Emotion Network Analysis During COVID-19 Quarantine - A Longitudinal Study

Abstract

The coronavirus disease 2019 (COVID-19) emergency imposed important challenges in the lives of individuals, particularly since the restriction of free movement and limitation of social contact started. This quarantine strategy has been used for centuries because self-isolation can help contain and control the spread of infectious diseases. However, both isolation per se and its uncontrollability have important negative psychological effects on individuals. Previous pandemics, such as those associated with severe acute respiratory syndrome (SARS), have been described as a mental health catastrophe due to the widespread psychopathology associated with the disease (Gardner and Moallef, 2015). In fact, some people become so anxious, distressed, avoidant, and functionally impaired under pandemics that end up requiring treatment due to the development of an emotional disorder (Wheaton et al., 2012). In this sense, although SARS was dangerous for the elderly and medically fragile, the psychological impact of SARS also inflicted a great deal of suffering in terms of the number of people affected by it and its duration (Chang et al., 2004; Washer, 2004). In another study, respondents who had been quarantined, those who worked in high-risk locations such as SARS wards, or individuals who had close friends or relatives who contracted SARS were 2–3 times more likely to have post-traumatic stress symptoms than people with lower exposure levels (Wu et al., 2009). Thus, it seems clear that mental disorders can be triggered or exacerbated by pandemic-related situational stressors (Wu et al., 2005; Gardner and Moallef, 2015; Shultz et al., 2015).

However, as evident as the effects of this quarantine during a pandemic in humans can be, we do not fully understand the psychological dynamics of mood during early quarantine stages and its longitudinal changes over the first 20 days of the COVID-19 quarantine period. Being quarantined is a complex psychological phenomenon that is hard to disentangle because there are numerous interactions between emotions and regulatory mechanisms in order to adapt to this strange and threatening new situation (Pfefferbaum and North, 2020; Suso-Ribera and Martín-Brufau, 2020). Cross-sectional studies fall short to investigate the psychological adaptation to quarantine and even pre-post studies have limitations in understanding what happens during the adaptation process (Brooks et al., 2020; Wang et al., 2020). For these reasons, longitudinal research seems to be the gold standard methodology to monitor these trajectories when attempting to better understand human psychological responses to pandemics. In addition and in contrast to the simplistic view of “one-size-fits-all” stress response to a potential traumatic situation, different trajectories have been proposed during the 2003 SARS outbreak using a latent class approach, namely, recovery, resilient, delayed, and chronic responses (Bonanno et al., 2008). To better understand these responses, a complex longitudinal analysis is needed to understand the variation and mutual influences of emotional network dynamic patterns during the early stages of the adaptation response to quarantine. This requires a new framework different to a latent approach.

Following an affective provocation, emotions interact as a dynamic and time-dependent system (Davidson, 2015). This network of emotions changes as a result of internal and external factors (Frijda, 2007). These fluctuations better characterize emotional response than mean levels of emotions (Kuppens et al., 2007; Sperry and Kwapil, 2019) and can be used to predict mood psychopathology (Wichers et al., 2015; Sperry et al., 2020). In fact, emotion dynamics may be key to understand pathways to psychopathology and well-being (Wichers et al., 2015). For these reasons, to study emotion fluctuations as a dynamic temporal network offers a good opportunity to study the response to stressful situations and increase our understanding of basic emotional responses and could suggest sooner and more successful interventions in the future.

The new field of network psychometrics has been used in recent years to investigate the complex structure of various psychiatric disorders (Fried, 2017), including depression (Fried et al., 2016), psychosis (Isvoranu et al., 2016), schizophrenia (Levine and Leucht, 2016), and anxiety (Beard et al., 2016), among others. The network perspective offers a novel way of understanding the dynamics of psychopathology (Borsboom, 2017). In contrast to viewing symptoms as reflective of underlying latent categories or dimensions, network analysis conceptualizes symptoms as constitutive of mental states, not reflective of them (McNally, 2016). At the heart of the theory lies the notion that psychopathological symptoms are causally connected through myriads of biological, psychological, and societal mechanisms. If these causal relations are sufficiently strong, symptoms can generate feedback that maintains symptomatology. In this case, the network can become stuck and develop into a disorder state (Borsboom, 2017). Ultimately, network analysis is a form of time-series analysis that has been recommended for its use in complex models where interactions between system components (e.g., different mood states) need to be modeled. This is done by graphically representing the interactions among system elements by means of edges and nodes (Gao et al., 2016). Thus, mood changes could be studied as networks, and this methodology could detect complex interactions between mood states over time that would be otherwise undetectable using pre–post methodology.

Repeated short-term assessments can detect variations in the presence and severity of states and reveal dynamic processes between them (Ebner-Priemer and Trull, 2009; Myin-Germeys et al., 2009; Bolger and Laurenceau, 2013). Network models can be used to investigate such dynamic processes in repeated assessment data from one participant [vector autoregression models (VARs)] or data from multiple participants (multilevel VAR; Epskamp et al., 2016). These models produce temporal networks depicting a directed network of the lagged associations of symptoms from one time point to the next for which Granger causal connections between symptoms are inferred (Schuurman et al., 2016). Temporal networks can then be used to identify symptoms with a high “out-strength,” that is, symptoms that are most predictive of other symptoms at the next time point (Epskamp et al., 2016).

The study of mood and its temporal evolution is important for several reasons. Moods, for example, are different to emotions in a number of characteristics, including the fact that they last longer (Ekman and Davidson, 1994). In fact, moods can have an impact on emotions (i.e., they lower the threshold that is required to trigger an emotion) (Thorndike et al., 1991). Therefore, moods can predispose individuals to experience situations in a certain manner, which can ultimately impact the way they cope with stressors (Berrocal and Extremera, 2008), such as being quarantined. Research into the determinants of mood states has been dominated by personality theories. For example, personality models like the five-factor model have shown that individuals high in neuroticism tend to present more unstable mood states (e.g., emotionality) and tend to be dominated by negative mood states (e.g., sadness and anxiety), while extraverted individuals tend to report more positive mood states (e.g., vigor; Garrity and Demick, 2001). The literature has shown, however, that mood states are influenced not only by internal factors (i.e., personality) but also by external elements (e.g., stress; Kudielka et al., 2004). As noted earlier, such changes in mood are important as they can lead to differential adaptation to adverse environments as they predispose to certain emotional states and coping efforts (Catanzaro and Mearns, 1999).

During the COVID-19 pandemic, several calls have been made to better understand the impact of the quarantine, an external stressor, on the mood status of individuals across time (Brooks et al., 2020; Lima et al., 2020). To do so, we would need to compare mood during the quarantine with mood prior to the quarantine. In the present study, however, only data after the quarantine were obtained, with the intention to explore how mood states develop over time under such strange situations using complex interaction statistical methods to study the evolution of networks of mood states under a pandemic, which can inform about human adaptation mechanisms under stressful conditions. Thus, the aim of this study was to explore the psychological dynamics of mood changes during the first stages of the COVID-19 quarantine in a sample of Spanish individuals from the general population using longitudinal data in a multilevel framework.