key: cord-0903551-smr0y5t9
authors: Wright, L.; Steptoe, A.; Mak, H. W.; Fancourt, D.
title: Do people reduce compliance with COVID-19 guidelines following vaccination? A longitudinal analysis of matched UK adults
date: 2021-04-15
journal: nan
DOI: 10.1101/2021.04.13.21255328
sha: dd196f2e6a573d92bdaf3da81da49978766fba54
doc_id: 903551
cord_uid: smr0y5t9

Background: Governments have begun mass vaccination programmes for COVID-19, but available vaccines do not confer immediate immunity and vaccinated individuals may still be at risk of transmitting the virus. The UK Government has not exempted vaccinated individuals from behavioural measures to reduce the spread of COVID-19, such as practicing social distancing. However, vaccinated individuals may have reduced compliance with these measures, given lower perceived risks. Methods: We used fives waves of monthly panel data from the UK COVID-19 Social Study to assess compliance following vaccination. Compliance was measured with two items on compliance with guidelines in general and compliance with social distancing. We used matching to create comparable groups of individuals by vaccination status and month of vaccination (January vaccination and February vaccination) and used fixed effects models to estimate changes in compliance between October 2020 - February 2021, a period which overlapped with the second wave of COVID-19 in the UK. Results: Vaccinated individuals were broadly keyworkers or older aged individuals. Compliance increased between October 2020 - February 2021, regardless of vaccination status or month of vaccination. There was some evidence that individuals vaccinated in January complied less with social distancing in January and February than matched non-vaccinated individuals, though associations were small and there were no similar clear differences in any month comparing January vaccinated with February vaccinated individuals and February vaccinated with non-vaccinated individuals. Conclusion: There was limited evidence that vaccinated individuals reduced compliance after receiving vaccination. Vaccinated individuals in the sample were older on average and the follow-up period post-vaccination was relatively short (1-2 months). Results require replication in other populations.

Governments have begun mass vaccination programmes for COVID-19, but it will be several months before herd immunity is achieved. The available vaccines do not confer immediate immunity and are not 100% effective (Public Health England, 2021a) . Vaccinated individuals may still be at risk of catching and transmitting the virus, including variants they have not been inoculated against (European Centre for Disease Prevention and Control, 2021) . Given this, the UK government has not exempted vaccinated individuals from behavioural measures to reduce the spread of such as the wearing of masks, practicing social distancing, and reducing household mixing.

International data show that, though compliance levels are high overall, not all individuals comply with recommended or mandated behavioural measures (YouGov, 2021) . While compliance has increased as countries have experienced second waves, overall compliance has decreased somewhat since the start of the pandemic (Petherick et al., 2021) . Vaccinated individuals, in particular, may feel less motivated to comply, given perceived lower health risks. Empirical evidence from the and previous epidemics (Barber & Kim, 2020; Bish & Michie, 2010; Harper et al., 2020) , and predictions from influential models of health behaviour, such as the Risk Compensation, Health Belief and COM-B models (Michie et al., 2011; Peltzman, 1975; Rosenstock et al., 1988) , suggest that individuals who are less concerned about catching a virus have lower compliance. Further, in the UK, citizens have expressed difficultly keeping abreast of latest rules (Denford et al., 2020; Williams et al., 2020a Williams et al., , 2020b Wright, Burton, et al., 2021) , due to variations in rules across areas and over time and (speculatively) due to "lockdown fatigue". Vaccinated individuals may therefore not be aware of nonexemption from government rules.

Early evidence from vaccine roll-out in Israel and the UK finds some increase in infection rates following first vaccination (Bernal et al., 2021; , and infection rates have risen in Chile despite high vaccination rates (Mander & Stott, 2021) . Some have argued that this may reflect lower compliance with protective behaviours (Independent SAGE, 2021; Rubin et al., 2021;  Scientific Pandemic Insights Group on Behaviours, 2021). This is supported by survey evidence from early December 2020 that 40% of respondents intended to comply less or not comply with government guidelines following vaccination (YouGov, 2020) and with recent evidence that a sizeable minority of vaccinated over 80s in the UK have subsequently broken household mixing rules (ONS, 2021) . Further, longitudinal evidence from influenza and Lyme's disease vaccination programmes shows reduced compliance with some protective behaviours (Brewer, 2007; Reiber et al., 2010 ). Yet, cross-sectional evidence inquiring about changes in behaviour following COVID-19 vaccination show more over-80s reporting increased compliance (8-15%) with hand-washing, face mask wearing, and social distancing rules than decreased compliance (1-2%) (ONS, 2021).

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 15, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 Given the risk of vaccinated individuals catching and transmitting the virus, understanding whether people comply less following vaccination is important for managing the pandemic (Ioannidis, 2021 ).

Yet, there is a notable lack of rigorous research on the consequences of COVID-19 vaccination for personal protective behaviours (Scientific Pandemic Insights Group on Behaviours, 2021). Therefore, in this paper, we used monthly panel data from a large sample of UK adults to explore changes in compliance following vaccination.

Data were drawn from the COVID-19 Social Study; a large ongoing panel study of the psychological and social experiences of over 70,000 adults (aged 18+) in the UK during the COVID-19 pandemic.

The study commenced on 21st March 2020 and involves online weekly (from August 2020, monthly) data collection from participants for the duration of the COVID-19 pandemic in the UK. The study is not random and therefore is not representative of the UK population, but it does contain a heterogeneous sample. Participants were recruited using three primary approaches. First, convenience sampling was used, including promoting the study through existing networks and mailing lists (including large databases of adults who had previously consented to be involved in health research across the UK), print and digital media coverage, and social media. Second, more targeted recruitment was undertaken focusing on (i) individuals from a low-income background, (ii) individuals with no or few educational qualifications, and (iii) individuals who were unemployed. Third, the study was promoted via partnerships with third sector organisations to vulnerable groups, including adults with pre-existing mental health conditions, older adults, carers, and people experiencing domestic violence or abuse. The study was approved by the UCL Research Ethics Committee [12467/005] and all participants gave informed consent. The study protocol and user guide (which includes full details on recruitment, retention, data cleaning and sample demographics) are available at https://github.com/UCL-BSH/CSSUserGuide.

For these analyses, we focused on participants aged 89 or younger who completed the monthly survey in each of the five months between 23 September 2020 and 22 February 2021 (n = 25,418; 76.1% of individuals with data collection between these dates; 35.5% interviewed at any point). Ages are capped at age 90 in the data, so we excluded participants aged 90 or above from this analysis. Though there is slight overlap in calendar months, for brevity, below we refer to the survey waves as October, November, December, January, and February waves, respectively. We used matching in this analysis and excluded participants with missing data on any variable used (n = 2,132; 8.4% of the eligible sample). This left a total sample size of 23,296.

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 15, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 4

The vaccine roll-out began in the UK on 8 December 2020. 768,000 individuals were vaccinated in England by 27 December 2020, 6.3 million by 28 January 2020 and 14.9 million by 25 February 2020

(1.4%, 11.4%, 27.0% of the population, respectively; NHS England, 2021). The COVID-19 Social Study does not contain information on the date of vaccination, but given few individuals reported being vaccinated on, or shortly after, 23 December 2020, we assume that no participants were vaccinated before this date (1.32% of participants recorded vaccination on 23 December 2020). The vaccine was initially rolled out in age order, beginning with over 80 olds, then over 75s, and over 70s.

Frontline health and social care workers, older adults in care homes, and clinically extremely vulnerable individuals were also offered the vaccine (Public Health England, 2021b).

The period studied here coincides with the (ongoing) second wave of COVID-10 in the UK. There have been a several changes to government rules across this period. Supplementary Figure Changes to government policy are described further in the Supplementary Information.

Compliance was measured with two questionnaire items, which we analysed separately. General compliance was measured with a single-item question, "Are you following the recommendations from authorities to prevent the spread of Covid-19?". Responses ranged from "1. Not at all" to "7. Very much so". Social distancing was measured with a single question "When you go out or meet with others have you been maintaining social distancing?". The responses categories ranged from "1. Yes, completely" to "4. Not at all" with an extra category for those who had not met with others or left their home in the last week. We reverse code this item so high scores indicate greater compliance and code those who did not leave them home or meet with others as the highest level of compliance (range 1-5).

Our analysis proceeded in three steps. First, we split our sample into three groups: individuals who first reported being vaccinated in the January wave; individuals who first reported being vaccinated in the February wave; and individuals who did not report being vaccinated in any wave. Second, given the rules used for roll-out of the vaccine, we used matching to obtain samples of similar individuals across the three groups. As our "treatment" variable (vaccination) had three levels, we carried out matching for each combination of two groups, obtaining three matched samples (January vs February vaccinators; February vs non-vaccinated; and January vs non-vaccinated). Observations were matched using Mahalanobis distance within a caliper of 0.25 SD in propensity scores. We used 1-to-1 matching without replacement and discarded observations outside the region of common support.

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 15, 2021. ; https://doi.org/10. 1101 In the Mahalanobis distance step, given vaccine eligibility criteria, we matched upon age, date of interview in the December wave, whether the participant was a keyworker, and whether they had a flu vaccine in the past year (an indicator of existing health problems and willingness to accept vaccination). To estimate propensity scores, we used variables for age (natural splines with degrees of freedom 3), date of data collection in December (natural splines with degrees of freedom 3), keyworker status, previous flu vaccination, sex, general compliance and social distancing in the December wave (inputted as categorical variables), attitudes to vaccination (exploratory factor analysis of 12 items), intention to receive COVID-19 vaccination (September wave; categorical variable), whether the participant reported shielding for health reasons at any point, and number of chronic health conditions (0, 1, 2+) and whether the participant had a diagnosis for a psychiatric condition. More detail on these variables is given in the Supplementary Information. We assessed match quality as bias < 10%for each covariate, Rubin's B < 25%, Rubin's R of 0.5-5, and visual inspection of the distributions for variables used in the Mahalanobis distance matching step.

In the third step, we estimated fixed effects regression models for each matched sample, separately, comparing within-person changes in compliance behaviour by wave of data collection across vaccination groups. Our model was of the form:

where i and t index individuals and waves respectively.

is a categorical variable for wave of data collection (December wave used as reference category). are person-specific and observation-specific random errors, respectively.

Our interest was in the sign and size of the coefficients ߚ . Our hypothesis was that, compared with non-vaccinated individuals, compliance would be lower among vaccinated individuals in the months that they were vaccinated, and, given that vaccination does not confer immediate immunity, progressively lower the more time had elapsed since vaccination. There should also be no differences in compliance levels in the months prior to vaccination. In our data, this hypothesis translated into no differences in compliance by vaccination status in the months October, November, and December; differences in compliance in January and February when comparing January vaccinators with February vaccinated or non-vaccinated individuals; and differences in compliance in February but not January when comparing February vaccinators with non-vaccinated individuals.

Data analysis was carried out in R v 4.0.3. (R Core Team, 2020). Matchings was carried out using the matchit package (Ho et al., 2011) , Due to stipulations set out by the ethics committee, data will be . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 15, 2021. ; https://doi.org/10.1101/2021.04.13.21255328 doi: medRxiv preprint 6 made available at the end of the pandemic. The code to replicate the analysis is available at https://osf.io/xghvb/.

Descriptive statistics for the full sample are displayed in Table 1 . There were several differences among the vaccination groups, most notably on age, keyworker status, and date of December interview. Differences were markedly smaller following matching (Supplementary Table S1 ). . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 15, 2021. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 15, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 8 Rubin's R of 0.5-2, and a bias of < 10% SD for each covariate. Figure 1 shows the trends in each compliance measure over the study period. As the UK entered a second wave, there were increases in both compliance measures, though with some decrease in social distancing over December (see also Ipsos MORI, 2020; YouGov, 2021) .

The results of the fixed effects regressions are displayed in Figure 2 . There were no statistically significant differences in either compliance comparing January vs February vaccinated groups, and . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 15, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 9 point estimates were small in size and close to zero. Similar results were obtained in the February vaccinated vs non-vaccinated matched sample, except vaccinated individuals reported greater (within-person) changes in general compliance in February, but not other months. There were statistically significant differences in social distancing when comparing January vaccinated vs nonvaccinated individuals, with January vaccinated individuals reporting smaller within-person changes in social distancing in January and February. Figure   S8 ). In fact, as shown in Supplementary Table S1, average compliance levels increased among all groups between October and February in line with the increase in compliance seen in the wider population.

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 15, 2021. ; https://doi.org/10.1101/2021.04.13.21255328 doi: medRxiv preprint Figure S3 : Distribution of compliance behaviours by vaccination status and wave, January vs February vaccinated matched sample.

Given that fixed effects regressions compare within-person changes in compliance levels across vaccination groups, we also repeated the model in (1) using mixed effects modelling, interpreting the term ‫ݑ‬ as a normally-distributed random intercept. These regressions tested differences in compliance levels by vaccination status and wave. The results are shown in Supplementary Figure S9 and are qualitatively similar to those shows in Figure 2 .

Using panel data from five months of the pandemic in the UK, we found no consistent evidence that receiving a COVID-19 vaccine reduced compliance behaviour. There was some evidence that . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 15, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 individuals vaccinated in January practiced less social distancing than (matched) non-vaccinated individuals. However, the quality of the matching appeared poor and the results were not replicated comparing individuals vaccinated in January with those vaccinated in February. There was also some (inconsistent) evidence that vaccinated individuals reported higher compliance with guidelines in general, though this could be explained by individuals considering vaccination a form of compliance behaviour. Descriptively, there was little evidence of vaccinated individuals reducing compliance altogether. In fact, vaccinated individuals -like non-vaccinated individuals -increased compliance from the beginning of the period as the UK experienced its second wave of COVID-19.

The results are striking given existing evidence that compliance levels are higher among those with greater health risks from -or greater fears of -catching COVID-19 (Barber & Kim, 2020; Harper et al., 2020) , and evidence of widespread intentions to reduce compliance following vaccination (YouGov, 2020 ). An explanation for the discrepancy may be the almost exclusive use of crosssectional data in the literature -a recent study shows that marked differences in between-person and within-person associations between compliance and several factors (Wright, Steptoe, et al., 2021) .

The results suggest that vaccinations do not crowd-out other preventive behaviours. However, it should be noted that we used a relatively short follow-up period -differences in compliance may take time to arise, especially as individuals are warned that vaccines do not take effect immediately and second vaccinations are required for full effectiveness. Vaccinated individuals in our sample were also relatively old. The results may have been different were vaccinations rolled out more widely. For instance, intentions to reduce compliance or not comply following vaccination are higher among younger age groups (YouGov, 2020) . Further, compliance was measured during a period of strict lockdown where the opportunities for non-compliance were limited. This study should be repeated as lockdowns are eased. We also only focused on two measures of compliance. Differences could potentially be observed for other behaviours, such as indoor or outdoor household mixing.

This study had a number of other limitations. First, we used two self-report measures of compliance which may be subject to biases such as recall bias or social desirability bias. Being vaccinated could be considered a form of compliance so our general compliance measure may not have been specific enough to pick up on differences in specific compliance behaviour. Second, our sample was not representative and, moreover, comprised of individuals who comply more than on average (Wright & Fancourt, 2020) . This may have biased associations toward the null. Third, the existence of the vaccine program may have induced behaviour changes in the non-vaccinated group, if these individuals were less concerned about transmitting the virus (Trogen & Caplan, 2021) . Fourth, compliance was changing over time, even in the absence of vaccination. Previous research has shown that the strength of several factors in predicting compliance differs over pandemics (van der Weerd et al., 2011; Wright & Fancourt, 2020) . Our matched samples may therefore not provide an appropriate counterfactual and results may be biased by unobserved confounding. Nevertheless, by exploiting the . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 15, 2021. ; https://doi.org/10.1101/2021.04.13.21255328 doi: medRxiv preprint 12 longitudinal nature of our sample, we were able to use compliance in months prior to vaccination as a placebo test. No statistically significant differences were found in these months, which may add confidence to our results.

Our results suggest that there is no immediate cause for concern of widespread non-compliance among vaccinated individuals. However, it is important to continue monitoring the situation as the vaccine is roll-out more widely, restrictions are lifted, and people receive second doses. Analyses using data from other populations and that examine the potential impact of widespread vaccination on the behaviour of those not yet vaccinated are also required in order to ensure that the gains of the vaccination program are not lost through increases in risky behaviour.

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. https://www.england.nhs.uk/statistics/statistical-work-areas/covid-19-vaccinations/ . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 15, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 

COVID-19 Worries and Behavior Changes in Older and Younger Men and Women

Early effectiveness of COVID-19 vaccination with BNT162b2 mRNA vaccine and ChAdOx1 adenovirus vector vaccine on symptomatic disease, hospitalisations and mortality in older adults in England

Demographic and attitudinal determinants of protective behaviours during a pandemic: A review

Risk compensation and vaccination: Can getting vaccinated cause people to engage in risky behaviors?

Understanding patterns of adherence to COVID-19 mitigation measures: A qualitative interview study

Functional Fear Predicts Public Health Compliance in the COVID-19 Pandemic

MatchIt: Nonparametric Preprocessing for Parametric Causal Inference

Estimating the effectiveness of the Pfizer COVID-19 BNT162b2 vaccine after a single dose. A reanalysis of a study of 'real-world' vaccination outcomes from Israel

Maintaining adherence to protective behaviours during vaccination rollout: Summary and recommendations (No. 30; The Independent SAGE Report

Benefit of COVID-19 Vaccination Accounting for Potential Risk Compensation (SSRN Scholarly Paper ID 3773950)

Britons increasingly abiding by the COVID-19 rules, with social responsibility and the NHS the primary drivers

Coronavirus and vaccine attitudes and behaviours in England: Over 80s population

februaryto20february2021#behaviours-of-over-80s-after-receiving-at-least-one-dose-of-acoronavirus-covid-19-vaccine

The Effects of Automobile Safety Regulation

A Worldwide Assessment of COVID-19 Pandemic-Policy Fatigue

Public Health England vaccine effectiveness report

Why you have to wait for your COVID-19 vaccine

R: A language and environment for statistical computing (3.6.3) [Computer software

R Foundation for Statistical Computing

Change in Human Social Behavior in Response to a Common Vaccine

Social Learning Theory and the Health Belief Model

Are people letting down their

SPI-B: Possible impact of the COVID-19 vaccination programme on adherence to rules and guidance about personal protective behaviours aimed at preventing spread of the virus

Risk Compensation and COVID-19 Vaccines

Monitoring the level of government trust, risk perception and intention of the general public to adopt protective measures during the influenza A (H1N1) pandemic in the Netherlands

Public perceptions and experiences of social distancing and social isolation during the COVID-19 pandemic: A UKbased focus group study

Public perceptions of nonadherence to COVID-19 measures by self and others in the United Kingdom

Public Opinion about the

The researchers are grateful for the support of a number of organisations with their recruitment efforts including: the UKRI Mental Health Networks, Find Out Now, UCL BioResource, HealthWise Wales, SEO Works, FieldworkHub, and Optimal Workshop.

Author contributions LW, AS, HWM, and DF conceived and designed the study. LW analysed the data and wrote the first draft. All authors provided critical revisions.

Data used in this study will be made publicly available once the pandemic is over. The code used to run the analysis is available at https://osf.io/xghvb/.. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)The copyright holder for this preprint this version posted April 15, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021