Modelling the complexity of pandemic-related lifestyle quality change and mental health: an analysis of a nationally representative UK general population sample

The strict COVID-19-related restrictions and regulations (e.g. lockdowns, social distancing) that have been put in place have considerably changed many individuals’ daily lives, routines, and relationships [1]. These measures have been effective in suppressing the transmission of the disease and will likely remain in place pending full rollout of vaccines that have been developed [2]. As a result, there has been widespread and ongoing concern that government-prescribed mitigation measures have, alongside virus-related morbidity, mortality and grief, negatively impacted the mental health of the general population [1, 3‐5]. A recent meta-analysis examining changes in mental health pre- and post-COVID-19 outbreak suggests that there has been a small overall increase in symptoms of anxiety and depression [6]. However, psychological responses since the outbreak appear to be highly heterogeneous [7, 8].

Specific COVID-19-related lifestyle changes have been documented in relation to health behaviours, including decreased physical activity, increased sedentary behaviour, screen time, food and alcohol consumption, unhealthier food choices, and poorer sleep [9‐11]. Moreover, disruptions to personal relationships and social activity have been common [12‐16]. Many individuals may also be under increased financial pressures as a result of unemployment or reduced working hours [17‐19], while employed individuals may be struggling with changes to work schedules, roles and adjusting to a home-working environment [15, 20‐22]. These negative lifestyle changes have been associated with poor mental health outcomes during this crisis in a number of studies [10, 22, 23]. Yet, it is underappreciated that individuals may also respond positively to the changes in life brought about by the COVID-19 pandemic. For example, some studies have identified groups of individuals who are living healthier lifestyles during the pandemic [9], reported improvements in their relationship satisfaction [16], and have decreased their overall spending [19, 24], all of which may positively impact their lives.

Importantly, however, these changes do not affect everyone equally. For example, unemployment and financial difficulties appear to be affecting the young, least educated, ethnic minorities, women and those with pre-existing financial difficulties the most [17, 19, 25]. Parents are negatively affected by school closures, particularly mothers who appear to be shouldering more caring responsibilities and are more likely to experience negative employment outcomes as a result of home-schooling demands [19, 26, 27]. Additionally, those living in cramped or poor housing conditions may also be more negatively affected [28]. Furthermore, particular lifestyle changes may be likely to co-occur with one another. For example, financial strain experienced by couples during the pandemic may negatively affect their relationship quality and satisfaction [29, 30].

To model and describe the complexity of pandemic-related lifestyle quality change, as well as its context and potential consequences, the following study had three objectives. First, using a broad array of social, health, relationship and economic lifestyle indicators that were designed to capture pandemic-induced lifestyle changes, we sought to identify the primary dimensions of lifestyle quality change within the general adult population during the first 4 months of the COVID-19 pandemic. In doing so, we aimed to distil the complexity of these many and varied lifestyle changes into a more parsimonious representation of pandemic-related change in lifestyle quality. Second, we then sought to identify if there were distinct groups within the population who were characterised by the same profile of variation across the dimensions of lifestyle quality change. This afforded an opportunity to identify, not only, what aspects of lifestyle changed for distinct groups within the population, but whether the lifestyles of these groups were changing for better or worse. Third, we aimed to describe and differentiate the membership of the distinct lifestyle quality change groups using a variety of sociodemographic, COVID-19-related, psychological, and mental health variables to better understand whose lives were changing and potentially reveal some of the factors that may be underlying these changes.

Endorsement frequencies for the 19 lifestyle quality indicators are reported in Table 1. More than half of respondents (57.0–86.8%) indicated ‘no difference’ in each area of their life compared to before the pandemic, with the exception of ‘Exercise’ and ‘Socialising’. In relation to ‘Exercise’, a quarter of individuals (24.0%) indicated that this had improved, while three in ten reported that it had deteriorated. Additionally, almost half of the sample (45.1%) indicated that their ‘Socialising’ was worse now than before the pandemic, whereas only around 1 in 15 (6.4%) reported that it had improved. Other than ‘Socialising’, variables which had the largest worse:better ratios included ‘taking care of your mental health’ (2.4:1), ‘relationship with your friends’ (2.1:1), and ‘sex life’ (2.1:1), while ‘relationship with your children’ (0.4:1) and ‘time spent commuting’ (0.2:1) had the smallest worse:better ratios.

The seven classes identified in the current analyses revealed a number of interesting things about the lifestyle of the UK adult population 4 months after the first nationwide lockdown. These can be summarised as three key findings. First, the largest class, which comprised over two-thirds of the sample, reported virtually no change across the five lifestyle dimensions. The size of this class suggests that over the first 4 months of the pandemic, the quality of most individuals’ lifestyles remained largely unchanged compared to their pre-pandemic lives. Second, nearly one-in-six individuals in the sample reported that they had experienced improvements in two or more areas of their lifestyle compared to before the pandemic (Classes 1, 4 and 5). Overall lifestyle improvement, improvement in partner and family and friend relations only, and better lifestyle with the exception of partner relationship, were generally characterised by a unique pattern of sociodemographic, COVID-19 and psychological associations. Moreover, none of these classes were associated with any of the four mental health covariates. Third, Classes 2, 3, and 7 comprised nearly one-in-seven individuals in the sample, and represented those members of the UK population whose lifestyle had worsened in relation to at least one dimension of pandemic-related lifestyle change.

Class 3 exhibited the most extreme deterioration in lifestyle change and this was evident across all dimensions. Compared to Class 3, Class 2 was characterised by a similar but less severe pattern of negative change across all dimensions except the partner relations dimension, where little change was identified. Conversely, Class 7 exhibited relatively little change across all lifestyle dimensions, except partner relations, where the level of deterioration was similar to that identified in Class 3. Overall, therefore, class composition suggested that, when lifestyle quality changed during the pandemic, it centred around either positive/negative changes (i) across all aspects of lifestyle quality, (ii) across all areas except partner relationships and (iii) to partner relationships only (or in the case of Class 4, positive changes to the quality of relationships generally).

Of the negative lifestyle change classes, compared to the no-change baseline majority (i.e. Class 6), only Classes 3 and 7 were more likely to experience mental health problems. Members of both classes were more likely to meet caseness for COVID-19 traumatic stress and report suicidal ideation. Members of both classes were also more likely to be in a relationship, and live in an area that was under local lockdown. Classes 3 and 7 were, however, distinct from each other in relation to other covariates. Membership of Class 3 (deterioration across all dimensions) was uniquely predicted by city living, being white, an absence of an underlying health condition, and being a keyworker, while membership of Class 7 (deterioration in partner relations only) was predicted by a medium income level, children in the home, living in a two-bedroom property, lost income due to the pandemic, the presence of an underlying health condition, and higher levels of loneliness. Somewhat surprisingly, while Class 2 experienced deterioration in four of the five lifestyle dimensions, it was not at greater risk of mental ill health and was predicted only by economic activity status and loneliness.

Class 3 clearly demonstrates that some members of the population are experiencing extreme difficulties during the pandemic and that multiple aspects of their lives are being affected. Their city living and frontline worker status are likely to impact their lifestyles and contribute to fears of contagion, infection, and illness that commonly arise from higher rates of transmission in more populated areas [51], and greater exposure in frontline working environments [52]. Class 7 suggests that compromised partner relations may be a particularly meaningful aspect of lifestyle change during the pandemic that, in itself, signals other important difficulties in household circumstances such as having dependent children in the home, living in a smaller house, higher levels of loneliness, having a family member with a chronic health condition (who may or may not live in the home) and having lost income, all factors known to compromise relationships and well-being pre-pandemic [53‐57].

Compromised partner relations may also, however, be consequential to other pandemic related factors. A number of COVID-19-related studies have begun to explore this. Using data from an online nationally representative probability survey of 1,010 American adults in April 2020, for example, Luetke et al. [13] found that 34% of respondents in a relationship reported some degree of conflict with their romantic partners due to the spread of COVID-19 and its related restrictions. In addition, they reported that individuals who experienced frequent coronavirus-related conflict with their partner were significantly more likely to report decreased frequency of intimate and sexual behaviours compared to those not experiencing any such conflict, exhibiting a dose–response trend among partnered sexual behaviours. Pietromonaco et al. [30] suggest that facing COVID-19-related external stress is likely to increase harmful dyadic processes (e.g. hostility, withdrawal, less responsive support), which may undermine couples’ relationship quality. Moreover, these harmful effects are likely to be exacerbated by the broader pre-existing context in which couples’ relationships are situated (e.g. social class, minority status, age), and their individual vulnerabilities (e.g. attachment insecurity, depression). Prime, Wade and Browne [58] state that the COVID-19 pandemic poses an acute threat to the well-being of families due to challenges related to social disruption such as financial insecurity, caregiving burden, and confinement-related stress (e.g. crowding, changes to routine). According to these authors, the consequences of these difficulties are likely to be longstanding, in part because of the ways in which contextual risk permeates the structures and processes of family systems.

Finally, a growing number of studies have reported increases in domestic and intimate partner violence during the pandemic [59‐61]. For example, a 60% increase in emergency calls from women subjected to violence by their intimate partner has been reported in the World Health Organization Europe member states. Comparing April 2020 with the same period last year, WHO reported that online inquiries to violence prevention support hotlines had also increased as much as fivefold [14]. Overall therefore, relationship complications (that may have existed before and/or been caused/compounded by the pandemic), alongside negative changes in other lifestyle dimensions in some cases, seem to constitute the more severe pandemic-related lifestyle quality change due to their association with poor mental health (COVID-19 PTSD and recent suicidal ideation) and are therefore an important marker for investigation.

The current study’s strengths include its sample size and use of weighting procedures to ensure the sample was nationally representative of the UK population across a number of sociodemographic indicators. Additionally, the sophisticated analytic strategy employed allowed for a better understanding of the complex ways in which the pandemic has affected people’s lives. There are, however, several limitations of the study to consider. Firstly, the cross-sectional design limits interpretations relating to causality. Secondly, the lifestyle indicator items asked participants to retrospectively reflect on their lives now compared to before the pandemic, and thus may be susceptible to recall bias. The current study used data from Wave 3 of the C19PRC Study, which was gathered during July 2020. At this time, the UK was past the peak of its first wave of coronavirus and the daily number of cases and deaths had reduced. As a result, lockdown measures were eased, and some individuals were able to engage with more ‘normal’ routines. Future research should consider pandemic-related lifestyle change in the light of further waves of coronavirus cases and the reinstatement of lockdown periods. Third, quota sampling was used to recruit participants via non-probability, opt-in online survey panels which excluded participants who did not have access to the Internet and those who could not read or write in English. Relatedly, this sampling strategy may have been susceptible to a number of biases (sampling bias, non-response bias, demand characteristics bias, question order bias, personality biases, psychometric biases) that have been shown to undermine confidence in online survey research findings [62‐64]. Fourth, the low number of cases in some of the latent classes in the best fitting LPA model may have impeded the detection of some effects. Finally, although framed around the pandemic, we do not know with certainty whether these changes were as a result of the pandemic, either directly or indirectly.