Research on disasters have progressed the science around psychological resilience, as they are specific measurable moments in time. Studies show that the majority of persons who experience adversity demonstrate the ability to bounce back from the trauma and disruption such as disasters (Hansel et al., 2020). The ability to rebound eventually leads to acceptance and incorporation of the emotional response into their new normal—re-envisioning or re-establishing a vision for the future (Norris et al., 2008). However, large scale disasters with longer recovery periods, and thus more chronic or accumulation of stressors for an uncertain or unknown end point can exhaust the natural coping methods and the ability to rebound. Previous trauma, bereavement, and loss (Osofsky et al., 2011) and financial setbacks (Drydakis, 2015) are all risk factors that should be of concern in the context of prolonged sequela. As such, increased anxiety and depression are common following disasters (Norris et al., 2002; Kessler et al., 2006) but this does not mean the absence of resilience. For example, studies following the Gulf oil spill in 2010 found increased mental health problems met with high resilience scores (Osofsky et al., 2011; Shenesey and Langhinrichsen-Rohling, 2015). Thus, understanding patterns of recovery, inclusive of resilience, are needed. It is likely that survivors will experience resilience, however the patterns of recovery, including resilience, is still not well understood – particularly when considered over time.

Social support is a strong predictor of resilience following disaster and trauma (Hall et al., 2010; Xu & Ou, 2014; Saltzman et al., 2017; Matthews et al., 2019; Osofsky et al., 2022); likewise, disruption of many support systems can challenge resilience (Cattan et al., 2005; Xu et al., 2011; Valtorta et al., 2016). In a systematic review of general social isolation, Leigh-Hunt et al. (2017) found associations with both poorer mental and physical health, yet the impact on resilience is not as well understood. Similar to social support, interdependence, or the belief that one’s actions are part of collective community change, has been noted as an important protective measure (Bollyky and Bown, 2020; Lo & Hsieh, 2020). Another study (Vashdi et al., 2020) noted the buffering effect of interdependence on more negative outcomes and the importance of connectedness as a mechanism toward collective resilience.

Resilience was first used to describe adaptations of the ecosystem and overtime extended to human systems and the built environment (Alexander, 2013). While scholars have tended to view them as siloed, research is beginning to demonstrate the connectedness of both psychological and social resilience. Abramson et al. (2015) provide a conceptual framework for resilience activation that considers human, social, economic and political capital at both the community and individual levels. Thus, resilience is not individual specific outcome but rather social determinants, such as support networks and available resources, are contributing attributes. A framework for resilience activation allows for complex and systematic factors that contribute to resilience, yet research is needed to support this theoretical model.

Similarly, collective resilience has been defined as a set of adaptive capacities that allow for consideration of and emphasize the importance of both individual and social attributes in the activation of resilience (Norris et al., 2008; Pfefferbaum, et al., 2008; Picou et al., 2009; Renschler et al., 2010). A collective framework also has the benefit of addressing some of the controversies around resilience regarding how it is defined and who is deemed resilient (Masten, 2014). Specific to disaster, the idea of bouncing back to unjust systems are also worth noting and require more consideration through a collective resilience framework. Clearly more research is needed to understand the psychological and social determinants that contribute to the resilience activation framework (Abramson et al., 2015).

The current study combines the frameworks of resilience activation and social resilience to explore the multi-system factors that influence psychological profiles of adaptation (including resilience) following a protracted disaster. In doing so, the current study aims to underscore the interaction of individual characteristics and contextual, social, and environmental factors that influence trajectories of adaptation. This work adds to the literature supporting both resilience activation and social resilience frameworks and expands the body of literature to include the application to large scale protracted disasters including public health crises.

Descriptive statistics for study measures and variables are presented in Table 1. Resilience was measured using the 2-item Connor Davidson Resilience Scale (Vaishnavi et al., 2007), asking participants if they (1) are able to adapt to change and (2) tend to bounce back from setbacks (α = 0.73, M = 8.23, SD = 1.37). General Anxiety Disorder assesses anxiety as (1) feeling nervous, anxious, or on edge and (2) being able to stop or control worrying (Sapra et al., 2020; α = 0.88, M = 4.80, SD = 1.96) in the past 30 days. The Patient Health Questionnaire assesses depression as being bothered by (1) little interest or pleasure in doing things and (2) feeling down, depressed, or hopeless (Kroenke et al., 2003; α = 0.86, M = 3.89, SD = 1.71).

Pandemic related stressors included asking if participants had experienced: loss of usual way of life, social isolation, work from home, community health concerns, children and adolescents being out of school, loss of income, personal health effects, participated in response or emergency services, loss of tourism, COVID-19 symptoms or diagnosis, loss of job or business. A stress index was created where 1 point was given for each of the experiences over the past year noted above consistent with previous research (Hansel et al., 2022). Life disruption was adapted from the Sheena Disability Scales (Sheehan et al., 1996) and included items that assessed to what degree the pandemic disrupted participants’ work or school life, social and leisure activities and family or home responsibilities activities (α = 0.61). Participants were also asked if they experienced a previous traumatic event (yes or no).

Social determinants included four life security questions (U.S. Department of Agriculture, 2012), including, in the last 2 weeks do participants: (1) have enough money to meet needs; (2) feel close to members of community; (3) worry whether food will run out before they get money to buy more; and (4) available information needed in day-to-day life. Interdependence was adapted for biological disaster relevance (Gerpott et al., 2018). Participants were asked to provide their level of agreement on the following indicators specific to the pandemic: (1) the outcome does not affect future interactions; (2) current behavior will have consequences for future outcomes; (3) personal behavior affects how others will behave in future situations; (4) what we do for [our actions] will not affect others; and (5) we need each other to get our best outcome. Items 1 and 4 were reverse coded for scale computation (α = 0.64).

Community impact was assessed by county level viral rate of COVID death and diagnosis through October 2020. Using the COVID Data Tracker, we matched viral rates to participants zip code (Centers for Disease Control, 2020). Geographic information (percentage rural) was assessed at the county level and matched to participants zip code (Robert Wood Johnson Foundation, 2018).

The mean age of participants was 46.3 years (SD = 14.7, 19–79 years of age); 84% identified as female, 16% as male and 1% as non-binary. The majority of participants identified as white (89%), 8% identified as black or African American, 5% as Latinx, 3% as Asian, and 1% as Native American (participants could select all that apply). Median income was $60,000 -$69,999 (2019 prompted as reference). Twenty percent have a high school education, 70% had a 4-year or professional degree, and 10% had a doctorate. Over half of the participants (63%) were married or cohabitating, 25% were single, 10% divorced or separated and 2% widowed. Participants (n = 334) resided in Louisiana (54%), Texas (6%), California (5%), Florida (4%), Georgia (3%), Illinois (3%). Two percent or less were from each of the following states: 2% from Mississippi, North Carolina, Pennsylvania, and Virginia; 1% from Arizona, Connecticut, Iowa, Kentucky, Minnesota, Massachusetts, New Jersey, New York, Maryland, Michigan, Missouri, Ohio, Oregon, South Carolina, and Washington; and 0.3% from Alabama, Alaska, Colorado, Delaware, Indiana, Nebraska, New Mexico, Rhode Island, Tennessee, Wisconsin, and Wyoming. Participants lived in urban and rural communities, with the minimum percent living in rural communities (M = 9%, SD = 14%, Mdn = 3.3%).

There were two phases of multivariate analysis; in one phase a series of latent profile models were estimated to identify unobserved groups in the data. A classify analyze approach using maximum posterior probabilities was used to assign respondents to their most likely latent class (Nagin, 2005). A new variable representing the latent class adaptation was then generated and used in phase two. The following measures were used as indicators of latent class membership: depression (PHQ-2); anxiety (GAD-2); and resilience (Connor & Davidson); higher values represented higher symptomology/resilience. The second phase included a step-wise multinomial logistic regression to predict class membership using a four group nominal variable representing class membership. Step one included the following intrapersonal predictors, age, marital status, race, education level, respondent sex, and income. Step two included community level data (viral rate and an indicator rural or urban residential area). Step three included the following psychosocial variables, perception of interdependence, disruption, stressors, trauma, and life security (financial, social support, food insecurity, and information access).

Given the uncertainties, life changes, and fear associated with the COVID-19 pandemic, psychological concerns were expected (Hansel et al., 2020). Participants reported COVID-19 resulted in a loss of their usual way of life, social isolation, and working from home. Further confirming expectations, the higher number of stressors resulted in increased anxious and depressive symptoms. However, these findings did not hold in the multivariate analyses, suggesting that the resilient group, experienced psychological resilience rather than less stressors. Consistent with the definition of resilience, exposure must be similar and the ones who are resilient are those who rebound or bounce back from adversity with lower mental health symptoms (Bonanno et al., 2007). Multivariate analysis revealed four main areas to focus efforts for resilience activation improvement - these areas include chronic stress, finances, loneliness and information access.

Other social determinant factors, such as demographics and community profiles, play a role in resilience activation. Age was confirmed as a protective factor in the multivariate results, where increased age decreased the likelihood of belonging to the depressed and anxious classes, thus older participants were more resilient. Other recent studies found youth were more susceptible to poorer mental health during the pandemic (Wilson et al., 2020; Browning et al., 2021). Similarly, being female increased the risk of being in the depressed or comorbid groups, a finding which has repeatedly been supported by past and recent literature (Barzilay et al., 2020; Xiong et al., 2020; Zhou et al., 2020). Conversely, identifying as White decreased the results of membership in the depressed or anxious groups, supporting concerns about growing health disparities (Bui et al., 2021; Masters et al., 2021; Young et al., 2021). Unsurprisingly, community indicators (i.e., viral rate) were predictive of belonging to both depressed and anxious, as compared to the resilient group, suggesting geographic location as a social determinant of resilience. Importantly, demographics and social determinants will vary greatly—while not specific to gender, age, and community indicators, service and resource allocation should be mindful of these indicators for implementation.