A longitudinal model for psychological distress in the COVID-19 crisis among brazilian graduate students

Recebido em: 19 jul. 2021. Aprovado em: 28 jul. 2021. Publicado em: 27 out. 2021. Abstract: As most evidence for mental health impacts of the COVID-19 crisis is cross-sectional, the present study aimed to analyze the longitudinal development of psychological suffering among 619 Brazilian adults by assessing mental health outcomes and individual factors in two periods: a year before and a month after the break of the pandemic. As major findings, pandemic psychological suffering was directly explained by previous-year suffering, conscientiousness, and pandemic perceived stress, and correlated with pandemic suicidal ideation. Pandemic perceived stress correlated with pandemic psychological distress, and was explained by previous-year suffering, neuroticism, and conscientiousness, as well as by pandemic life satisfaction and perceived pandemic impact. Finally, pandemic suicidal ideation variance was explained by prior ideation and pandemic life satisfaction. These findings are in line with current models of mental health and highlight the importance of integrating both more stable individual factors and more transient variables towards an explanation for mental health outcomes.

Palabras clave: salud mental, pandemia del CO- longitudinal,personalidad The World Health Organization (WHO) declared the COVID-19, the novel SARS-CoV-2 coronavirus disease, a pandemic in March 2020, after its fast spread around the globe as of 2019. In Brazil, a state of emergency was declared on February 3, 2020 (Ministério da Saúde, 2020a), the first case was confirmed in the 26th, and the first death, on March 17, 2020 (Ministério da Saúde, 2020b), with subsequent adoption of several sanitary policies for the containment of the pandemic. Pandemics and public health emergencies are associated not only with economic, sociopolitical, and human tolls, but also with important detrimental effects over the population's mental health (Brooks et al., 2000;Zortea et al. 2020).
Personality traits are also being investigated as potential predictors or moderators of mental health outcomes and attitudes towards the pandemic.
Neuroticism is associated with a greater individual tendency to experiencing negative emotions, vulnerability, emotional instability, and reactivity when facing distressing events, and regarded as an important independent predictor of several physical and mental health outcomes (McCrae & Sutin, 2018). During the pandemic, amounting evidence shows that the levels of neuroticism, even after controlling for socio-demographic characteristics, have positively and significantly correlated with intensity of fear towards the virus, feelings of boredom, and negative fantasies (Caci et al., 2020); to perceived stress and perceived threat (Liu et al., 2021); as well as to health-related anxiety, COVID related anxiety, and anxious and depressive symptoms (Nikčević et al., 2020).
Another Big-Five personality trait related to coping with COVID is conscientiousness, which, generally, describes a tendency towards feelings, thoughts and behaviors of control, self-efficacy, commitment, responsibility, discipline, parsimoniousness, and order (Sutin & McCrae, 2018). During the COVID-19 pandemic, conscientiousness has been associated with greater propensity and ease towards following and adhering to preventive measures, pandemic containment and social distancing (e.g., Carvalho et al., 2020). Furthermore, higher levels of conscientiousness are associated with greater self-efficacy for infection prevention, and with reduced perceived stress during the pandemic (Liu et al., 2021), as well as with greater subjective well-being (Modersitzki et al., 2020), lower levels of health anxiety and COVID related anxiety and, consequently, decreased symptoms of depression and generalized anxiety (Nikčević et al., 2020).
However, most of cumulative evidence regarding individual associated factors and potential mental health predictors in the context of the COVID-19 pandemic are restricted to cross-sectional data, which are not sufficient to establish causality. Among the few existing longitudinal studies, there is conflicting evidence regarding psychological impact (Prati & Mancini, 2021). Some did not find statistically nor clinically relevant change amidst mental illness measures or found a reduction of these (e.g., China, Wang et al., 2020b;Argentina, Juric et al. 2020), whereas other studies have identified increased fear, preoccupation, and symptoms related to stress, anxiety, and depression throughout the pandemic (e.g., Espanha, Multicultural, Lippold et al., 2020). Most of these studies, nonetheless, have demonstrated significant differences in respect of several COVID related variables (e.g., showing symptoms, being part of a risk group, knowing someone who has been infected, and others) as well as in respect of age and gender, with younger participants and women showing worse effects (e.g, Juric et al., 2020;Lippold et al., 2020;Planchuelo-Gómez et al., 2020). Other socio-demographic factors (e.g., educational level, marital status, occupational status, and others) yielded diverse results, and sexual orientation data were not analyzed in the identified longitudinal studies, despite cross-sectional evidence of LGBTQ+ groups being significantly more affected (Suen et al., 2020). Finally, only Lippold et al. (2020) included personality individual differences and concluded that neuroticism was the best predictor of fear and preoccupation related to the pandemic, after controlling for political orientation, risk group status, gender, and educational level. Importantly, although the four cited studies had a large number of participants, their follow-up intervals were of approximately 2 to 5 weeks, starting from the beginning of the pandemic, and two of them analyzed changes between non-coinciding groups across time (Lippold et al., 2020;Wang et al., 2020b). Thus, only two of the studies have performed totally or partially paired analysis, i.e., for the exact same participants along a time period (Juric et al., 2020;Planchuelo-Gómez et al., 2020).

Participants
The participants of the current study were part of both a 2019 project, which investigated mental health among post-graduate students, and a

4/15
Psico, Porto Alegre, v. 52, n. 3, p. 1-15, jul.-set. 2021 | e-41332 Measures Socio-demographic characterization and pandemic perception questionnaire. The authors developed a structured questionnaire that included socio-demographic (age, gender, race/ethnicity and sexual orientation) and COVID-19 related questions. The only COVID-19 related item included in the present study evaluated the subjective perception of the impact of the pandemic over the participants' life, the Pandemic Impact variable. The item asked, "How much has your life and yourself been affected by the COVID-19 pandemic" and provided 4 possible answers, ranging from "Neutral, I do not think I have been affected as much as other people yet" to "Very much affected". (Lovibond & Lovibond, 1996) (Chang & Chang, 2016). FSII comprises 5 items that evaluate the frequency of suicidal ideation along the last 12 months through a 5-point Likert scale that varies from 1 ("never") to 5 ("almost every day"). The Brazilian version, FSII-Br, shows good psychometric properties and a Cronbach's Alpha of 0.88 for its single factor (Teodoro et al., 2020). FSII total scores vary from 5 to 25 points, and higher scores mean higher frequency of suicidal ideation in the last 12 months. This scale was completed by participants at both periods, 2019 and 2020, and total scores correspond to FSII2019 and FSII2020 variables.

& Mermelstein, 1983).
We used the 14-item version, in which answers are organized in an ordinal scale of 5 points that vary from "never" to "very frequently". Each item can receive values from 0 to 4 and, accordingly, minimum total score is 0 and maximum total score is 56. The greater the total score, the smaller is the degree of confidence that one's life has been unpredictable, uncontrollable and overwhelmed in the month before the assessment. The Brazilian version of PSS shows a satisfactory level of intern consistency (α = 0.83, Dias et al., 2015). This measure was collected only in 2020 and coded as PSS2020.

Procedure
The 2019 assessment was part of a broader study which aimed to characterize mental health and psychological suffering among Brazilian postgraduate students and took place between March and July -therefore, before the first documented global case of COVID-19, which occurred in to 20 minutes to be answered. All respondents provided informed consent at each period of the study. The study protocol was approved by Ethics Committee (CAAE 07077019.3.0000.5149).

Data Analysis
Univariate descriptive analysis was conducted using the psych package, version 2.0.9 (Revelle, 2020), and multivariate normality analysis was (iii) invariance of intercepts and regressors. The compared groups were the male (n = 169) and female (n = 450) participants (gender 2020 variable) and the heterosexual (n = 482) and LGBTQ+ (n = 129) orientation participants (sexual orientation 2020 variable; 8 "other"/"no response" participants were excluded from this group analysis). Here, the packages used are the same as in the invariance analyzes of the questionnaires.

Descriptive Analysis
Participants' scores in SWLS items did not present multivariate normality.

Path Analysis
Model 1 showed inadequate fit to the data    FSII2020, and PSS2020 were, respectively, 64%, 45%, and 41%. DASS2020 was directly explained by the scores of the participants in the previous year (DASS2019), conscientiousness in the previous year (Consc2019; although with low predictive value = -0,05), as well as by the scores in PSS2020. Also, DASS2020 was correlated with 2020 FSII scores, but not explained by them. As for PSS2020 scores, they correlated with DASS scores of the same year, and its variance was explained by the scores in DASS (DASS2019), neuroticism (Neurot2019) and conscientiousness (Consc2019) from the previous year, as well as by the SWLS scores (SWLS2020), and Pandemic Impact, both from 2020. FSII2020 was correlated with DASS2020, and its variance was explained by scores FSII scores from the previous year (FSII2019), as well as by SWLS (SWLS2020) from the same period. The subjective perception of Pandemic Impact was explained in 8% of its variance, and 2020 ESV was its most important predictor, suggesting that increases in self-reported life satisfaction in 2020 were accompanied by reduction on participants' perception that their life was affected by the pandemic.

Discussion
The main objective of the present study was For this purpose, invariance analysis of the scales from each period have been conducted to assure that 2019 and 2020 total scores were comparable.
First, the frequency of suicidal ideation in 2020 was positively correlated with general psychological suffering by 2020. This result has also been found in other studies (Ammerman et al., 2020;Franklin et al., 2017;Wetherall et al., 2020) and points out to the fact that thoughts of hopelessness and sadness, and intense experiences of suffering and grief are probably comorbid with anxiety and depression, even though there is no necessary causal relation between them. Interestingly, in the longitudinal model that demonstrated the best fit indices, frequency of suicidal ideation in 2020 (during the pandemic) had 45% of its variance predicted uniquely by the frequency of suicidal ideation in 2019 (before the pandemic), and life satisfaction in 2020 (negatively). Frequency of suicidal ideation in 2019 was the best predictor for the same variable in 2020, which might suggest that, although the assessment of suicidal ideation is based on shortterm temporal criteria (e.g. last week, last month), there is some aspect of psychological functioning that is more stable, and that maintains these kind of self-injurious thoughts for longer periods of time, and does not seem to correspond neither to a personality trait (such as neuroticism), nor to the existence of previous mental distress (as measured by 2019 DASS-21). In this sense, it could be argued that this stability is not restricted to the COVID-19 context, but more evidence to support this claim should be gathered in future studies.
Notably, subjective perception of pandemic impact over one's life was not capable of predicting frequency of suicidal ideation in 2020, which goes against previous findings from crosssectional studies carried out in other countries (Ammerman et al., 2020). The present results might be explained by the data collection period that took place a month after the start of social  (Galderisi et al., 2015). This observation seems to be corroborated not only by other studies in the field of mental health (Fergusson et al., 2015;Siegmann et al., 2017), but also by other findings of the present study.
Life satisfaction during the pandemic moderately and negatively predicts both perceived stress in the same period (2020) and the perception of how much the changes that come with the pandemic control measures affect daily life. Therefore, the most satisfied with life post-graduate students showed lower levels of perceived stress and felt like their lives had been less affected by the changes they had been through during the first month of the COVID-19 pandemic. Similar findings were also found in cross-sectional studies (Ammerman et al., 2020;Casali, 2020;Pradhan et al., 2020). Collectively, these results strengthen the account of perceived well-being as a protective factor that reduces negative effects of stress. It should be noted that a sense of satisfaction with life does not imply that individuals will not experience negative life events or setbacks. Conversely, it is associated with the acknowledgement that, even though bad things happen to everyone, they can still see that there are reasons for living, and that they have further resources to help them overcome challenges. In this sense, life satisfaction might be associated with values, with personality characteristics linked to optimism and positive affect, and also with self-efficacy, which may collectively bestow life satisfaction with its aforementioned protective effects (Dyrenforth et al., 2010;Marcionetti & Rossier, 2016). containment measures, and in the heightening of fear, perceived risk, and perceived stress levels (Carvalho et al., 2020;Lippold et al., 2020;Liu et al., 2021;Mikcevic et al., 2020). Hence, it could be suggested that people with higher levels of neuroticism (negative affect) are more susceptible to negative bias and lack of emotional regulation strategies when facing unexpected events, which, associated with reduced self-efficacy and ability to take on self-protection measures (lower level of conscientiousness), leaves them more prone to experiencing higher levels of stress during the pandemic, which, in turn, may add to the psychological suffering. It is also worth highlighting that life satisfaction scores (cross-sectional and longitudinal) seem to indirectly impact, as seen in the modeling done, the degree of psychological suffering among participants, through the (negative) influence they exert on the levels of perceived stress and subjective pandemic impact. As mentioned before, life satisfaction is, currently, one of the most important components of mental health models (Galderisi et al., 2015;Marcionetti & Rossier, 2016).
Finally, the best fitting model was tested for invariance regarding sex (feminine and masculine) and sexual orientation (heterosexual and LGBTQ+), since literature indicates that women and LGBTQ+ groups were the most affected by the pandemic, showing higher levels of psychological suffering (Di Crosta et al., 2020;Suen et al., 2020). The findings of the present study demonstrate invariance of the longitudinal model for both gender and sexual orientation. This is highly informative for two main reasons. First, because it follows the assertion that these groups' averages can truly be compared. However, above all, it demonstrates that even though men and women, and heterosexuals and LGBTQ+, have distinct levels of psychological suffering, their predictive factors seem to be shared -at least when considering the pandemic context analyzed in the present study. Nonetheless, it should be noted that other factors that were not included in the present analyses (such as excess of responsibilities and work overload due to the pandemic; triple journey; need for confinement with family or people from unhealthy, unaccepting and uncaring relationships) could add to the explained variance of the model, and offer an even more comprehensive understanding of the documented differences between men, women, heterosexuals and LGBTQ+ (Auerbach et al., 2018;Kneale & Becares, 2020;Suen et al., 2020).
There are a few limitations to the present study.
Data collected in 2020 correspond to the first four weeks of social distancing in Brazil. In this sense, even though there is evidence that acute effects over mental health take place during health emergencies and crisis, a greater negative effect due to the pandemic, and especially due to social distancing measures, might take a longer period to become apparent. Moreover, because participants were post-graduate students, specific variables related to difficulties experienced in graduate school could interfere with general psychological suffering levels yet were not controlled in the present study. Finally, we used only self-report measures, and our sample size did not allow the use of a SEM strategy that could account for latent variables, and thus our analyses are based on scales' total scores. Despite these limitations,