Coping behaviors and mental health during the COVID-19 pandemic: situational and habitual effects on mental distress with hybrid model analysis

Abstract

Background

The COVID-19 pandemic brought about widespread mental health challenges. Yet, its mental health impacts vary depending on the coping strategies people adopt to manage stress. The COVID-19 pandemic, with its rapidly changing circumstances, provides an opportune context to examine how different coping behaviors are linked to mental distress. This study explores four coping approaches—social connection, substance use, social media use, and relaxation techniques—to assess how they relate to mental distress over time during the pandemic at both situational (within-person) and habitual (between-person) levels.

Methods

Using a hybrid model, this study analyzed longitudinal data from the Understanding American Study (UAS), spanning from April 2020 to June 2021. This approach allowed for differentiating between within-person (how changes in an individual’s coping behaviors related to their own mental health over time) and between-person effects (how individuals with different coping behaviors, on average, compare in terms of mental health outcomes).

Results

Adaptive coping behaviors like social connection and relaxation coping were linked to lower mental distress at both within-person and between-person levels, with stronger between-person effects. Substance use and social media use were associated with increased mental distress, suggesting potential risks in their use, especially when these behaviors become habitual. Roughly 20–52.39% of the between-person effects of coping behaviors were explained by sociodemographic characteristics. Conclusion. By recognizing the value of stable, adaptive coping habits—while also accounting for how situational changes impact well-being—policymakers and practitioners can craft more effective interventions to foster mental resilience during public health crises.

Supplementary Information

The online version contains supplementary material available at 10.1007/s44192-025-00263-w.

Introduction

The COVID-19 pandemic has caused widespread mental distress, but not everyone has been affected in the same way [1, 2]. One important reason for this variation lies in the coping strategies people adopt to manage stress. Generally, coping methods are split into adaptive approaches (e.g., social connection, relaxation) and maladaptive ones (e.g., substance use, avoidance) [3]. Past research has shown that adaptive coping tends to protect mental health, while maladaptive ones often exacerbate distress [4]. Yet the pandemic introduced unique challenges that complicated these categories. For instance, while connecting with others usually helps reduce loneliness, it could also heighten distress if emotional support was unavailable [5, 6]. Likewise, although spending more time with family together is typically considered beneficial for mental health, it sometimes leads to heightened interpersonal conflict, contributing to a “psychosocial stress contagion” effect during the pandemic, where stress spreads within family units [7]. Social media, too, emerged as a double-edged sword—offering connection and information during isolation, but also exposing users to misinformation and harmful social comparisons that can intensify psychological strain [8–10]. These complexities show that simple labels like “adaptive” or “maladaptive” don’t always capture how coping works in real-world contexts. What helps or harms may often depend on the context and the individual.

Another domain that warrants special attention during the pandemic is substance use, commonly recognized as a maladaptive coping strategy. Studies have documented a rise in alcohol and drug consumption as people sought ways to manage isolation, boredom, and heightened anxiety during the pandemic [11–14]. These changes were particularly pronounced among those experiencing economic hardship or limited access to mental health resources [12, 14]. While substance use may provide temporary relief, it often leads to long-term negative consequences, such as impaired emotional regulation, greater psychological distress, and a diminished capacity to engage in more adaptive coping behaviors [15]. What makes this trend particularly concerning in the context of the pandemic is not just the increase in substance use, but the way in which chronic stress and constrained coping options may have shifted people toward more harmful strategies over time. Unlike one-time stress events, the prolonged and evolving nature of the pandemic created conditions where maladaptive behaviors could become entrenched. This underscores the need to go beyond static categorizations of coping strategies and consider how individuals’ responses may shift based on evolving circumstances and available resources.

In contrast to substance use, physical activity is typically regarded as an adaptive coping strategy. As Sher and Wu [16] observed, many people initially engaged in exercise for stress relief in the early stages of the pandemic, but it became harder to maintain these habits as time went on and circumstances changed. These insights suggest the need to distinguish between situational coping —responses to immediate stressors, such as those triggered by the pandemic—and habitual coping, which reflects more stable, ingrained behaviors. This becomes particularly relevant in prolonged crises like the pandemic, where people adjust their coping behaviors as stressors evolve. Despite this, most research on pandemic-related coping has been cross-sectional, providing only snapshots rather than tracking how coping unfolds over time.

To address this gap, the present study employs a hybrid modeling approach to differentiate between situational and habitual coping strategies and examines how each relates to mental distress. This distinction is grounded in Lazarus and Folkman’s [3] transactional model of stress and coping, which views coping as a dynamic process that unfolds in response to person-environment interactions. In their framework, coping involves cognitive and behavioral efforts to manage internal or external demands that are appraised as taxing or exceeding one’s resources. In contrast, other scholars have emphasized the more trait-like, dispositional aspect of coping, suggesting it remains relatively stable over time [17]. Moss and Holahan [18] thus proposed an integrative view in which coping encompasses both adaptive, momentary responses and enduring, habitual tendencies.

Building upon this integrative framework, the present study models coping as both a within-person (situational) and a between-person (habitual process). Situational coping refers to how individuals adapt their coping behaviors in response to momentary or contextual stressors —captured in the model as time-varying deviations from their typical patterns. Habitual coping reflects general or trait-like use of particular coping behaviors over time, often shaped by socialization, personality, or broader structural conditions. By separating within-person from between-person variance, the hybrid model allows for a more nuanced understanding of how both short-term fluctuations and long-term tendencies in coping are associated with mental distress.

This study also explores whether sociodemographic characteristics (e.g., age, gender, education, income) are associated with coping behavior. Prior work has shown that access to resources can influence the adoption of certain coping behaviors [e.g.,11,12]]. Examining the structural and demographic factors is expected to provide crucial context for understanding individual differences in coping—and by extension, disparities in mental health outcomes during the pandemic.

Methods

Data

This study drew on de-identified, publicly available data from the Understanding America Study (UAS) COVID-19 Survey, a nationally representative sample of U.S. adults [19]. Internet service and tablets were provided to participants without Internet access to ensure inclusivity [20]. By May 2024, 33 biweekly waves of data had been collected. For the present analyses, data from 26 waves were used, specifically from April 2020 (Wave 4) through June 2021 (Wave 29). The first three waves were excluded because coping behavior questions were not available. The analytic sample consists of 157,319–157,377 observations from 8,272 participants. Of those, 39.5% (n = 3,267) had completed data across all 26 waves, while 60.5% (n = 5,004) contributed at least some data. The median number of waves completed was 24, with 75% completing at least 24 waves. Even the lower quartile (25th percentile) completed 13 waves, reflecting a high panel retention rate. Missingness at the within-person level was also low. On average, fewer than 5% of participants had more than 10% missing data on any of the key variables. The PHQ-4 scale, for instance, was missing in 2.1% of observed person-wave records. The average within-person missingness across waves was 1.7% for PHQ-4 and under 3% for all coping variables. Missing data were addressed using full-information maximum likelihood (FIML) in the analyses, which leverages all available data under the assumption of missing at random (MAR).

Measures

Mental health was measured with the 4-item Patient Health Questionnaire (PHQ-4), a validated instrument for anxiety and depression [21, 22]. Participants were asked to report how often they’ve felt nervous, worried, depressed, or uninterested in activities over the past two weeks, using a 4-point scale (1 = “not at all” to 4 = “nearly every day”). Total scores range from 0 to 12, with higher scores indicating greater distress. To evaluate the reliability of this scale in a longitudinal context, both between-person and within-person consistency were assessed, following the approach recommended by Nezlek [23]. The PHQ-4 demonstrated strong internal consistency across both levels, with Cronbach’s alpha estimates of α_between = 0.87 and α_within = 0.82, respectively.

Participants reported how many days in the past week they engaged in 13 different activities, captured across multiple survey waves. Exploratory factor analysis produced three main coping dimensions: (1) social connection (in-person, phone, or text interactions) (α_between = 0.86 and α_within = 0.82), (2) relaxation (exercise, meditation) (α_between = 0.58 and α_within = 0.34), and (3) substance use (drugs, cigarettes, alcohol, cannabis, vaping) (α_between = 0.40 and α_within = 0.37). Full items of each coping category are listed in the Supplementary Information (Table S1).Each dimension was calculated as the mean number of days (0–7) per week.

These dimensions were derived empirically and represent behavioral groupings rather than reflective latent constructs. Internal consistency was high for social connection (α_between = 0.86 and α_within = 0.82), but lower for relaxation (α_between = 0.58 and α_within = 0.34) and substance use coping (α_between = 0.40 and α_within = 0.37). These lower reliability values likely reflect the episodic and heterogeneous nature of these behaviors, which do not necessarily co-occur consistently either across people or within individuals over time. Despite this, these factors were retained to capture broader behavioral tendencies relevant to coping.

Social media use emerged as a separate coping factor, given its positive correlation with mental distress, whereas other social coping connection items were negatively correlated. This divergence aligns with prior research and theoretical frameworks such as compensatory internet use theory [24], which posits that individuals may turn to digital media to manage negative emotions, often with mixed or adverse psychological effects [8, 25]. Based on empirical and theoretical rationale, social media use was analyzed as a distinct coping dimension in the primary models. To ensure the robustness of this decision, a parallel model was also estimated in which social media use was included within the social connection factor. Results from this specification are presented in the Supplementary Information (see Table S2).

Finally, the models controlled for the following covariates: age, gender, educational level (1–16), race/ethnicity, marital status (married), and household income (scale, 1–16).

Analytic strategy

The present study utilized a hybrid model to estimate within-person and between-person effects of coping behaviors on mental distress [26]. This method is particularly well-suited for longitudinal panel data, as it allows for the decomposition of time-varying predictors into two components: (1) within-person effects, which capture how fluctuations in an individual’s coping behavior over time relate to changes in their own distress, and (2) between-person effects, which assess whether individuals who, on average, engage more frequently in certain coping strategies report different levels of distress than others.

Although both hybrid models and hierarchical linear modeling (HLM) can be used to analyze longitudinal data, the hybrid model offers a key advantage for the present study. HLM typically assumes that random effects are uncorrelated with covariates and often requires strong distributional assumptions [27, 28]. Moreover, unless random slopes are explicitly specified, HLM may conflate within-person (momentary) and between-person (trait-like) effects, making interpretation less straightforward [26]. In contrast, the hybrid model explicitly separates these two sources of variance by decomposing time-varying predictors into within- and between-person components. This decomposition enables direct testing of whether coping behaviors exert differential effects on mental distress within individuals over time versus across individuals. Such a distinction is central to the aims of the current study, which seeks to examine both intra-individual variability and inter-individual differences in coping and mental distress. The hybrid model was estimated using the xthybrid command in Stata 18.5, which automates this decomposition process [26, 29].

The model is specified as follows:

  

Where Y_it is the mental distress for individual i at time t. X_it−_i represents the within-person deviation from the individual’s own mean on the coping behavior. Coing variables were person-mean centered to isolate within-person variation, consistent with best practices in hybrid modeling. β₁ and β₂ represent within-person and between-person effects, respectively. Z_iγ denotes time-invariant covariates (e.g., gender, race/ethnicity) and their associated fixed effects. µ_0i and µ_1i denote random intercept and slope, respectively. To test whether the within- and between-person effects of coping behaviors significantly differed, the Wald test of coefficient equality was performed. It should be noted that in all models, coping variables were standardized prior to estimation to facilitate interpretation and comparability. Using raw scores of the coping behavior variables produces consistent results. Accordingly, all coefficients for coping behaviors can be interpreted as the expected change in mental distress associated with a one-standard-deviation increase in the coping variable, at the within-person or between-person level (Table 1).

Table 1.

Weighted summary statistics of variables in analyses

	Mean/% (SD)
PHQ-4 scores (0–12)	1.73 (2.82)
Social connection coping (0–7)	4.31(2.04)
Social media use coping (0–7)	4.04(2.97)
Relaxation coping (0–7)	2.94 (1.74)
Substance use (0–7)	0.61 (0.91)
Age	48.67 (16.58)
Gender
Male	48.30%
Female	51.70%
Race/ethnicity
White	62.60%
Black	12.00%
Hispanic	15.20%
Asian	5.40%
Pacific Islanders	0.20%
Native Americans	0.80%
Mixed race	3.80%
Married	54.60%
Household income (scale, 1–16)	10.89 (4.297)
Highest education (level, 1–16)	10.82 (2.46)

Four families of models corresponding to each coping dimension were analyzed: social connection, relaxation, substance use, and social media use. Each domain included two nested models: The first model included only the within-person and between-person components of the coping behavior. The second model added sociodemographic characteristics to examine the extent to which these characteristics explained variance in between-person effects. This analytic structure is reflected in Table 2, which presents results for eight models labeled M1a-M4b. M1a and M1b correspond to social connection coping, M2a and M2b to relaxation coping, M3a and M3b to substance use coping, and M4a and M4b to social media use coping. In each pair, Model “a” includes coping effects only, and Model “b” includes coping plus covariates.

Table 2.

Hybrid models estimating between-person and within-person coping behaviors and mental distress associations

	Social Connection		Substance use		Social Media use		Relaxation Coping
	M1a	M1b	M2a	M2b	M3a	M3b	M4a	M4b
	Coef.	Coef.	Coef.	Coef.	Coef.	Coef.	Coef.	Coef.
Within-social connection	-0.035**	-0.032**
	(0.011)	(0.011)
Between-social connection	-0.172**	-0.122**
	(0.032)	(0.032)
Within-substance use			0.115***	0.107***
			(0.017)	(0.017)
Between-substance use			0.522***	0.455***
			(0.029)	(0.028)
Within-social media					0.080***	0.073***
					(0.012)	(0.012)
Between-social media					0.292***	0.139***
					(0.031)	(0.031)
Within-relaxation							-0.119***	-0.130***
							(0.012)	(0.012)
Between-relaxation Covariates							-0.485***	-0.308***
							(0.032)	(0.032)
Age		-0.030**		-0.027**		-0.028***		-0.025***
		(0.002)		(0.002)		-0.002		(0.002)
Male		-0.491**		-0.550**		-0.407***		-0.448**
		(0.055)		(0.054)		(0.055)		(0.054)
Race/ethnicity (ref. white)
Black		-1.031**		-0.915*		-0.964***		-1.010***
		(0.101)		(0.099)		(0.101)		(0.100)
Hispanic		-0.389**		-0.181*		-0.328***		-0.426***
		(0.081)		(0.081)		(0.081)		(0.081)
Asian		-0.003		0.236*		0.053		-0.006
		(0.119)		(0.118)		(0.119)		(0.119)
Pacific Islander		-0.091		0.020		-0.088		-0.125
		(0.312)		(0.307)		(0.312)		(0.310)
Native American		-0.199		-0.047		-0.157		-0.207
		(0.180)		(0.178)		(0.180)		(0.179)
Mixed		-0.001		-0.005		0.024		0.025
		(0.117)		(0.115)		(0117)		(0.116)
Married		-0.514**		-0.439***		-0.515***		-0.529***
		(0.058)		(0.057)		(0.058)		(0.058)
Household income		-0.098**		-0.096***		-0.107**		-0.103***
		(0.008)		(0.008)		(0.008)		(0.008)
Education		-0.044***		-0.054***		-0.041**		-0.053***
		(0.012)		(0.011)		(0.011)		(0.011)
Constant	1.932**	4.664**	1.921**	4.300**	1.926**	4.626**	1.922**	4.321***
	(0.027)	(0.166)	(0.027)	(0.165)	(0.027)	(0.166)	(0.027)	(0.170)
Random-effects parameters
Random slope	0.348***	0.348***	0.598***	0.588***	0.320***	0.317***	0.457***	0.458***
	(0.012)	(0.012)	(0.027)	(0.027)	(0.014)	(0.014)	(0.016)	(0.016)
Random intercept	6.062***	5.403***	5.853***	5.247***	6.013***	5.398***	5.908***	5.353***
	(0.098)	(0.087)	(0.094)	(0.085)	(0.097)	(0.087)	(0.095)	(0.087)
Variance	2.015***	2.010***	2.027***	2.023***	2.046***	2.042***	2.004***	1.998***
	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)	(0.007)	(0.007)
N (Individual-wave observations)	157,370	157,370	157,377	157,377	157,363	157,363	157,319	157,319
N (Individual observations)	8,272	8,272	8,272	8,272	8,272	8,272	8,272	8,272
Ma-Mb % ΔWP coping coef.		8.57%		6.95%		8.75%		9.24%
Ma-Mb % ΔBP coping coef.		20.06%		12.45%		52.39%		36.49%

Note: WP refers to within-person, while BP refers to between-person

Robust standard errors are in parentheses

***p <.001 ** p <.01, * p <.05 (Two-tailed test)

To determine the appropriate modeling strategy, both combined and separate hybrid models were initially tested. While the random intercept version of the combined model (including all coping strategies simultaneously) converged, the random slope version—essential for estimating individual variability in within-person effects—failed to converge. This is likely due to model complexity and limited within-person variability when included together. In contrast, separate hybrid models for each coping domain with random slopes not only converged successfully but also showed better model fit (e.g., for the social connection model, AIC = 594,700; BIC = 594,800) than the combined model (AIC = 596,400; BIC = 596,500). Modeling each coping strategy independently allowed for clearer interpretation of the distinct within- and between-person effects associated with each coping behavior, avoiding potential issues of overlapping variance or suppressor effects that can arise when related predictors are modeled simultaneously. Given these statistical and conceptual considerations—model convergence, improved fit, and clarity of interpretation—each coping strategy was modeled independently.

Finally, to explore potential drivers of between-person differences in coping behaviors themselves, a set of mixed-effects regression analyses was conducted, regressing each coping strategy on sociodemographic characteristics. These analyses help contextualize the observed associations between coping and mental distress by identifying the structural or demographic influences. Survey weights and full-information maximum likelihood with robust standard errors were applied in Stata18.0 for all analyses to ensure population representativeness.

Results

Table 1 presents the weighted summary statistics of variables. The mean PHQ-4 score was 1.73 (SD = 2.81). On average, participants reported engaging in social connection and social media activities for 4 days per week, relaxation coping for 2.4 days, and substance use coping for 0.61 days. Distributions of the between-person averages and within-person deviations for mental distress and coping behaviors are presented in Supplementary Figures S1 and S2.

Table 2 summarizes the results of the hybrid models on coping behaviors and mental distress. Social connection coping was negatively associated with distress at both within-person and between-person levels. A one-standard-deviation increase in within-person social connection was associated with a statistically significant decrease in distress (Model 1a: β = -0.035; Model 1b: β = -0.032, p <.01), while a one-standard-deviation increase at the between-person level was associated with an even greater reduction in distress (Model 1a: β = -0.172; Model 1b: β = -0.122, p <.01). Substance use was associated with increased distress both within-person (Model 2a: β = 0.115; Model 2b: β = 0.107, p <.01) and between-person (Model 2a: β = 0.522; Model 2b: β = 0.455, p <.01). Social media coping was positively associated with distress both within-person (Model 3a: β = 0.080; Model 3b: β = 0.073, p <.01) and between-person (Model 3a: β = 0.292; Model 3b: β = 0.139, p <.01). Relaxation coping reduced distress at both levels, with within-person increases (Model 4a: β = -0.119; Model 4b: β = -0.130, p <.01) and between-person differences (Model 4a: β = -0.485; Model 4b: β = -0.308, p <.01) associated with lower distress.

INSERT Table 2 Hybrid models estimating between-person and within-person coping behaviors and mental distress associations.

Sociodemographic covariates included in the hybrid models (Table 2) were also significantly associated with mental distress. Higher educational attainment and greater household income were associated with lower levels of distress across all models. Being married and older age also predicted lower distress. Interestingly, however, non-White respondents—particularly Black and Hispanic individuals—tended to report lower levels of distress relative to White respondents, after adjusting for coping strategies.

In summary, both within-person changes and between-person differences in all coping behaviors were significantly associated with anxiety and depression, with between-person effects consistently greater. Formal statistical tests with the Wald test confirmed that between-person effects were significantly larger than within-person effects for social coping (χ²(1) = 7.24, p <.01), substance use coping (χ²(1) = 109.08, p <.001), social media coping (χ²(1) = 4.18, p <.05), and relaxation coping (χ²(1) = 26.4, p <.001).

As a robustness check, a parallel model was conducted in which social media use was included within the social connection coping dimension, consistent with the original factor analytic structure (see Supplementary Table S1). This specification yielded slightly larger within-person effects (β = -0.046) and attenuated between-person effects (β = -0.119), compared to the primary model (Models 1b in Table 2) that excluded social media use from the social connection coping, but the direction and statistical significance remained consistent across both specifications. This suggests that excluding social media slightly strengthens the association between habitual social connection and lower distress, but the core findings are robust.

Time-invariant characteristics like gender, race, and economic status together explained a significant portion of the between-person effects in Models 1b–4b. As reported in the bottom rows of Table 2, between-person effects for social media and relaxation coping decreased by 20–52%, while substance use coping decreased by 12.45%. To further examine these patterns, a series of mixed-effects regressions were performed to examine the association between each coping behavior and sociodemographic factors. As reported in Table 3, the coefficient for age-estimating social media use was substantially larger (b = -0.028, p <.001) than for other coping strategies (e.g., social connection: b = -0.001; relaxation: b = 0.012). A formal test using seemingly unrelated estimation (SUEST) showed that the coefficient for age predicting social media use was significantly larger than for social connection coping (χ²(1) = 8296.86, p <.001) and relaxation coping (χ²(1) = 21700.33, p <.001). Gender differences were also evident. Men were more likely than women to rely on substance use coping (b = 0.159, p <.001) and less likely to use social connection (b = -0.235, p <.001) or social media use (b = -0.307, p <.001), indicating distinct gendered patterns in coping behavior. A follow-up SUEST analysis confirmed that the gender effect on substance use coping was significantly greater than its effect on social media use (χ²(1) = 7128.84, p <.001) and social connection coping (χ²(1) = 4274.07, p <.001), further highlighting distinct gendered patterns in coping behavior.

Table 3.

Mixed-effects regression estimates of coping behaviors by sociodemographic variables

	Social Connection	Substance use	Social media	Relaxation	Social Connection Social Media
Age	-0.001	-0.008***	-0.028***	0.012***	-0.003***
	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)
Male	-0.235***	0.159***	-0.307***	0.030	-0.259***
	(0.022)	(0.032)	(0.024)	(0.021)	(0.022)
Race/ethnicity (ref. white)
Black	-0.331***	-0.169***	-0.318***	-0.062	-0.381***
	(0.037)	(0.048)	(0.034)	(0.038)	(0.038)
Hispanic	-0.295***	-0.321***	-0.232***	-0.268***	-0.362***
	(0.032)	(0.040)	(0.031)	(0.033)	(0.035)
Asian	-0.279***	-0.512***	-0.287***	-0.109**	-0.316***
	(0.055)	(0.064)	(0.051)	(0.038)	(0.055)
Pacific Islander	-0.165	-0.242	-0.357**	-0.043	-0.252
	(0.154)	(0.125)	(0.121)	(0.125)	(0.159)
Native American	-0.288***	-0.293***	-0.204***	-0.075	-0.307***
	(0.065)	(0.063)	(0.058)	(0.071)	(0.068)
Mixed	-0.045	0.056	-0.149*	0.034	-0.083
	(0.039)	(0.055)	(0.063)	(0.033)	(0.043)
Married	0.051	-0.146***	0.007	-0.003	0.026
	(0.038)	(0.028)	(0.024)	(0.023)	(0.036)
Household income	0.012***	0.000	0.003	-0.003	0.009**
	(0.003)	(0.003)	(0.002)	(0.003)	(0.003)
Education	0.036***	-0.038***	0.014**	0.035***	0.039***
	(0.005)	(0.006)	(0.005)	(0.005)	(0.005)
Constant	-0.337***	0.944***	0.886**	-0.952***	-0.220***
	(0.071)	(0.082)	(0.067)	(0.070)	(0.072)
Random-effects parameters
Random intercept	0.643***	0.783***	0.664***	0.603***	0.669***
	(0.009)	(0.022)	(0.008)	(0.010)	(0.010)
Variance	0.283***	0.210***	0.215***	0.247***	0.254***
	(0.005)	(0.012)	(0.004)	(0.005)	(0.005)

***p <.001 ** p <.01, * p <.05 (Two-tailed test)

Racial and ethnic differences emerged in all domains. Black and Hispanic respondents consistently reported lower use of social connection and social media than White respondents. Asian and Native American respondents also reported reduced use of social connections, social media, and substance use coping. Socioeconomic status (SES), particularly education and income, significantly shaped coping strategies. Individuals with higher education levels were more likely to turn to social connection (b = 0.036, p <.001) and relaxation coping (b = 0.035, p <.001), but less likely to rely on substance use (b= -0.038, p <.001). Those with greater household income also reported more social connection use (b = 0.012, p <.01).

INSERT Table 3 Mixed effects models estimating each coping behavior by sociodemographic characteristics.

Discussion

This study examined the relationship between coping strategies—social connection, substance use, social media use, and relaxation—and mental distress during the COVID-19 pandemic, distinguishing between situational (within-person) and habitual (between-person) effects using a nationally representative longitudinal sample. The longitudinal analysis with the hybrid model revealed that how people cope, both in the short term (within-person) and as a habitual strategy (between-person), were significantly related to mental distress during the COVID-19 pandemic.

Social connection and relaxation strategies, often viewed as adaptive, showed robust negative associations with anxiety and depression. Social connection likely provided emotional support and alleviated feelings of isolation [30], while relaxation methods were associated with lower distress levels, potentially by providing moments of calm or tension reduction [31]. These associations suggest that such behaviors may play a beneficial role in managing stress during high-uncertainty periods.

In contrast, substance use was associated with higher levels of distress, probably because it can disrupt sleep, impair cognitive functioning, and increase physiological stress responses [15]. These effects may be particularly pronounced during prolonged crises, such as the pandemic, when access to healthier coping alternatives is limited. Social media use was also positively associated with increased distress, possibly due to the amplification of stressors such as misinformation or social comparison, despite its potential to facilitate connection [32].

The comparison between the primary model and the robustness specification reveals that including social media in the social connection factor generated a slight attenuation of the between-person effect. While the difference is small, this shift is directionally consistent with the idea that social media may obscure some of the beneficial effects of traditional forms of social connection. This aligns with theoretical frameworks such as compensatory internet use theory [24] and empirical research [8, 25], which highlighted that social media may have more ambivalent or even adverse mental health consequences compared to in-person or direct social contact. As such, separating social media as a distinct coping domain remains defensible. At the same time, the robustness check lends additional support to the main findings by demonstrating that results are not sensitive to the way social media use is categorized.

The finding that between-person effects surpass within-person effects suggests that ingrained, long-term coping habits have a notable impact on mental health. Individuals who habitually rely on adaptive coping show overall lower levels of distress, whereas those leaning on substance use or social media coping experience consistently higher distress. These results corroborate with prior research showing that enduring patterns of coping and emotional regulation are key predictors of long-term mental health outcomes [33]. At the same time, the significance of within-person effects also underscores that even day-to-day changes in coping behavior can meaningfully impact mental well-being [18]. Together, this highlights the value of interventions that not only promote adaptive coping habits but also encourage flexibility and responsiveness to situational demands.

Sociodemographic characteristics accounted for a substantial portion of the between-person variance in coping behaviors and their associations with mental distress, echoing broader evidence that structural factors, like economic and educational disparities, shape the ability to adopt and maintain healthy coping strategies [16]. Among these, age stood out: the association between age and social media coping was significantly stronger than for other coping domains, suggesting that generational patterns may be particularly influential in shaping reliance on social media, especially during a period of heightened digital engagement like the COVID-19 pandemic. While structural inequities remain relevant, age-specific media habits and digital literacy may offer additional explanatory power.

Individuals with higher socioeconomic status (i.e., higher income and education levels) were more likely to engage in adaptive coping strategies, including relaxation and social connection, and less likely to rely on maladaptive behaviors such as substance use [34]. This socioeconomic gradient in coping may account for the inverse association we observed between SES and mental distress, which aligns with prior findings [35, 36]. Interestingly, this group also reported greater use of social media as a coping mechanism. This may reflect broader access to digital resources or distinct patterns of media engagement. While this raises questions about the mixed role of social media, it is possible that higher-SES individuals benefit from a buffering effect due to their concurrent use of more protective, adaptive coping strategies.

Racial and ethnic differences in coping were also evident in the findings, aligning with the stress and health disparities framework, which posits that coping behaviors are critical pathways through which structural and social inequalities impact mental health outcomes [37]. In this study, non-white respondents reported lower overall in all types of coping behaviors, including both adaptive (e.g., social connection, relaxation) and maladaptive (e.g., substance use) strategies, compared to White respondents. At the same time, these groups also reported lower levels of distress in the primary hybrid models. This finding may appear paradoxical but resonates with prior research highlighting complex interactions between culture, resilience, and mental health, such as the so-called Black -White and Hispanic mental health paradox, in which Black Americans and Hispanic Americans report lower distress despite greater exposure to structural adversity, including during the pandemic [37–40]. Several potential explanations may help contextualize this pattern. Culturally rooted norms around emotional expression and self-reliance may shape how stress is internalized or expressed, leading to underreporting of distress or the use of coping strategies not captured by survey items [37, 38, 41]. Structural barriers—including limited access to culturally responsive mental health services, digital infrastructure, or discretionary time—may constrain the ability to engage in the coping behaviors measured in this study [42]. These findings also invite a reframing of how distress and coping are understood in racially marginalized communities. Rather than viewing lower reported distress solely through a deficit lens, such as underreporting or limited coping resources, it is equally important to consider the cultural strengths, community resilience, and adaptive strategies that have long supported emotional well-being in these populations. For instance, strong kinship ties, faith-based practices, and collective cultural values may promote psychological resilience in ways that are not fully captured by standard survey instruments or clinical assessments [37].

In contrast, the smaller reduction in mental distress associated with the between-person effect of substance use (12.45%) may reflect its universal risk across populations. While coping behaviors are often shaped by social position, the psychological consequences of maladaptive strategies like substance use appear to be broadly detrimental, regardless of one’s background.

These findings should be interpreted with several limitations in mind. First, all measures were self-reported, which introduces the potential for reporting bias. In addition, the unique context of the COVID-19 pandemic may limit the generalizability of these findings to other types of social or environmental stressors. Future research could use objective measures to explore coping strategies in various contexts and investigate the differential effects of social media behaviors, such as passive scrolling versus active engagement, on mental health. Although the hybrid model helps differentiate within- and between-person effects, causality cannot be firmly established as unmeasured confounders or bidirectional relationships may influence the findings. Moreover, some of the variance attributed to between-person effects may reflect structural inequalities—such as unequal access to mental health care, digital resources, or safe recreational spaces—rather than individual-level coping tendencies alone.

While sociodemographic covariates explained much of the between-person effects in coping behaviors, some residual variance likely reflects unmeasured factors not captured in this study. These may include cultural norms, structural barriers, or environmental conditions that influence coping behavior but were not directly assessed. Similarly, though within-person effects are assumed to be driven by situational changes in coping behaviors in the model, we cannot rule out the possibility that short-term fluctuations or transient contextual factors not captured in the data may also come into play. Future studies would benefit from considering cultural and structural variables more directly to detangle environmental constraints from personal coping patterns.

Another limitation concerns how coping strategies were operationalized. The factor-analytic structure used to define coping domains was shaped by the available survey items, which may not fully capture the breadth of real-world coping behavior. For instance, our maladaptive coping focused exclusively on substance use and did not capture other relevant strategies such as avoidance, rumination, or self-blame. Likewise, relaxation coping, composed of exercise and meditation, may not fully reflect individuals’ relaxation experiences, some of which may overlap with social engagement (e.g., spending time with loved ones). These conceptual overlaps point to a potential construct ambiguity, especially in lived experiences of coping, calling for need for more nuanced, multidimensional assessments of coping in future research.

Conclusion

This study highlights the importance of distinguishing between habitual and situational coping strategies in understanding mental distress during the COVID-19 pandemic. By applying a hybrid model to a nationally representative longitudinal dataset, the study revealed that both between-person differences and within-person changes in coping behaviors are significantly associated with mental health outcomes. The consistent finding that between-person (habitual) effects are stronger than within-person (situational) effects suggests that enduring coping tendencies—shaped by personal history and structural conditions—play a significant role in shaping distress than short-term fluctuations alone. While sociodemographic factors explain a meaningful portion of these differences, additional structural and cultural variables may further account for variance. These insights underscore the value of promoting stable, adaptive coping routines while also recognizing the power of situational adaptations. Public health efforts should be designed to support both the development of long-term coping habits and the capacity for flexible coping responses during crises.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

S.K. conceived and designed the study, performed the data collection and analysis, prepared the figures, wrote the manuscript, and approved the final version of the work for publication.

Funding

The author(s) received no financial support for the research.

Data availability

The data used in this study were obtained from the Understanding America Study (UAS), a nationally representative longitudinal panel maintained by the University of Southern California’s Dornsife Center for Economic and Social Research. The UAS data are publicly available to registered researchers upon request at https://uasdata.usc.edu. Access to the data requires approval from the UAS team and adherence to their data use agreement policies.

Declarations

Conflict of interest

No potential conflict of interest was reported by the author(s).

Ethical approval

This study utilized secondary data from the Understanding America Study (UAS), which was originally collected by the University of Southern California’s Center for Economic and Social Research. The UAS adheres to ethical guidelines and was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board (IRB) of the University of Southern California (IRB Protocol No. 22-030-1044). This study utilized de-identified secondary data from the UAS. Since no direct interaction with participants occurred and the data were anonymized, the secondary analysis conducted for this study was deemed exempt from further ethical review by the author’s institution.