219–220. https://doi.org/10.1590/2237-6089-2020-0081
Reger MA, Stanley IH, Joiner TE (2020) Suicide mortality and coronavirus disease 2019-a perfect storm? JAMA Psychiat 77:1093–1094. https://doi.org/10.1001/jamapsychiatry.2020.1060
Wu T, Jia X, Shi H et al (2021) Prevalence of mental health problems during the COVID-19 pandemic: a systematic review and meta-analysis. J Affect Disord 281:91–98. https://doi.org/10.1016/j.jad.2020.11.117
Steel Z, Marnane C, Iranpour C et al (2014) The global prevalence of common mental disorders: a systematic review and meta-analysis 1980–2013. Int J Epidemiol 43:476–493. https://doi.org/10.1093/ije/dyu038
Robinson E, Sutin AR, Daly M, Jones A (2022) A systematic review and meta-analysis of longitudinal cohort studies comparing mental health before versus during the COVID-19 pandemic in 2020. J Affect Disord 296(January):567–576. https://doi.org/10.1016/j.jad.2021.09.098
Prati G, Mancini AD (2021) The psychological impact of COVID-19 pandemic lockdowns: a review and meta-analysis of longitudinal studies and natural experiments. Psychol Med 51:201–211. https://doi.org/10.1017/S0033291721000015
Chmielewska B, Barratt I, Townsend R et al (2021) Effects of the COVID-19 pandemic on maternal and perinatal outcomes: a systematic review and meta-analysis. Lancet Glob Health 9:e759–e772. https://doi.org/10.1016/S2214-109X(21)00079-6
Shevlin M, Butter S, McBride O et al (2021) Refuting the myth of a “tsunami” of mental ill-health in populations affected by COVID-19: evidence that response to the pandemic is heterogeneous, not homogeneous. Psychol Med. https://doi.org/10.1017/S0033291721001665
Ahrens KF, Neumann RJ, Kollmann B et al (2021) Differential impact of COVID-related lockdown on mental health in Germany. World Psychiatry 20:140–141. https://doi.org/10.1002/wps.20830
Iob E, Frank P, Steptoe A, Fancourt D (2020) Levels of severity of depressive symptoms among at-risk groups in the UK during the COVID-19 pandemic. JAMA Netw Open 3:e2026064. https://doi.org/10.1001/jamanetworkopen.2020.26064
Batterham PJ, Calear AL, McCallum SM et al (2021) Trajectories of depression and anxiety symptoms during the COVID-19 pandemic in a representative Australian adult cohort. Med J Aust. https://doi.org/10.5694/mja2.51043
Pierce M, McManus S, Hope H et al (2021) Mental health responses to the COVID-19 pandemic: a latent class trajectory analysis using longitudinal UK data. Lancet Psychiatry. https://doi.org/10.1016/S2215-0366(21)00151-6
Hyland P, Vallières F, Daly M et al (2021) Trajectories of change in internalizing symptoms during the COVID-19 pandemic: a longitudinal population-based study. J Affect Disord 295:1024–1031. https://doi.org/10.1016/j.jad.2021.08.145
Worldbank (2019) Low and middle income population. In: The Worldbank. https://data.worldbank.org/country/XO. Accessed 15 Jun 2021
Demyttenaere K, Bruffaerts R, Posada-Villa J et al (2004) Prevalence, severity, and unmet need for treatment of mental disorders in the World Health Organization World Mental Health Surveys. JAMA 291:2581–2590. https://doi.org/10.1001/jama.291.21.2581
Bromet E, Andrade LH, Hwang I et al (2011) Cross-national epidemiology of DSM-IV major depressive episode. BMC Med 9:90. https://doi.org/10.1186/1741-7015-9-90
Lim GY, Tam WW, Lu Y et al (2018) Prevalence of depression in the community from 30 countries between 1994 and 2014. Sci Rep 8:2861. https://doi.org/10.1038/s41598-018-21243-x
Schmidt MI, Duncan BB, Mill JG et al (2015) Cohort profile: longitudinal study of adult health (ELSA-Brasil). Int J Epidemiol 44:68–75. https://doi.org/10.1093/ije/dyu027
Brunoni AR, Suen PJC, Bacchi PS et al (2021) Prevalence and risk factors of psychiatric symptoms and diagnoses before and during the COVID-19 pandemic: findings from the ELSA-Brasil COVID-19 Mental Health Cohort. Psychol Med. https://doi.org/10.1017/S0033291721001719
Aquino EML, Barreto SM, Bensenor IM et al (2012) Brazilian Longitudinal Study of Adult Health (ELSA-Brasil): objectives and design. Am J Epidemiol 175:315–324. https://doi.org/10.1093/aje/kwr294
Nunes MA, Pinheiro AP, Bessel M et al (2016) Common mental disorders and sociodemographic characteristics: baseline findings of the Brazilian Longitudinal Study of Adult Health (ELSA-Brasil). Braz J Psychiatry 38:91–97. https://doi.org/10.1590/1516-4446-2015-1714
Harris PA, Taylor R, Thielke R et al (2009) Research electronic data capture (REDCap): a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform 42:377–381. https://doi.org/10.1016/j.jbi.2008.08.010
Suen PJC, Bacchi PS, Razza L et al (2022) Examining the impact of the COVID-19 pandemic through the lens of the network approach to psychopathology: analysis of the Brazilian Longitudinal Study of Health (ELSA-Brasil) cohort over a 12-year timespan. J Anxiety Disord 85:102512. https://doi.org/10.1016/j.janxdis.2021.102512
von Elm E, Altman DG, Egger M et al (2007) The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Ann Intern Med 147:573–577. https://doi.org/10.7326/0003-4819-147-8-200710160-00010
Lewis G, Pelosi AJ, Araya R, Dunn G (1992) Measuring psychiatric disorder in the community: a standardized assessment for use by lay interviewers. Psychol Med 22:465–486. https://doi.org/10.1017/s0033291700030415
Nunes MA, de Mello Alves MG, Chor D et al (2011) Adaptação transcultural do CIS-R (Clinical Interview Schedule - Revised Version) para o português no estudo longitudinal de saúde do adulto (ELSA). Clin Biomed Res 31(4):487–490
Brunoni AR, Santos IS, Passos IC et al (2020) Socio-demographic and psychiatric risk factors in incident and persistent depression: an analysis in the occupational cohort of ELSA-Brasil. J Affect Disord 263:252–257. https://doi.org/10.1016/j.jad.2019.11.155
Brunoni AR, Szlejf C, Suemoto CK et al (2019) Association between ideal cardiovascular health and depression incidence: a longitudinal analysis of ELSA-Brasil. Acta Psychiatr Scand 140:552–562. https://doi.org/10.1111/acps.13109
Busatto GF, de Araújo AL, da Duarte AJ, S, et al (2021) Post-acute sequelae of SARS-CoV-2 infection (PASC): a protocol for a multidisciplinary prospective observational evaluation of a cohort of patients surviving hospitalisation in Sao Paulo, Brazil. BMJ Open 11:e051706. https://doi.org/10.1136/bmjopen-2021-051706
Antony MM, Bieling PJ, Cox BJ et al (1998) Psychometric properties of the 42-item and 21-item versions of the Depression Anxiety Stress Scales in clinical groups and a community sample. Psychol Assess 10:176–181. https://doi.org/10.1037/1040-3590.10.2.176
Vignola RCB, Tucci AM (2014) Adaptation and validation of the depression, anxiety and stress scale (DASS) to Brazilian Portuguese. J Affect Disord 155:104–109. https://doi.org/10.1016/j.jad.2013.10.031
Pasqualucci PL, Damaso LLM, Danila AH et al (2019) Prevalence and correlates of depression, anxiety, and stress in medical residents of a Brazilian Academic Health System. BMC Med Educ 19:193. https://doi.org/10.1186/s12909-019-1621-z
Serafim AP, Durães RSS, Rocca CCA et al (2021) Exploratory study on the psychological impact of COVID-19 on the general Brazilian population. PLoS ONE 16:e0245868. https://doi.org/10.1371/journal.pone.0245868
Librenza-Garcia D, Passos IC, Feiten JG et al (2020) Prediction of depression cases, incidence, and chronicity in a large occupational cohort using machine learning techniques: an analysis of the ELSA-Brasil study. Psychol Med. https://doi.org/10.1017/S0033291720001579
Xiong J, Lipsitz O, Nasri F et al (2020) Impact of COVID-19 pandemic on mental health in the general population: a systematic review. J Affect Disord 277:55–64. https://doi.org/10.1016/j.jad.2020.08.001
Benjamin DJ, Berger JO, Johannesson M et al (2018) Redefine statistical significance. Nat Hum Behav 2:6–10. https://doi.org/10.1038/s41562-017-0189-z
Johnson VE (2013) Revised standards for statistical evidence. Proc Natl Acad Sci USA 110:19313–19317. https://doi.org/10.1073/pnas.1313476110
Ioannidis JPA (2018) The proposal to lower P value thresholds to.005. JAMA 319:1429–1430. https://doi.org/10.1001/jama.2018.1536
Albano AD (2015) A general linear method for equating with small samples. J Educ Meas 52:55–69. https://doi.org/10.1111/jedm.12062
Albano A (2018) Equate: observed-score linking and equating. https://cran.r-project.org/web/packages/equate/index.html. Accessed 24 Aug 2021
Kolen MJ, Brennan RL (2014) Test equating, scaling, and linking: methods and practices (statistics for social and behavioral sciences), 3rd edn. Springer, Berlin
Nagin DS, Odgers CL (2010) Group-based trajectory modeling in clinical research. Annu Rev Clin Psychol 6:109–138. https://doi.org/10.1146/annurev.clinpsy.121208.131413
Nagin DS (2005) Group-based modeling of development. Harvard University Press, Cambridge
Twisk JWR (2013) Continuous outcome variables. Applied longitudinal data analysis for epidemiology: a practical guide, 2nd edn. Cambridge University Press, Cambridge
Azur MJ, Stuart EA, Frangakis C, Leaf PJ (2011) Multiple imputation by chained equations: what is it and how does it work? Int J Methods Psychiatr Res 20:40–49. https://doi.org/10.1002/mpr.329
Kenward MG, Carpenter J (2007) Multiple imputation: current perspectives. Stat Methods Med Res 16:199–218. https://doi.org/10.1177/0962280206075304
Rubin DB (1987) Multiple imputation for nonresponse in surveys. Wiley, New Jersey
Graham JW, Olchowski AE, Gilreath TD (2007) How many imputations are really needed? Some practical clarifications of multiple imputation theory. Prev Sci 8:206–213. https://doi.org/10.1007/s11121-007-0070-9
Jakobsen JC, Gluud C, Wetterslev J, Winkel P (2017) When and how should multiple imputation be used for handling missing data in randomised clinical trials-a practical guide with flowcharts. BMC Med Res Methodol 17:162. https://doi.org/10.1186/s12874-017-0442-1
Musliner KL, Munk-Olsen T, Eaton WW, Zandi PP (2016) Heterogeneity in long-term trajectories of depressive symptoms: patterns, predictors and outcomes. J Affect Disord 192:199–211. https://doi.org/10.1016/j.jad.2015.12.030
de la Torre-Luque A, de la Fuente J, Prina M et al (2019) Long-term trajectories of depressive symptoms in old age: relationships with sociodemographic and health-related factors. J Affect Disord 246:329–337. https://doi.org/10.1016/j.jad.2018.12.122
Schubert KO, Clark SR, Van LK et al (2017) Depressive symptom trajectories in late adolescence and early adulthood: a systematic review. Aust N Z J Psychiatry 51:477–499. https://doi.org/10.1177/0004867417700274
Fukuti P, Uchôa CLM, Mazzoco MF et al (2020) How institutions can protect the mental health and psychosocial well-being of their healthcare workers in the current COVID-19 pandemic. Clinics 75:e1963. https://doi.org/10.6061/clinics/2020/e1963
Bower P, Kontopantelis E, Sutton A et al (2013) Influence of initial severity of depression on effectiveness of low intensity interventions: meta-analysis of individual patient data. BMJ 346:f540. https://doi.org/10.1136/bmj.f540
Otte C, Gold SM, Penninx BW et al (2016) Major depressive disorder. Nat Rev Dis Primers 2:16065. https://doi.org/10.1038/nrdp.2016.65
Craske MG, Stein MB, Eley TC et al (2017) Anxiety disorders. Nat Rev Dis Primers 3:17024. https://doi.org/10.1038/nrdp.2017.24
Lorant V, Deliège D, Eaton W et al (2003) Socioeconomic inequalities in depression: a meta-analysis. Am J Epidemiol 157:98–112. https://doi.org/10.1093/aje/kwf182
Gallo LC (2009) The reserve capacity model as a framework for understanding psychosocial factors in health disparities. Appl Psychol Health Well Being 1:62–72. https://doi.org/10.1111/j.1758-0854.2008.01000.x
Niemeyer H, Bieda A, Michalak J et al (2019) Education and mental health: do psychosocial resources matter? SSM Popul Health 7:100392. https://doi.org/10.1016/j.ssmph.2019.100392
O’Connor RC, Wetherall K, Cleare S et al (2020) Mental health and well-being during the COVID-19 pandemic: longitudinal analyses of adults in the UK COVID-19 Mental Health & Wellbeing study. Br J Psychiatry. https://doi.org/10.1192/bjp.2020.212
Huang Y, Li L, Gan Y et al (2020) Sedentary behaviors and risk of depression: a meta-analysis of prospective studies. Transl Psychiatry 10:26. https://doi.org/10.1038/s41398-020-0715-z
Schuch FB, Vancampfort D, Firth J et al (2018) Physical activity and incident depression: a meta-analysis of prospective cohort studies. Am J Psychiatry. https://doi.org/10.1176/appi.ajp.2018.17111194
Stubbs B, Vancampfort D, Firth J et al (2018) Relationship between sedentary behavior and depression: A mediation analysis of influential factors across the lifespan among 42,469 people in low- and middle-income countries. J Affect Disord 229:231–238. https://doi.org/10.1016/j.jad.2017.12.104
Choi KW, Chen C-Y, Stein MB et al (2019) Assessment of bidirectional relationships between physical activity and depression among Adults: a 2-sample mendelian randomization study. JAMA Psychiat. https://doi.org/10.1001/jamapsychiatry.2018.4175
Rosenbaum S, Tiedemann A, Sherrington C et al (2014) Physical activity interventions for people with mental illness: a systematic review and meta-analysis. J Clin Psychiatry 75:964–974. https://doi.org/10.4088/JCP.13r08765
Pan K-Y, Kok AAL, Eikelenboom M et al (2020) The mental health impact of the COVID-19 pandemic on people with and without depressive, anxiety, or obsessive-compulsive disorders: a longitudinal study of three Dutch case-control cohorts. Lancet Psychiatry. https://doi.org/10.1016/S2215-0366(20)30491-0
Hughes K, Bellis MA, Hardcastle KA et al (2017) The effect of multiple adverse childhood experiences on health: a systematic review and meta-analysis. Lancet Public Health 2:e356–e366. https://doi.org/10.1016/S2468-2667(17)30118-4
Petruccelli K, Davis J, Berman T (2019) Adverse childhood experiences and associated health outcomes: a systematic review and meta-analysis. Child Abuse Negl 97:104127. https://doi.org/10.1016/j.chiabu.2019.104127
Golm D, Maughan B, Barker ED et al (2020) Why does early childhood deprivation increase the risk for depression and anxiety in adulthood? A developmental cascade model. J Child Psychol Psychiatry 6:63. https://doi.org/10.1111/jcpp.13205
Shonkoff JP, Garner AS, Committee on Psychosocial Aspects of Child and Family Health et al (2012) The lifelong effects of early childhood adversity and toxic stress. Pediatrics 129:e232–e246. https://doi.org/10.1542/peds.2011-2663
Kendler KS, Karkowski LM, Prescott CA (1999) Causal relationship between stressful life events and the onset of major depression. Am J Psychiatry 156:837–841. https://doi.org/10.1176/ajp.156.6.837
Kessler RC (1997) The effects of stressful life events on depression. Annu Rev Psychol 48:191–214. https://doi.org/10.1146/annurev.psych.48.1.191
Kraaij V, Arensman E, Spinhoven P (2002) Negative life events and depression in elderly persons: a meta-analysis. J Gerontol B 57:P87-94. https://doi.org/10.1093/geronb/57.1.p87
Pierce M, Hope H, Ford T et al (2020) Mental health before and during the COVID-19 pandemic: a longitudinal probability sample survey of the UK population. Lancet Psychiatry 7:883–892. https://doi.org/10.1016/S2215-0366(20)30308-4
IBGE (2019) Pesquisa Nacional por Amostra de Domicílios (PNAD). IBGE, Rio de Janeiro
McCabe-Beane JE, Segre LS, Perkhounkova Y et al (2016) The identification of severity ranges for the Edinburgh Postnatal Depression Scale. J Reprod Infant Psychol 34:293–303. https://doi.org/10.1080/02646838.2016.1141346
Samara MT, Engel RR, Millier A et al (2014) Equipercentile linking of scales measuring functioning and symptoms: Examining the GAF, SOFAS, CGI-S, and PANSS. Eur Neuropsychopharmacol 24:1767–1772. https://doi.org/10.1016/j.euroneuro.2014.08.009
Stevens AL, Ho KY, Mason WA, Chmelka MB (2021) Using equipercentile equating to link scores of the CBCL and SDQ in residential youth. Resid Treat Child Youth 38:102–113. https://doi.org/10.1080/0886571X.2019.1704670
Nagin DS, Jones BL, Lima Passos V, Tremblay RE (2016) Group-based multi-trajectory modeling. Stat Methods Med Res. https://doi.org/10.1177/0962280216673085
Tmgh-Global Covid-Collaborative (2021) Psychological impacts and post-traumatic stress disorder among people under COVID-19 quarantine and isolation: a global survey. Int J Environ Res Public Health. https://doi.org/10.3390/ijerph18115719
Abdalla SM, Ettman CK, Cohen GH, Galea S (2021) Mental health consequences of COVID-19: a nationally representative cross-sectional study of pandemic-related stressors and anxiety disorders in the USA. BMJ Open 11:e044125. https://doi.org/10.1136/bmjopen-2020-044125
Acknowledgements
We thank the ELSA-Brasil staff for administrative support.
Funding
This study was supported by a São Paulo Research StateFoundation (FAPESP) grant (20/05441-9). DF receives support from the International Health Cohort Consortium (IHCC) and Open Society Foundations. ARB receives scholarships and support from FAPESP, the Brazilian National Council of Scientific Development (CNPq-1B), University of São Paulo Medical School (FMUSP), the UK Academy of Medical Sciences (Newton Advanced Fellowship), and the International Health Cohort Consortium (IHCC).
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Ethical standards
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
Supplementary Information
Below is the link to the electronic supplementary material.
Rights and permissions
Springer Nature or its licensor holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Fatori, D., Suen, P., Bacchi, P. et al. Trajectories of common mental disorders symptoms before and during the COVID-19 pandemic: findings from the ELSA-Brasil COVID-19 Mental Health Cohort. Soc Psychiatry Psychiatr Epidemiol 57, 2445–2455 (2022). https://doi.org/10.1007/s00127-022-02365-0
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00127-022-02365-0