1 INTRODUCTION

The 2019 novel coronavirus (COVID-19) pandemic has led to unprecedented loss and change. Since December 2019, the virus has caused over 5 million documented deaths globally, including almost 800,000 in the United States alone (Johns Hopkins University, 2021). To mitigate the spread, public health officials recommended mask wearing, physical distancing and limiting operations of non-essential businesses. While protecting physical health, quarantine and isolation are known to have especially deleterious impacts on mental health (Henssler et al., 2021), and an emerging body of literature documents the pandemic's negative effects (e.g. Pfefferbaum & North, 2020). However, there is a specific need for research on psychotherapy clients' functioning in the pandemic context, as those with psychiatric diagnoses are at greater risk of negative psychological impacts during epidemics, natural disasters and community threats (Esterwood & Saeed, 2020; Yao et al., 2020).

Positive mental health is more than the absence of symptoms, however. Well-being is a multidimensional construct that can relate to mental health in complex ways (O'Connor et al., 2012, 2015). Further, individual experiences of the pandemic vary not only in functioning as measured by mental health and well-being indicators but also in what feels uniquely challenging, beneficial or even conducive to new learning and growth. With an eye towards capacities for resilience, efforts to identify how people have fared during this public health crisis should include nuanced attention to diverse sets of experiences. To examine psychotherapy clients' functioning and adaptation in the ever-changing pandemic context, this study employed a concurrent triangulation mixed-methods design to identify subgroups of clients based on mental health symptoms (hereafter, symptoms) and indicators of well-being, and to assess clients' perceived challenges, benefits and areas of self-learning related to the COVID-19 pandemic.

1.1 COVID-19 pandemic mental health effects

The current public health crisis has introduced unique stressors to daily life. Restrictions that protect the public have also created significant disruption, and changing government restrictions, dynamic scientific understanding and ongoing multiple waves of infection have contributed to an ever-changing ‘normal’. Research from the 2002–2004 SARS epidemic documented the vulnerability of some populations in times like these, noting that ‘while psychological consequences are widespread, not all individuals are affected universally’ (Douglas et al., 2009, p. 3). Studies conducted during previous outbreaks suggest that healthcare workers, people with prior mental health challenges, those with fewer social and economic resources or those who are impacted more directly (e.g. becoming sick, losing a loved one) are at heightened risk for chronic psychological distress (Esterwood & Saeed, 2020). Since the beginning of the pandemic, researchers and public health experts have raised alarm about rising mental health concerns, including depression, anxiety, traumatic stress, insomnia, substance abuse, intimate partner violence and suicidality (Pfefferbaum & North, 2020). A study conducted across 194 cities in China in early 2020 found that 54% of participants reported moderate to severe psychological impacts, with 29% describing moderate to severe anxiety. A meta-analysis of community-based studies conducted during the pandemic found depression to have a 25% prevalence rate (Bueno-Notivol et al., 2020). In particular, people with mental health vulnerabilities are likely to be ‘more substantially influenced by the emotional responses brought on by the COVID-19 epidemic’ (Yao et al., 2020, p. e21), including prolonged isolation imposed by stay-at-home orders, while potentially having less access to treatment (Simon et al., 2020). While a number of community studies have measured mental health in convenience samples, clinical populations are underrepresented.

Groups who experience systemic oppression, such as racial/ethnic minorities and economically disadvantaged individuals, are also at increased risk on multiple levels (Martin-Howard & Farmbry, 2020). The pandemic has disproportionately affected Black, Indigenous and People of Color (BIPOC) communities, where individuals are more likely to die from COVID-19 infection (Thakur et al., 2020). Tangible (e.g. financial security) and internal (e.g. loss of control over one's life) losses can create and perpetuate adversity, as vulnerable persons often have less access to resources needed to regain stable functioning (Hobfoll, 2011). Historically disadvantaged populations often have less access to healthcare, resulting in pre-existent conditions; rely on hourly/service positions, which have been hardest hit by economic downturn; reside in smaller living spaces and more population-dense areas, making physical distancing difficult; and may experience discrimination in seeking treatment. Women have also left the workforce at exponentially higher rates because of childcare needs (Madgavkar et al., 2020).

While some groups may be more vulnerable to negative effects, others may maintain stable functioning or improve or grow under the pandemic's circumstances, and a variety of risk and protective factors may influence these outcomes (Mancini, 2020). Despite their unique vulnerability, psychotherapy clients present as heterogeneous subgroups, and more nuanced conceptualisations of their experiences are necessary for effective treatment planning during and following this public health crisis.

1.2 Coping and adaptation in disaster situations

There has long been interest in how people fare in response to widespread disaster. While situations like the COVID-19 pandemic impose new circumstances and experiences, they also invoke a need for individual and systemic coping and adaptation. Evidence suggests that epidemics can adversely affect mental health and well-being, but considerably less is known about how people adapt to the unique effects of infectious disease outbreaks relative to natural and human-caused disasters. Coping strategies involve active ‘efforts to regulate emotions, behaviours, cognitions, psychophysiology, and environmental aspects’ in response to stress (Morales-Rodríguez & Pérez-Mármol, 2019, p. 2). Adaptive coping takes many forms. Emotion-focused coping can help manage difficult affect and reduce feelings of isolation, problem-focused coping is well-suited for challenges over which one has some control, and meaning-based coping may be optimal in situations of chronic hardship (Folkman & Greer, 2000). Research investigating responses to disasters have salient application to infectious disease outbreaks. Subjective appraisals (e.g. the pandemic is a threat versus a challenge) can affect responses to stress (e.g. paralysed with fear versus rallying internal resources to adapt; Lazarus & Folkman, 1984), and perspectives (e.g. perceived risk) can impact willingness to follow recommendations (e.g. wearing a mask; Rogers & Prentice-Dunn, 1997).

Researchers conceptualise adaptive responses to adversity in a variety of ways. Resilience is sometimes defined as bouncing back, that is, resuming and maintaining stable functioning despite adversity (Herrman et al., 2011). Smith (2020) emphasised the need to distinguish ‘bouncing back’ from ‘going beyond’ (p. 84), with the former referring to resilience and the latter thriving or post-traumatic growth. Bonanno and Diminich (2013) differentiated types of resilience, distinguished by whether the outcome was in response to chronic adversity or single-incident trauma. Such conceptualisations largely frame resilience as an outcome and predominantly a characteristic of an individual that can be directly (e.g. Smith, 2020) or indirectly assessed via measures of symptoms and/or indicators of well-being. However, defining resilience in this way fails to capture systemic factors and the variety of ways people adapt and respond to adverse situations.

Resilience has more recently been defined as a process. Walsh (2020) conceptualised resilience as a socio-ecological construct dynamically unfolding at family and community levels, and Ungar and Theron (2020) describe it as the interaction among ‘biological, psychological, social, and ecological systems … that help individuals to regain, sustain, or improve their mental well-being’ (p. 441). Such formulations draw attention to individuals' relational contexts and the ways in which individual-contextual interactions promote resilience. Wong (2011) suggested a process definition of resilience that went beyond bouncing back, and posited resilience as stemming from the dynamic interaction of an individual's internal capacities and contextual factors to foster positive growth and eudaimonic well-being, which he defined as ‘meaning plus virtue’ (p. 75). In fact, Wong suggested that meaning-making was central to resilience and introduced spirituality as potentially important to meaning-making. Like Wong (2011), Walsh (2020) has recognised the role of virtues (e.g. hope), meaning-making and spirituality in the process of resilience towards positive growth.

Fortitude may be an aspect of a process definition of resilience with particular salience during the pandemic. The term fortitude refers to ‘positive (fortigenic) appraisals of one's self, family, and external sources of support’ (Pretorius & Padmanabhanunni, 2021a, p. 159), and some definitions emphasise this capacity when ‘a positive outcome is not guaranteed (i.e. terminal illness) or may be difficult for a prolonged period (i.e. disaster impacted populations)’ (Van Tongeren et al., 2019, p. 7). Several studies have demonstrated the positive influence of fortitude on reduced mental health symptoms and greater subjective well-being (for a review, see Pretorius & Padmanabhanunni, 2021a), and emerging literature suggests the importance of fortitude during this public health crisis. Early in the pandemic, Pretorius and Padmanabhanunni (2021b) found that loneliness and anxiety indirectly predicted life satisfaction through fortitude in a sample of South African undergraduate students. In the context of community disasters, fortitude was found to indirectly predict lower symptoms through facilitating meaning-making (Zhang et al., 2021). These findings suggest an interplay of coping, symptoms and well-being, as fortitude appears to buffer the negative effects of adversity by helping people re-appraise ongoing hardship within the context of their worldview.

1.3 Mental health and well-being

Dual-factor models of flourishing that attend to symptoms and well-being represent an emerging trend in psychotherapy research (Fosha & Thoma, 2020; Jankowski et al., 2020; Rusk et al., 2018; Trompetter et al., 2017). Mental health care typically focuses on reducing symptoms and improving hedonic well-being, or a subjective state of positive affect; however, a diversity-sensitive approach to treatment prioritises holistic forms of emotional, psychological and social well-being that fit with clients' values and concerns. Eudaimonic well-being is a widely studied construct that includes psychological and social well-being dimensions, such as healthy relational connections, a sense of meaning and purpose in life, self-acceptance and contributions to community well-being. Keyes (2005) foundational work demonstrated that symptoms and well-being tend to be inversely related; however, these dimensions can constellate within individuals in complex ways (O'Connor et al., 2012, 2015). As an example, an individual with severe symptoms might have a sustaining life purpose, or someone may have few symptoms but lack a sense of meaning or social connection. Positive mental health, or flourishing, has been defined as the clinical goal of reducing symptoms and promoting greater subjective and eudaimonic well-being, but there is currently a lack of empirical evidence documenting treatment effectiveness for flourishing (Jankowski et al., 2020).

Research is beginning to investigate relationships between mental health, well-being and capacities for resilience during the pandemic. In a European study early in the COVID-19 outbreak, resilience was associated with less perceived stress and greater well-being, controlling for demographics and health vulnerabilities (Kavčič et al., 2020). The authors proposed that resilience may ‘inoculate individuals against elevated stress levels and decreased mental health, as well as weaken the negative impact of potential risk factors’ (Kavčič et al., 2020 p. 2). However, with the pandemic lasting more than a year, bouncing back from adversity may be less feasible, particularly in communities with pervasive social structural disadvantages. Zhang et al. (2020) found that spiritual fortitude buffered the association between resource loss and mental health distress during the pandemic. Relatedly, emerging evidence from Columbia and South Africa during lockdown suggests that positive religious coping and cultivating hope can support mental health by buffering the psychological toll of pandemic-related spiritual struggles (Captari et al., 2020). Further, Landi et al.'s (2020) study on coping during lockdown in Italy found psychological flexibility, openness to inner discomfort and engaging in values-based actions attenuated the negative effects of health-related anxiety on psychological functioning. These findings highlight the need to attend to divergent ways clients may be affected by, and respond to, the pandemic's effects.

The complex relationship between mental health and well-being invokes a need to employ person-centred data analytic approaches (e.g. Burton et al., 2018), which explore diversity within a sample by identifying subgroups based on similar scores on multiple constructs. The resulting subgroups are ‘homogeneous within a given category and are heterogeneous across categories’ (Muthén & Muthén, 2000, p. 883). Most of the pandemic-related literature so far has employed variable-centred analytic methods, which examine group-level associations among constructs across a sample. Employing person-centred analyses to examine mental health and well-being may provide a more nuanced understanding of pandemic functioning and help identify groups of clients with unique treatment needs. Further, most of the literature has reported quantitative findings exclusively, which effectively depicts broad trends but has not captured the variety of ways the pandemic has uniquely affected people's lives or how they have responded. There are likely a variety of risk and protective factors that influence how clients fare (Mancini, 2020), and while there has been justifiable attention to pandemic-related challenges, inquiring about benefits and learning may offer insight into factors and processes relevant for pandemic functioning.

1.4 The current study

The purpose of this study was to examine psychotherapy clients' functioning and adaptation in the early months of the COVID-19 pandemic. A mixed-methods, concurrent triangulation design was used (Hanson et al., 2005), whereby quantitative and qualitative data were collected at the same time, given equal priority and analysed so results from each method informed interpretation of the other's results. Given the exploratory nature of the study and person-centred analysis, and evidence that symptoms and well-being are distinct dimensions that can be related to each other in different ways in clinical samples (e.g. Jankowski et al., 2021), our first aim was to use person-centred analysis to empirically identify distinct subgroups on indicators of symptoms, specifically self-reported severity of depressive and anxious symptoms, and levels of emotional (i.e. hedonic or subjective well-being), and psychological and social well-being (i.e. eudaimonic well-being). Our second aim involved using qualitative data analyses to describe client experiences of coping and adaptation during the early phase of the pandemic. Our final aim was to integrate quantitative and qualitative findings in the interpretation phase, by (a) examining responses to the COVID impact item by subgroup, and (b) descriptively comparing subgroups on our coding of participants' responses to the questions about challenges, benefits and learning during the pandemic.

2 METHOD 2.1 Study participants

Participants were outpatient clients at a psychodynamic-oriented community mental health clinic in a large urban area of the northeastern United States. Ninety-five clients completed the study measures. Clients whose mental health diagnoses included symptoms of psychosis or severe forms of dissociation (e.g. dissociative identity disorder), and whose responses suggested that they were in a dissociative state at the time of data collection, were excluded from the analysis. One such case was identified and subsequently dropped from the dataset. The remaining clients (N = 94) ranged from 20 to 81 years old (M = 41.53, SD = 15.35), and they identified as female (66%), male (27.7%), transgender (1%), genderqueer (2.1%), and other or more than one gender (e.g. ‘intersex and bigender tending to female’, 3.2%). Their sexual orientations included heterosexual (67%), bisexual (12.8%), gay (7.4%), lesbian (2.1%), pansexual (2.1%), asexual (1%), and other or more than one sexual orientation (e.g. ‘bisexual/polyamorous’, 7.4%). A majority of clients identified their race as White (76.6%), whereas others identified as Asian (6.4%), Black or African American (6.4%), Middle Eastern/North African (2.1%), biracial (4.3%) or unreported (4.3%). Five (5.3%) reported being Hispanic or Latino/a.

2.2 Procedures

The study clinic assesses mental health, well-being and indicators of virtue and flourishing as part of ongoing clinical routine outcome monitoring (Lambert et al., 2018). Four questions about the pandemic's effects were added to the previously established battery of measures. In mid-May 2020, clients received an encrypted email from a university-sponsored, HIPAA-compliant survey tool, REDCap (Harris et al., 2009, 2019), and were directed to an online form where they reviewed consent information and completed the measures. Data were collected over four weeks.

2.3 Measures 2.3.1 Depression

The Patient Health Questionnaire (PHQ-9; Kroenke & Spitzer, 2002) assessed depressive symptoms. The PHQ is a 9-item self-report measure frequently used in psychiatric and medical settings (e.g. Arroll et al., 2010; Beard et al., 2016) to assess symptoms of major depression. Clients reported symptom severity for each item (e.g. ‘little interest or pleasure in doing things’) on a 4-point scale ranging from 0 (not at all) to 3 (nearly every day). Internal reliability for the PHQ-9 in this study was α = 0.87. Higher sum scores represented greater levels of symptoms.

2.3.2 Anxiety

The Generalized Anxiety Disorder scale (GAD-7; Spitzer et al., 2006) assessed anxiety symptoms using a 4-point, 7-item measure with response items ranging from 0 (not at all) to 3 (nearly every day). Sample items include ‘feeling nervous, anxious, or on edge’ and ‘feeling so restless that it's hard to sit still’. Internal reliability for the GAD-7 in this study was α = 0.89. Higher sum scores represented greater symptoms.

2.3.3 Well-being

The 14-item Mental Health Continuum-Short Form (MHC-SF; Lamers et al., 2011) assessed three dimensions of well-being: hedonic/emotional (EWB; three items; e.g. ‘happy’), eudaimonic/psychological (PWB; six items; e.g. ‘confident to think and express your own ideas and opinions’) and eudaimonic/social (SWB; five items; e.g. ‘that you had something important to contribute to society’). Participants rated frequency of each feeling on a 6-point scale ranging from 1 (never) to 6 (every day). Internal reliability scores for the MHC-SF sub-scales were α = 0.86 (EWB), α = 0.87 (PWB) and α = 0.82 (SWB). Higher sum scores on each subscale represented greater well-being.

2.3.4 COVID-19 impact

Similar to Klaiber et al.'s (2021) rating of pandemic-related stress, a single item assessed the overall impact of the COVID-19 pandemic on clients' lives on a sliding scale ranging from 0 (negatively) to 100 (positively).

2.3.5 COVID-19 challenges, benefits and learning

Participants responded to three questions about their functioning during the pandemic and were encouraged to be as detailed as possible: (a) ‘In your own words, what, if anything, has been most challenging about the ways the COVID-19 situation has impacted you?’; (b) ‘In your own words, what, if anything, has been most beneficial about the ways the COVID-19 situation has impacted you?’; and (c) ‘In reflecting on the changes that have been involved in trying to deal with the COVID-19 situation, what have you been learning about yourself?’

2.4 Data analysis plan 2.4.1 Quantitative

The aims of the quantitative analysis were to: (a) identify different client subgroups based on symptoms and well-being early in the pandemic context, and (b) describe and interpret the subgroups to inform the qualitative analysis. We used latent profile analysis (LPA) to identify client subgroups based on the similarity of their responses across the two symptom measures and three well-being domains. Table 1 presents bivariate correlations for the key study variables. The number of subgroups was determined by lowest Akaike Information Criteria (AIC) and Bayesian Information Criteria (BIC) values. BIC was also used to test the assumption of local independence (Asparouhov & Muthén, 2014). Consistent with Celeux and Soromenho's (1996) recommendations, we also considered an entropy estimate above 0.80 as an acceptable level of separation between classes.

We examined the clients' response to the COVID impact item and available demographic variables, specifically age, race, gender and sexual orientation, as predictors of class membership using the automated 3-step method (R3STEP in Mplus; Asparouhov & Muthén, 2013; Vermunt, 2010). This automated method allows dichotomous or continuous covariates to be included in the model without influencing the enumeration phase and uses multinomial logistic regression to assess whether an increase in a given covariate is associated with a higher probability that a participant belongs to one subgroup over another (Asparouhov & Muthén, 2013; Vermunt, 2010). Data were cleaned (e.g. assessed for outliers) in SPSS v24 before being exported to Mplus v8.4 for all statistical analyses.

TABLE 1. Bivariate correlations among mental health and well-being variables Measure 1 2 3 4 5 6 1. Depression – 2. Anxiety 0.71** – 3. Emotional well-being −0.55** −0.38** – 4. Psychological well-being −0.57** −0.53** 0.75** – 5. Social well-being −0.56** −0.45** 0.72** 0.80* – 6. COVID−19 pandemic impact −0.36** −0.27* 0.27* 0.16 0.14 – 2.4.2 Qualitative

Procedures outlined for thematic analysis (Braun & Clarke, 2006) informed the process of identifying the range of client experiences. Without knowledge of LPA subgroup membership, we (first and second authors) immersed ourselves in the data by inductively reading participants' responses to the open-ended questions about challenges, benefits and self-learning. Several meetings followed to discuss observations and potential patterns, which informed the development of preliminary codes. Through this process, we identified six dimensions of experience and subsequently organised clients' responses along these dimensions: psychological (cognitive effects and processes), emotional (named or implied feelings), relational (relevant to interpersonal effects), physical (bodily well-being and health), ecological/systemic (physical location and surroundings) and behavioural/lifestyle (routines, activities and vocational/educational arrangements). These dimensions provided structure and served as major themes, into which we sorted preliminary codes as sub-themes (e.g. isolation/loneliness as a form of relational challenge; see Tables 4–6). After organising the preliminary codes along these dimensions, we independently coded the data a second time in NVivo 12.6.0 (released 2019); inter-coder reliability estimates within each of the major themes were all greater than 0.90. Finally, we met to discuss discrepancies and reach consensus on the final sub-theme codes for each participant's response.

Next, we integrated the LPA subgroup membership into the analysis to identify trends within and differences across subgroups. First, we sorted the responses by LPA subgroup membership and each read all responses within each subgroup. Each of us distilled our understanding of the essence of each subgroup's experience into a summary paragraph. We then met to compare and discuss our observations before reaching a consensus about how to summarise the patterns within each subgroup. The numerical frequencies of themes (see Tables 4–6) helped corroborate these summative interpretations. In addition to considering symptom and well-being levels as measured for the LPA, the results from the qualitative analysis, detailed below, informed each subgroup's name.

3 RESULTS 3.1 Quantitative

After assessing model fit indices for two-, three-, four-, five-, six- and seven-class models, a five-class model was selected as best fitting the data (see Table 2). We made our decision based on the lowest BIC value, which indicated that a 5-class solution was the best fit. The BIC appears to be the most commonly used and best performing tool for enumeration in mixture modelling contexts (Nylund-Gibson & Choi, 2018; Sterba, 2016). In addition, the size of one subgroup in the 6-class solution contained only 4% of the total sample (n = 4), which fell below a recommended minimum of 5% per class, and two subgroups below this threshold in the 7-class model (Masyn, 2013). Furthermore, two classes in the 6- and 7-class solutions were deemed comparable across indicators; that is, no qualitative differences existed, so we deemed the 5-class solution as best fitting the data. Further, the assumption of local independence was not violated after comparing BICs for a model with uncorrelated indicators (BIC = 2546.41) to one with correlated indicators (BIC = 2554.23).

TABLE 2. Model fit indices for latent profile analysis model selection Number of classes AIC BIC Entropy # Classes <5% 2-Class 2625.35 2589.59 0.87 3-Class 2532.94 2588.89 0.89 4-Class 2503.44 2574.65 0.89 5-Class 2477.93 2546.41 0.87 6-Class 2467.92 2569.65 0.90 1 7-Class 2466.72 2583.49 0.87 2 Abbreviations: AIC, Akaike Information Criteria; BIC, Bayesian Information Criteria.

We labelled subgroup 3, our reference class, Flourishing, given that they reported the lowest symptom levels and highest ratings across all three well-being dimensions, consistent with prior clinical research employing person-centred analyses of client data (Jankowski et al., 2021). We labelled subgroup 1 Stagnant, given that they scored mid-range on symptoms and well-being indicators. Also consistent with prior research (Jankowski et al., 2021), we labelled the second subgroup Languishing, given their high symptom levels and lowest levels of well-being across subjective, social and psychological well-being dimensions. Subgroup 4 reported comparable levels of symptoms as the Languishing and comparable levels of eudaimonic well-being as the Stagnant, but higher subjective well-being than the Stagnant, yet generally scoring mid-range on well-being. Such a pattern of higher well-being despite higher symptoms has been defined as resilience (Jankowski et al., 2021). However, we opted for the narrower label Fortitudinous because fortitude primarily refers to contexts of prolonged adversity (Van Tongeren et al., 2019) and has been used in prior research on the ongoing COVID-19 pandemic as a context of prolonged adversity (Pretorius & Padmanabhanunni, 2021b), and because the qualitative data suggested positive self (i.e. benefiting from engaging with activities, projects and learning) and other (i.e. appreciating time with their quarantine ‘pod’) appraisals despite high symptoms. Subgroup 5, we labelled Mobilized because of its seeming engagement with active coping (e.g. developing new hobbies; Lin, 2016). This subgroup reported comparable symptom scores as the Stagnant, but reported greater well-being across dimensions relative to the Stagnant. The Mobilized class also reported significantly lower well-being across all dimensions compared to the Flourishing, and comparable levels of emotional and social well-being as the Fortitudinous, and yet higher levels of psychological well-being relative to the Fortitudinous. See Table 3 for descriptive statistics of indicator variables, class specific means and significant differences.

TABLE 3. Full sample and class-specific descriptive statistics and differences between classes on indicator variables Full sample (N = 94) Class 1 (n = 15) Class 2 (n = 8) Class 3 (n = 28) Class 4 (n = 12) Class 5 (n = 31) Stagnant Languishing Flourishing Fortitudinous Mobilized Indicators M SD Range M M M M M Anxiety 6.20 4.76 0–21 5.29a 12.61b 3.04c 13.42b 4.82a Depression 6.83 5.24 0–27 6.75a 16.26b 2.44c 12.58b 5.92a EWB 9.46 3.26 0–15 6.08a 3.18b 12.24c 9.44d 10.45d PWB 19.49 6.32 0–30 14.90a 7.75b 25.71c 16.27a 21.05d SWB 11.20 5.44 0–25 7.36a 2.83b 17.22c 8.55ad 11.18d Note Values with different superscripts in same row represent significantly different values at p ≤ .05. Abbreviations: EWB, Emotional well-being; PWB, Psychological well-being; SWB