The COVID-19 pandemic has indisputably had significant consequences for physical and mental health. In addition to the threat posed by the virus itself, individuals have experienced disruptions to daily living on a massive scale, including changes in work or employment status, economic difficulties, and other financial hardships (Panchal et al., 2021). That is notwithstanding the highly significant loss of life and long-term medical disability associated with contracting the disease itself (Hodgson et al., 2021; Lopez-Leon et al., 2021). Moreover, these pressures have created even greater difficulties for individuals already contending with other stressors (Manchia et al., 2022). Large-scale community disasters like pandemics have been shown to threaten otherwise adaptive adjustment processes (Goldmann and Galea, 2014). In particular, diminished access to social resources that might otherwise help to mitigate stress could greatly compound the risk for mental illness (Belsher et al., 2012). In this investigation, we sought to explicate the impacts of the COVID-19 pandemic on social support processes in adults recovering from injuries sustained during traumatic events (e.g., motor vehicle crashes, violence, industrial accidents, etc.).

Social support is a consistent and highly protective factor in psychological adjustment to traumatic events as demonstrated by decades of research (Cohen and Wills, 1985; Joseph, 1999; Guay et al., 2006). We focused specifically on social support, given explicit, broadly administered social recommendations that served as countermeasures to slow the spread of the COVID-19 virus, including stay-at-home orders, self-isolation/quarantine periods, limited public social interactions, and being otherwise less likely to see close friends and other important social contacts (Centers for Disease Control, 2020). These “social distancing” recommendations may have impacted individuals’ experiences of social support (Mancini, 2020). As prevailing theories suggest that social support serves a “buffering” effect on adjustment to stress, understanding to what degree the pandemic has impacted its influence on psychological adaptation to traumatic events is critical (Cohen and Wills, 1985; Arnberg et al., 2012; Evans et al., 2013).

The current study sought to test the contextual impact of the COVID-19 pandemic on social support processes and psychological adjustment. Because the protective role of social support may be most evident in the aftermath of acute stress (Orlas et al., 2021), we focused our investigation on adults admitted to the hospital following injury caused by traumatic events (e.g., serious motor-vehicle crashes, violence, etc). We applied a case–control design defined by context so that comparisons could be made between a “control” cohort recruited before the active phase of the pandemic in the United States (2018 to March 9, 2020) and a “case” cohort recruited during the pandemic (December 2020 to April 2022). Specifically, our twofold goal was to examine whether (1) experiences of social support were affected by the pandemic and (2) whether the process of adjustment was influenced by any resulting changes in social support. Keeping with a nuanced conceptualization of social support, we considered three different operationalizations: perceived social support, social network size, and report of social constraints. Given prior evidence suggesting the unique impacts of the pandemic on particular groups (e.g., lower SES groups, the elderly; Patel et al., 2020; Grolli et al., 2021), we matched our samples carefully based on demographic, economic, and injury-related characteristics to ensure a rigorous comparison. Moreover, we not only tested cohort differences in social support and psychological health indicators over time, but also tested potential moderation by COVID-19 context in the expected associations between social support and psychological adjustment over a 5-month period. For these latter tests, we expected, based on extensive prior research, that early reports of social support (perceived social support, social network size) would predict fewer psychological symptoms at 5-months post-trauma, but that social constraints on disclosure would predict higher symptoms. Although we did not develop any specific a priori predictions for whether COVID-19 would impact adjustment or not, we did anticipate that any impact would likely be deleterious to adjustment.

First, as preliminary analyses, we confirmed the matching of the two cohorts, on demographic and injury severity variables (see Table 1). Next, we ran zero-order correlations between all key variables, followed by independent-samples t-tests and repeated measures analysis of covariance (ANCOVA) to determine whether the cohorts (pre-COVID-19 vs. COVID-19) differed on any of the key variables, including perceived social support at times 1 and 2, social network (time 1 only), social constraints (time 1 only), depression symptoms at times 1 and 2, and post-traumatic stress symptoms (PTS) at times 1 and 2. In addition, because of our modest sample sizes, we confirmed any null findings with tests of equivalence using the two one-sided tests (TOST) approach with the “TOSTER” package in R (Lakens et al., 2018). Equivalence tests allowed us to determine whether non-significant effects reflect true equivalence or, rather, were best explained as the result of underpowered analyses. This approach consisted of setting equivalence bounds that allowed us to test whether we could statistically reject the presence of meaningful differences (Lakens, 2017). In the absence of precedent for effect sizes in this context, equivalence bounds were set using the smallest effect size detectable by our sample with 80% power (Lakens, 2017). An a priori power analysis to determine the equivalence testing boundary was executed using G*Power 3.1 (Faul et al., 2007). This simulation indicated that we would have 80% power to detect an effect size of d = 0.55, hence equivalence bounds were set to d = −0.55 and d = 0.55. This boundary is largely consistent with clinical efficacy effect size recommendations (e.g., Jakobsen et al., 2014) and therefore is a useful estimate of clinically meaningful difference. We did consider matching variables as potential covariates if they correlated with any of the outcome indices, however, only age did and thus all other matching variables were not considered in analyses.

Next, we conducted longitudinal models to test whether the protective effects of social support on adjustment depended on the cohort (Pre-COVID-19 versus COVID-19). We conducted linear ordinary least squares (OLS) regressions with depression or PTS symptoms (time 2) as the outcome variable. Social support (time 1) was conceptualized as the primary independent variable and cohort (cases v. controls) as the moderator. For each model, we co-varied age and symptoms at time 1, which were added in step 1. The social support variable was added in step 2, cohort in step 3, and their interaction term in step 4. Variables were centered as recommended. We tested all three measures of social support at time 1 (perceived social support, social network, and social constraints) as predictors in separate models. Finally, we conducted linear OLS regression with perceived social support at time 2 as the main predictor of symptoms at time 2, while controlling for symptoms at time 1. For this model, we followed the same iterative approach as above and tested for moderation by cohort. For any significant moderation effects in the above models, we conducted tests of simple slopes and plotted the moderation by cohort.

Given the number of OLS regression models conducted, we applied the Benjamini-Hochberg False Discovery Rate (FDR) Correction to correct for Type 1 error (Benjamini and Hochberg, 1995). This method consists of rank-ordering value of ps of each independent variable within a set of statistical tests in ascending order (smallest to largest value of p), then applying the following formula to each variable: (i/m)*q, where i is the variable’s value of p rank, m is the number of variables within the set of hypothesis tests, and q is the critical value of p chosen to interpret statistical significance (or the FDR). The largest value of p within the set of variables that was lower than its corresponding Benjamini-Hochberg critical value (based on (i/m)*q) was then interpreted as the critical value of p for determining statistical significance. We applied this method for two sets of hypothesis tests: (1) the four OLS regression models with depression symptoms at time 2 as the outcome variable, and (2) the four OLS regression models with PTS symptoms at time 2 as the outcome variable. We chose an initial FDR of p = 0.05 (the value used for q in the above formula), and both sets of hypothesis tests contained 20 variables each (m = 20). The results of this approach indicated that (1) any tests with depression symptoms as the outcome with a value of p of p ≤ 0.003 should be interpreted as statistically significant, and (2) any tests with PTS symptoms as the outcome with a value of p of p ≤ 0.008 should be interpreted as statistically significant. See Supplementary Tables S1, S2 for a full breakdown of the Benjamini-Hochberg Method applied to these two sets of hypothesis tests.

In this investigation, we tested the impact of the COVID-19 pandemic on social support and the process of adjustment following traumatic events in United States adults. Using a matched case–control design, we compared reports of social support and psychological symptoms in United States adults in the first 6 months following traumatic events (e.g., motor vehicle crashes, violence). The two cohorts (pre-COVID-19 and during-COVID-19) were defined by recruitment and data collection date to ensure no overlap for the comparison and were extracted from a larger sample. We carefully matched cohorts based on key characteristics relevant to psychological adjustment (e.g., income, gender). The results suggested that the cohorts were equivalent in reports of social support and psychological symptoms at either the initial stage after the event (approximately 1-month post-event) and 4 months later. Although comparison of symptom reports by cohort were not statistically significant, equivalence tests suggested that differences, particularly for depression, were still possible. Longitudinal analyses of adjustment suggested that, as expected, social support indicators were generally associated with improved adjustment over time (including reduced depression and post-traumatic stress symptoms). However, there was key evidence that cohort, or the specific timing of data collection (pre-versus during-COVID-19) moderated the process of adjustment. Specifically, the results indicated that individuals in the COVID-19 cohort (as compared to the pre-COVID cohort) did not receive as much benefit from social support over time in relation to symptoms of post-traumatic stress. In sum, although generally there was evidence that reports of symptoms and social support were equivalent in participants before or during the COVID-19 pandemic, and that social support largely enhanced psychological adjustment to trauma, there was some evidence suggesting that during the COVID-19 pandemic, social support offered significantly less protection from psychological symptoms to U.S. adults recovering from traumatic events.

The COVID-19 pandemic has been an international crisis, with lock-down and social distancing orders altering the social lives of billions of people leading to what many have called a “loneliness epidemic” (Horigian et al., 2020; Killgore et al., 2020; Okruszek et al., 2020; Tso and Park, 2020). However, consistent with our findings, cross-sectional and longitudinal data have shown little overall change in loneliness after the onset of COVID-19 (Aknin et al., 2022). There has also been evidence that demographic factors have influenced individual variability in adjustment to this stressor. For example, there is clear evidence of sex differences in adjustment, potentially due to differences in parenting roles and the burden of remote schooling on mothers in particular (Connor et al., 2020). Moreover, there is evidence, particularly in the United States, that economic status greatly impacted not only the psychological stress of individuals, but even the risk for the disease (Finch and Hernández Finch, 2020). However, importantly, in our investigation, we found no difference in reports of social support even when holding key factors such as economic status and gender, constant through careful cohort matching. Specifically, our tests of equivalence suggested that three different reports of social support—perceived social support, social network size, and social constraints—were equivalent regardless of whether made prior to, or during, the pandemic. Moreover, there was no evidence of a significant difference in psychological symptoms between the two cohorts at the initial stages of adjustment.

Our analysis of longitudinal adjustment revealed some more complex findings, suggesting that the process of adjustment for adults following trauma, may have been impacted by the pandemic. For example, although our null hypothesis tests of both depression and post-traumatic stress (PTS) symptoms at 5 months post-event were not significant, our equivalence tests indicated that we could not presume that there were no differences between cohorts. Moreover, we did find significant moderation by COVID-19 cohort when predicting PTS symptoms at 5 months, suggesting that social support was less protective during adjustment to trauma during the pandemic, as compared to just prior to the pandemic. There is limited research examining social support during the pandemic in relation to PTS symptoms. One prior investigation of perceived social support during COVID-19 found that higher perceived social support predicted less post-traumatic stress symptoms (Zaken et al., 2022). However, this investigation was cross-sectional and did not test a longitudinal model of adjustment. Importantly, the authors suggested that social support is effective at buffering against some stressors but may not be effective at buffering against the stressors directly related to COVID-19. Our findings support this hypothesis by providing evidence that social support was less effective at reducing post-traumatic stress symptoms during COVID-19 compared to pre-COVID-19.

One explanation may be that although individuals reported the same amount of social support, the quality of that support may have differed. During social distancing orders, many social interactions were moved to virtual settings, which may provide less social reward than in-person interactions (Ferreira et al., 2022; Towner et al., 2022). It is also plausible that other stressors related to COVID-19 reduced the protective effects of social support. Indeed, there is evidence of increased reports of anxiety around opportunities for exposure in social contacts (Etilé and Geoffard, 2022; Saeed et al., 2022), therefore social interactions may have fueled more anxiety compared to pre-COVID-19 interactions. Although our cohorts were matched based on age, sex, education, income, and injury severity, there may be other variables that were influential within the context of COVID-19. Research in the first year of the pandemic has suggested that personal experience with, or proximity to, COVID-19 was a predictor of psychological health (Aknin et al., 2022). Specifically, being personally diagnosed with, or having symptoms of, Covid-19 was associated with increased psychological distress; having a friend or family member with COVID-19 increased reported anxiety; and being around COVID-19 (e.g., healthcare workers) was a risk factor for greater psychological distress (Aknin et al., 2022). Research has also suggested that how people spent their time during the pandemic was influential, with activities like gardening, exercising, and reading leading to lower rates of psychological distress and higher rates of life satisfaction (Disabato et al., 2022). Thus, many factors beyond social support contributed to individuals’ adjustment to the pandemic.

One complicated finding from the current investigation was that COVID-19 erased the protective effects of social support against PTS symptoms, but not depression symptoms. That is, social support was equally protective against depression symptoms for both cohorts but was less protective against PTS symptoms for the COVID-19 group. This difference could be due to qualitative differences between PTS and depression symptoms. Specifically, PTS symptoms are thought to differ from depression on dimensions of arousal and avoidance (Metzger et al., 2004; American Psychiatric Association, 2013), and past research (though limited) has investigated the differential relationships between social support and different clusters of PTS symptoms (McGuire et al., 2018). Indeed, Joseph et al. (1997) previously presented a theoretical model proposing that social support should impact PTS symptom clusters differently. For example, they suggested that arousal-based symptoms likely benefit from emotion regulatory resources received from positive social interactions, whereas PTS symptoms related to distressing thoughts (e.g., symptoms more closely related to depression) may be affected by others’ interpretations. Importantly, McGuire et al. (2018) investigated the protective effects of social support on depression and PTS symptom clusters following trauma exposure in Hurricane Katrina victims. Results showed protective effects of social support for depression symptoms as well as avoidance-and arousal-based PTS symptoms for individuals who were not displaced by the hurricane. However, the protective effects of social support on arousal-and avoidance-based PTS symptoms were not evident in victims displaced by the hurricane, whereas the protective effect nonetheless remained for depression. Thus, both our and McGuire and colleagues’ findings show that the protective effects of social support on PTS symptoms (but not depression symptoms) are diminished under conditions of higher stress (COVID-19 for the present study, displacement by a hurricane for McGuire and colleagues). Moreover, our exploratory analyses specifying the relationship between social support and arousal-and avoidance-based PTS symptoms further support this pattern of findings. However, more research is needed to delineate the differential impact of social support on PTS and depression symptoms in environmental contexts of high stress.

Importantly, some interesting differences in outcomes emerged depending on which measure of social support was included in the model. Indeed, the three indices of social support have important conceptual differences. For example, perceived social support (or functional social support) is thought to directly buffer against stress, whereas the benefits of a larger social network (or structural social support) are thought to come from a greater sense of belongingness and access to resources (Cohen and Wills, 1985; Kawachi and Berkman, 2001; Stewart et al., 2022). Overall, perceived social support had the most reliable protective effect across analyses, but individuals in the COVID-19 cohort did not appear to receive this benefit when PTS symptoms were the outcome. Further, results indicated that a larger social network may have even been harmful for individuals in the COVID-19 cohort (but protective for the pre-COVID-19 cohort). It is possible that these two findings go hand-in-hand. Given that individuals in the COVID-19 cohort did not benefit from the perceived receipt of social support from close others, having a larger social network could have increased feelings of frustration with less access to close others during the time of their adjustment period. In addition, some, though limited, research has highlighted contexts in which social support is experienced as burdensome (e.g., individuals with chronic illness; Palant and Himmel, 2019). However, this line of research has not been extended to the context of the COVID-19 pandemic, and any additional commentary would be conjecture. Rather, our findings highlight a need for further investigation into potential nuanced effects of social support in the context of normal adjustment to PTSD-qualifying events during a broader stressful context such as the COVID-19 pandemic. In particular, it seems highly relevant that research consider broader contextual processes as well as multiple operationalizations of social support in future investigations of this kind.

This study provides compelling evidence that social support was equivalent regardless of timing in relation to the COVID-19 pandemic. A major strength of our investigation is the case–control design in which cohorts were rigorously matched based on age, gender, race/ethnicity, education, reported income, and injury severity status. We included multiple indices for comparison and operationalized social support in three ways: perceived social support, social network size, and social constraints within important relationships. Given the relative consistency in our findings across operationalizations of social support, we can have more confidence in our results. We were also able to model symptoms longitudinally, with symptoms reported at two time points approximately 4 months apart, in a high-risk sample recruited shortly after traumatic events and injury.

Despite these key strengths, there were a number of limitations. The rigor of the case–control matched design did, unfortunately, limit the overall sample size included, and there was some indication that null findings in symptom comparisons by cohort at 5 months may have been underpowered. Thus, replication studies will be essential to establish the reliability of our results. Importantly, future research should aim for samples with enough power to model multiple independent and dependent variables within a single structural equation model to better control for shared variance between depression and post-traumatic stress symptoms and social support predictors. Moreover, we were not powered to test 3-way interactions. In particular, we were not able to consider whether other factors known to influence risk for symptoms, such as gender, also moderated the interaction between social support and cohort in the prediction of symptoms. Further, we were not able to measure different forms of social interaction (e.g., in-person versus virtual), which may have differed by cohort. One limitation related to the matching of cohorts was that we did not have injury severity scores (ISS) for 6 participants—5 from the COVID-19 cohort—which may have impacted the results of the matching process. Finally, our analyses did not consider other dimensions of psychological adjustment, namely indices of psychological well-being or daily functioning, both of which would be relevant to this question. Despite these limitations, the rigor of the matched cohort longitudinal design makes the findings novel and impactful, and certainly warrants replication and expansion.

In sum, this investigation aimed to examine how social support may have deviated in form and/or function during the COVID-19 pandemic in United States adults recovering from potentially traumatic events. By applying a matched case–control design, we were able to test if reports of social support differed between patients who either experienced trauma before or during the pandemic. The matched design allowed us to isolate effects that could otherwise have been accounted for by key factors known to impact adjustment during the pandemic, namely economic status and gender. Moreover, we were able to test how social support operated in the service of psychological adjustment and found some evidence that social support, although reported at equivalent levels across cohorts, was less protective when traumatic events occurred within the context of the COVID-19 pandemic when compared to pre-pandemic. In sum, we found compelling evidence that the process of psychological adjustment to trauma may have shifted in part, due to pandemic-related impacts. These findings are novel and impactful, as they shed light on the specific impacts of the COVID-19 pandemic on psychological processes, but also test the role of broader, perhaps more chronic, environmental stressors on adjustment to an acute traumatic event.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary material.

The studies involving human participants were reviewed and approved by Kent State University Institutional Review Board. The patients/participants provided their written informed consent to participate in this study.

BM, EG, BB, and KC contributed to the conceptualization and methodology, and wrote the original draft. BM and KC did the formal data analyses. KC, RG, FM, AM, JG, and DD contributed to the acquisition of resources and funding. BM did the data curation and created the tables and figures. KC oversaw supervision and project administration. All authors contributed to the investigation, review, and editing of the manuscript.

This study was supported by a National Institutes of Health grant number (no. R01 MH113622) to KC.

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.