Deprivation, essential and non-essential activities and SARS-CoV-2 infection following the lifting of national public health restrictions in England and Wales

Patient and public involvement

Due to constraints related to conducting research with a wide remit during the COVID-19 pandemic, patients and/or the public were not involved in the design or dissemination of this study.

Study design and setting

The analyses are based on the Virus Watch Community Cohort in England and Wales, the detailed methodology of which, including eligibility criteria, recruitment and follow-up methods, is described elsewhere³⁰. Briefly, the study recruits whole households with detailed baseline information, weekly surveys of symptoms and self-reported positive SARS-COV-2 tests (PCR or lateral flow) conducted through the national tracing programme, linkage to the national testing data-set, and monthly questionnaires on social activity patterns during the preceding week. At the time of this study, Virus Watch had recruited 58, 628 adult and child individuals.

Study population

Inclusion criteria. Within the Virus Watch community cohort study, participants were included in the current study if they were aged 18 years and above and had completed three monthly behavioural surveys following the declaration of “Freedom Day” (completed during the periods 22/09/2021 – 29/09/2021, 19/10/2021 – 26/10/2021 and 16/11/2021 – 23/11/2021) and cases were included if testing PCR or lateral flow positive between 01/09/2021 and 16/12/2021 (before the wide-spread circulation of the Omicron variant).

Exclusion criteria. Participants were excluded if there was evidence of recent infection in the previous three months (reported a positive PCR or lateral flow test in the 90 days before 01/09/2021) signally likely natural immunity. We did not include responses from the August survey as many participants were on holiday and survey completion rates were low.

Exposure

The exposure of interest, deprivation, was derived using English or Welsh Index of Multiple Deprivation (IMD) quintiles 2019^31,32. The IMD combines official data on small local areas for seven dimensions of deprivation (i.e., income, employment, education, health, crime, barriers to housing and services, and the living environment). Overall scores across these dimensions are used to rank areas from the most deprived to the least deprived. Virus Watch participants’ postcodes were linked with the May 2020 ONS Postcode Lookup file³³ to derive IMD quintiles. Consequently, only participants who provided a valid postcode at the beginning of the study were included in our analyses. IMD were provided as quintiles in this analysis (1=most deprived, 5=least deprived).

Outcomes

Survey respondents were requested to report the number of days on which they engaged in various activities during the week leading up to each monthly activity survey, including attending work outside their homes. Composite variables based on multiple items were created for the following activities: taking public transport (use of taxi, bus, over and underground rail or tram and air travel), use of shared car with a non-household member, indoor hospitality (eating in an indoor restaurant, café or canteen; going to an indoor bar, pub or club; and going to an indoor party), outdoor hospitality (eating in an outdoor restaurant, café, or canteen; going to an outdoor bar, pub or club; and going to an outdoor party), indoor leisure (attending a gym, the theatre, the cinema, a concert or sports event), outdoor leisure (outdoor team sport), non-social activities (visiting a barber, hairdresser, beautician or nail salon).

To estimate overall activity patterns during the period immediately after Freedom Day, the weekly frequency of each activity was calculated by averaging relevant data from the three surveys. The following binary outcomes were then classified based on the frequency distribution for each composite activity variable: leaving home to go to work or education (no, any), using public or shared transport (none, any), visiting indoor hospitality settings (up to once a week, more than once a week), outdoor hospitality settings (none, any), indoor leisure settings (none, any), undertaking outdoor leisure activities (none, any), and visiting non-social settings (none, any) in the week prior to the survey.

To examine the association between deprivation and infection, SARS-CoV-2 infection status was binary coded (yes/no evidence of infection) based on any of the following: i) a positive self-reported PCR test, ii) a positive self-reported lateral flow test, iii) a positive PCR or lateral flow test from data linkage to the Second Generation Surveillance System (SGSS), which contains official data regarding SARS-CoV-2 test results from hospitalisations (Pillar 1) and national community testing (pillar 2). Linkage was conducted by NHS Digital.

Statistical methods

We used logistic regressions models to examine the association between deprivation and undertaking non-household activities and between deprivation and infection. For both sets of analyses, the least deprived quintile (IMD 5) was used as the reference category for level of deprivation.

Deprivation and activities. We undertook separate logistic regression models for each of the activities (leaving home to go to work, car sharing and each of the composite activities). Univariable analyses was performed to examine the relationship between deprivation and the proportion undertaking each activity on a weekly basis. Multivariable models were adjusted for variables relevant to each activity. Where included, age was classified as adult of working age (18–64 years) or adult of retired age (65 years and above). Region was derived from linking participants’ postcode to ONS national region using the May 2020 National Statistics Postcode Lookup file (14).

The model examining the association between deprivation and leaving home for work was adjusted by sex, region, living with children, and area of residence, with a model additionally examining the effect of age. Sex was considered a relevant a priori potential confounder. Geographic region and area of residence (rural, urban, or conurbation) were controlled for as the local prevalence of infection likely determines the risk associated with doing any activities (separate to the actual risk of the activity itself). We controlled for the presence of children (<18 years) in the household due to the likely influence of COVID-19-related school closures on the working patterns of parents and carers. The additional model adjusting for the effect of age was included due to a plausible relationship between both IMD and working status.

We adjusted the use of the public transport and car share models by sex, region and area to take account of likely differential use of public transport in different geographical regions and differences between rural, urban and conurban areas. We conducted a further model additionally adjusted for age and employment status as, although having to leave home for employment is likely on the causal pathway between deprivation and use of public transport or car sharing, employment status is likely also related to both deprivation status and public transport use.

All other activity models (indoor hospitality use, outdoor hospitality use, indoor leisure, outdoor leisure and non-social activities) were each adjusted for age, on the basis that age is significantly related to activity levels, sex a priori, and living with children and living alone under the hypothesis that living with others is likely to affect the ability to undertake other activities (either reducing the opportunities to go out if looking after children or increasing the need to seek social engagement outside the home if living alone).

Missing data were sparse and for ease of comparisons to the adjusted activity models participants with missing data were not included in the univariate analysis of the association between deprivation and activity, nor in the multivariate adjusted models.

Deprivation and infection. For the infection analysis, we undertook univariable analyses comparing the proportion with evidence of infection according to deprivation status. We used multivariable logistic regression to adjust for age, sex and vaccination status, a priori, region due to the likely influence of regional variations in prevalence rates affecting infection risk, area of residence, living alone or living with children due to known associations with both deprivation and infection. Missing data were sparse and while shown in the univariate analyses and for each co-variable, participants with missing data were not included in the multivariate adjusted model.

All analyses were carried out using STATA version 16.

Ethical considerations

This study has been approved by the Hampstead NHS Health Research Authority Ethics Committee, Ethics approval number—20/HRA/2320. Written consent or assent was obtained at study registration for all aspects of the study. Participants were informed when providing consent that their de-identified data would be processed by the study team within a secure research environment for research purposes and overall study results without identifiable information would be published as scientific articles and presentations at scientific meetings.

Deprivation and activities (Figure 1)

There was strong evidence that living in areas with any level of deprivation was associated with having to leave home to go to work, the odds increasing with each level of deprivation, the greatest found in those who are most deprived (OR 1.59 (1.37 – 1.86) (Table 2.1). This association remained strong for all levels of deprivation after multivariable adjustments (aOR 1.56 (1.33 – 1.83) for the most deprived). The strength of the increase in odds was reduced across all strata (aOR 1.26 (1.06 – 1.49) for most deprived) when additionally adjusting for age but remained significant.

Figure 1. Fully adjusted relationship between deprivation and activities (adjustments as per Table 2.1–Table 2.8).

There was strong evidence that deprivation was independently associated with public transport use, the effect being particularly strong among those living with the greatest level of deprivation (OR 1.71 (1.47 – 1.99)) (Table 2.2). After controlling for the effects of sex, region and area, the size of the effect reduced (aOR 1.33 (1.13 – 1.57) but remained significant. There was little change after additionally controlling for age and employment status. Correspondingly, those living with the greatest levels of deprivation were less likely to report car-sharing with a non-household member than the least deprived (OR 0.81 (0.69 – 0.94) (Table 2.3). The size and strength of the effect remained largely unchanged (aOR 0.83 (0.71 – 0.97) after adjusting for sex, region, area, employment status and age.

Those living across nearly all strata of deprivation reported undertaking fewer hospitality, leisure or non-social activities compared to the least deprived, with decreasing odds of undertaking activities with increasing levels of deprivation (Table 2.4 – Table 2.8). The effect sizes and strength of associations were minimally altered after multivariable adjustments, with the following reduced odds observed for the most deprived compared to the least deprived: indoor hospitality (aOR 0.82 (0.70 – 0.96)), outdoor hospitality (aOR 0.56 (0.48 – 0.66)), indoor leisure (aOR 0.63 (0.54 – 0.74)), outdoor leisure (aOR 0.64 (0.46 – 0.88)), non-social activities (aOR 0.72 (0.61 – 0.85)) (Table 2.4 – Table 2.8).

Deprivation and infection

During the period (1^st September – 16^th December 2021), among 11,231 participants, 482 cases were identified (Table 3). Although an increase in odds of infection was observed across all strata of deprivation (OR 1.36 (0.96 – 1.92) among the most deprived), the observed association for deprivation was non-significant in this time-period. After adjusting for all potential confounders, those living in the greatest level of deprivation appeared to continue to have an increased odds of infection but the results remained non-significant (aOR 1.25 (0.87 – 1.79) (Table 3).

Strengths and limitations

Our exposure measure, deprivation, was measured during baseline surveys and our outcomes (either activities or infection) were measured during the post-Freedom Day period. It is possible that participants deprivation status changed throughout the pandemic due to residential moving but we did not capture updated deprivation area data. Activities and behaviours are self-reported and therefore subject to recall bias and social desirability bias although data examined during the first wave of the pandemic in Germany were found to support the use of self-reported contact survey data to reflect infection dynamics³⁷. We tried to minimise recall bias by asking about activities in the previous seven days. The activities were sampled at three points during the period post Freedom Day but did not include survey results from the month of August, a holiday period in the UK, as response rates were low. We may have missed any immediate increase in activities around Freedom Day. Moreover, data taken from one-week time-periods in a monthly survey may not be representative of activities throughout a month. We sought to minimise this bias by taking an average of activities engaged in across the three months. Using self-reported and linked data on test results from the national testing system allowed better ascertainment of infections than cohort data alone and increased ascertainment of infection data supports increasingly accurate assessment of the importance of deprivation. However, the period of follow-up for infections between Freedom Day and before the onset of the more infectious Omicron variant was short so we may have been underpowered to examine an association between deprivation and infection in the post-Freedom Day period. Virus Watch survey respondents were not demographically representative of the population, with a lower proportion of the survey samples drawn from the most deprived communities (8%) compared with the least deprived (30%), potentially affecting generalisability of responses from those living in the most deprived areas as well as statistical power. Finally, Virus Watch is a voluntary cohort likely subject to a degree of self-selection bias and it is possible that our cohort may include individuals with a higher socioeconomic status living within the more deprived areas and we may be underestimating the true effect size. Future studies that are able to capture time-updated information on deprivation status and examine infections over a longer period of time, or throughout the Omicron wave when prevalence and case numbers were higher will be of value to add power to this work and to refine this research further.