Outcome-based resilience refers to the maintenance or quick recovery of mental health despite exposure to adversity, presumably resulting from a dynamic process of adaptation1. While resilience has been predominantly studied in the context of natural disasters, accidents, terror attacks, and other potentially traumatizing events1–3, the outbreak of the COVID-19 pandemic in 2020 has brought up new qualities and quantities of stressors affecting broad majorities of populations globally. This is illustrated by the surge in stress-related mental disorders such as depression and anxiety during the pandemic4. Especially people who were mentally stable before the pandemic showed large symptom increases during the crisis compared to those already affected by a mental disorder5. This underlines the urgent need to improve preventive mental health care and to identify predictors, processes, and potential intervention targets for strategies to promote mental resilience, not only during the current pandemic1,6 but also in anticipation of future global stressors.
Many studies worldwide have addressed questions of mental resilience during the COVID-19 pandemic via online surveys, conducted in China and other Asian countries7–13, Iraq14, Turkey15,16, Israel17, European countries18–25, the USA26–28, and Canada29. Increased levels of depressive symptoms and anxiety were frequently reported compared to population norms, while higher scores on trait resilience measures, behavioral coping strategies, and social support were cross-sectionally associated with lower symptoms of distress and/or better mental health. However, with the exception of our previous cross-sectional survey study “DynaCORE-C”24 , none of these studies took into account stressor exposure at the individual level, which is crucial in the operationalization of resilience as the maintenance of mental health despite exposure to such stressors30,31.
In DynaCORE-C24, we employed a residualization approach31–33, regressing internalizing mental health problems, retrospectively reported for a past two-week time window, against stressor exposure during that same time window. Using this method, individuals with a negative regression residual (a negative stressor reactivity, or SR, score) can be seen as showing lower-than-expected symptom severity given their level of stressor exposure (i.e., an indication of higher resilience), while individuals with positive residuals (positive SR score) show higher-than-expected stressor-related symptom severity (i.e., an indication of lower resilience). This approach addresses the problem that individuals may well show different degrees of mental health impairments in the Corona crisis, but that this may also be trivially explained by different degrees of experienced adversity (rather than by different resilience capacities).
Using this methodology, DynaCORE-C tested predictions of Positive Appraisal Style Theory of Resilience (PASTOR)30. According to PASTOR, individuals exhibiting a positive appraisal style have the general tendency to set the values that they on average attribute to stressors (that is, to potential threats to their goals and needs) on the key threat appraisal dimensions of threat magnitude or cost, threat probability, and coping potential to levels that realistically reflect the threat or even slightly underestimate it. Positive appraisers usually avoid catastrophizing (magnitude/cost dimension), pessimism (probability dimension), and helplessness (coping dimension), but also do not normally produce unrealistically positive (delusional) threat perceptions that might lead to trivialization, blind optimism, or over-confidence. As a result, their average stress reactions tend to be optimally regulated, in the sense that positive appraisers are well able to generate stress reactions where necessary but also avoid spending resources unnecessarily, that is, into unnecessarily strong, long, or repeated stress responses. This gives them enough time for recovery, resource rebuilding, and exploration and limits deleterious allostatic load effects and resource depletion as much as possible.
Thus, PAS includes constructs such as optimism (probability dimension) and general self-efficacy (coping dimension), but is broader and allows that, for instance, the negative effects of a person’s habitual pessimism may be compensated by positive effects of their low catastrophizing or good self-confidence. Insofar, optimism and general self-efficacy questionnaires, for instance, may be helpful but are potentially not sufficient for assessing PAS. The PAS questionnaire employed in DynaCORE-C took an alternative approach to indexing appraisal tendencies in the threat appraisal dimensions of PASTOR, by instead trying to measure beneficial cognitive processes, or mental operations, that people may habitually use in stressful situations and that help generate positive appraisal contents24. Notably, the processes addressed by the scale include variations of cognitive positive reappraisal, such as trying to find positive aspects or potential good outcomes of a situation, to put the situation in perspective, to accept it, or to detach from it.
The study found the dedicated PAS scales and also the employed optimism and self-efficacy scales to be positively associated with resilience (approximated by a negative cross-sectional SR score). In addition to these measures of habitual appraisal styles, situational positive appraisal specifically of the current Corona crisis was associated with resilience. Further, the effect of PAS on resilience was statistically mediated by a subjective measure of good stress response regulation (perceived good stress recovery)24, in accordance with theory.
Another claim of PASTOR is that the effects of other social, biological, and psychological RFs on outcome-based resilience are mediated by PAS. That is, RFs other than PAS are only beneficial for resilience to the extent that they shape someone’s appraisal style towards the positive30,34. For instance, certain genetic or biological factors may render the brain circuits mediating positive appraisal and reappraisal processes more effective; spirituality may help find meaning in hardships; or trust in one’s social networks may allow one to perceive many stressors as manageable. In this regard, DynaCORE-C observed that the effects of perceived social support were mediated by PAS 24.
Finally, DynaCORE-C found a weak cross-sectional association between behavioral coping style and resilience 24.
In addition to testing these eight different RFs, the study confirmed the well-known role of neuroticism as a negative RF, that is, risk factor, neuroticism showing an inverse association with resilience 24.
These RF-SR associations from the cross-sectional DynaCORE-C study would be substantiated if one could show that i) RFs also prospectively predict SR, ideally over an extended time window, and that ii) fluctuations in RFs are accompanied by fluctuations of SR, contemporaneously and/or prospectively (that is, with a time lag). Prospective associations in particular would help control for mood-congruency or other state-dependent effects that may have exaggerated the previously reported cross-sectional associations29.
To achieve this, we conducted a longitudinal study (DynaCORE-L) with repeated weekly measures of above RFs and of stressor exposure and mental health (to repeatedly calculate SR) over five consecutive weeks (Figure 1). To be able to address the potentially dynamic associations of RFs with SR over time, RFs were assessed on two different time scales: RF styles and RF modes. At baseline, most RF questions asked about participant’s typical or usual behavior. They presumably reflect properties or qualities that are relatively durably associated with a person or constitute a typical way or tendency in which a person reacts to life experiences, but may still gradually change over time, for instance through learning experiences and environmental changes. To demarcate these RFs from more trait-like RFs, we termed them “styles”, in keeping with30, and denoted them with the subscript S. Compared with traits, which are here denoted with the subscript T, styles are more likely to show adaptation over time and can be hypothesized as the basis for allostatic resilience processes31. Next to styles, at the weekly follow-ups, these same RFs were assessed as “modes” (denoted with the subscript M). With this new measurement approach, we assessed to what extent a particular RF was used or experienced in a given week. Complementary to RF style measures, RF mode measures may be more sensitive to changes in the strength of an RF, which would not become apparent from enquiring typical or usual behavior. Thereby, repeated RF mode assessments allow for examining how an RF potentially is associated with stressor reactivity on a shorter timeframe.
With this approach, we addressed the following five sets of hypotheses (Figure 1):
First (H1), we aimed to replicate the associations of RF styles and SR found in DynaCORE-C24, using the cross-sectional data assessed at baseline.
Second (H2), we aimed to extend the cross-sectional DynaCORE-C findings24 by exploring whether RF styles at baseline prospectively predict resilience, as approximated by the average SR score over all follow-up time points.
Third (H3), we investigated the relation between RF modes and SR scores within individuals longitudinally across weekly time points, predicting contemporaneous co-fluctuations.
Fourth (H4), in our primary hypothesis, we aimed to investigate the temporal dynamics of RFs and SR scores, namely, whether the use of RF modes is prospectively associated with the SR score assessed one week later (lagged association). For all analyses, we hypothesized a negative association between RFs and SR (except for neuroticism).
In line with PASTOR30 and previous results24, we further hypothesized that the statistical effect of perceived social support on SR is positively mediated by positive appraisal (PA). The mediation hypothesis was tested for each type of associations, that is, in a cross-sectional (H1_MED), prospective (H2_MED), contemporaneous (H3_MED), and lagged fashion (H4_MED).
Fifth (H5) and based on the consideration that the experience of stressors may compromise or, as in the phenomenon of stress inoculation35–37, potentially also strengthen RFs, we longitudinally investigated stressor exposure-dependent fluctuations in the RF modes measured in the subsequent week, hypothesizing that stressor exposure would be associated either negatively or positively with RFs in a time-lagged fashion.
All hypotheses, except for H2, were preregistered (https://osf.io/qm4ba). For simplification and better explanation of concepts, we changed the numbering of hypotheses relative to the preregistration and introduced the subscripts S, T, and M in this paper.