Since 28 January 2020, before the first UK cases of COVID-19 were reported, a series of cross-sectional surveys has been conducted by BMG Research and then Savanta on behalf of the Department of Health and Social Care, England. We analysed these data as part of the COVID-19 Rapid Survey of Adherence to Interventions and Responses (CORSAIR) study. Surveys were initially weekly, then usually fortnightly from mid-July 2020. The methods for these surveys have been described in detail elsewhere . In this study, we used data from 10 February 2020 (wave 3) to 20 January 2022 (wave 66).
Approximately 2000 participants (people aged 16 years or over living in the UK) completed each survey wave (≈1700 from England per wave). Quota sampling was used, with quotas applied based on age and gender (interlocked) reflecting targets from Office for National Statistics data . From wave 8 of the survey onwards, two specialist research panel providers were used to recruit participants: Respondi (n = 50,000) and Savanta (n = 31,500). Before then, participants were recruited solely from the Respondi panel. Participants could complete multiple survey waves, but after having completed the survey once were excluded from the subsequent three waves. An error with one of the survey panels meant this exclusion rule was not applied in the first waves of data collection, leading to a small number of people (at least n = 133, 0.2% of sample) completing the survey nine times or more up to 18 November 2020 (wave 30). We have adjusted statistically for participants appearing in multiple waves, where possible. Due to an error, some repeat participants were not identified in the data available. We were unable to identify repeat participants from one of the panels before 20 September 2021 (wave 58). The impact of adjusting for repeat participants on the results is minimal. Therefore, this inability to identify some repeat participants should not have any material impact on results. Participants were reimbursed for having completed the survey in points, which could be redeemed as cash, gift vouchers or charitable donations (up to 70p per survey).
Questionnaire items reported were developed for the CORSAIR study. Full survey items for outcome measures are reported in the supplementary materials.
Hand cleaning: From 10 January 2020 (wave 3), participants were asked if they had washed their hands thoroughly and regularly with soap and water in the past 7 days. Response options were “done this, same amount as usual,” “done this, more than usual,” “not done this,” and “not applicable.” Between 27 April 2020 (wave 14) to 13 May 2020 (wave 16), the sample was split, with half the sample being asked to respond using the original options, and half the sample being asked to respond on a five-point scale from “never” to “very frequently.” From 18 May 2020 (wave 17) onwards, all participants responded on the five-point scale. From 26 October 2020 (wave 31) onwards, the item was amended so that participants were asked if they had “washed [their] hands thoroughly and regularly with soap and water, or used hand sanitising gel”. We created binary variables to indicate if people reported washing their hands frequently or very frequently (compared to never, rarely or occasionally). This item was removed from the survey after 19 May 2021 (wave 50). For these variables, “not applicable” was coded as missing (n = 847, 0.9%).
Number of outings: Participants were asked to state the number of times they had been out of their home in the last 7 days to go: to the shops, for groceries/pharmacy; to the shops, for things other than groceries/pharmacy; for a walk or some other exercise; to spend time outdoors for recreational purposes; out to work; to meet up with friends and/or family they did not live with; and to a restaurant, café or pub. These questions were introduced to the surveys on 30 March 2020 (wave 10), with the exception of spending time outdoors for recreational purposes, which was introduced on 18 May 2020 (wave 17), and going out to a restaurant, café or pub, which was introduced on 6 July 2020 (wave 24). This measure was amended on 1 June 2021 (wave 51), and from this date on, asked participants to state the number of times they had done each of the following activities in the past 7 days, including having: been to the shops, for groceries/pharmacy; been to the shops, for things other than groceries/pharmacy; spent time outdoors for exercise or recreational purposes (including to sit in parks etc.); left the house to go out to work (number of days); met up with friends and/or family that they did not live with; and been to a restaurant, café or pub. To investigate total number of outings where people were likely to come into close contact with someone from another household indoors, we created three separate variables. The first variable summed outings for shopping and to see friends or family from another household to give a “total number of outings” variable. As the item investigating outings to hospitality venues was introduced later, we created a second variable that also included going out to a restaurant, café or pub. A third variable, summing outings for shopping, to see friends or family from another household, visiting hospitality venues and for work was also created, but only calculated for those who reported being in employment.
Wearing a face covering: Participants were asked if they had worn a professional face mask or a homemade, cloth or improvised face covering (such as a scarf) when out and about in the last 7 days on a five-point scale from “never” to “very frequently.” Between 27 April 2020 (wave 14) to 13 May 2020 (wave 16), this question was only asked to half the sample. From 18 May 2020 (wave 17), all participants were asked to respond on the five-point scale. We coded participants as wearing a face covering if they selected “very frequently” or “frequently” for either or both of the face covering items. On 26 October 2020 (wave 31), the two items were replaced by a single item asking how frequently participants had “worn a face mask or another face covering (such as a scarf) when out and about”. This item was removed from the survey after 19 May 2021 (wave 50). For these items, we treated answers of “not applicable” as not having worn a face covering frequently or very frequently (n = 432, 0.8%).
Participants were asked to report their sex, age, ethnicity, highest level of educational or professional attainment, employment status and socio-economic grade . Region was derived from participants’ postcodes.
This work was conducted as part of a service evaluation of the marketing and communications run by the Department of Health and Social Care, and so did not require ethical approval. We sought advice from the Psychiatry, Nursing and Midwifery Research Ethics Office, King’s College London and they confirmed this position.
For analyses investigating hand cleaning, we used data from all participants completing the survey between 10 February 2020 and 19 May 2021 (waves 3 to 50), as these were the dates that the relevant outcome question was included (n = 95,998 responses).
Due to differences in rules on outings in the devolved nations, we restricted the sample to include only participants living in England (included in surveys between 30 March 2020 and 20 January 2022 [waves 10 to 66], n = 100,736 responses) for analyses of outings. For the number of outings for work, we restricted analyses to those who reported they were in full-time, part-time or self-employment (n = 54,504 responses).
For analyses of wearing a face covering, due to differences in rules in the devolved nations, we restricted the sample only to those in England who reported having been out shopping in the last 7 days (included in survey between 27 April 2020 and 19 May 2021 [waves 14 to 50], n = 52,285 responses).
We described engagement with protective behaviours descriptively and graphically, using line graphs to show the percentage of people who reported engaging in hand cleaning frequently or very frequently, the number of times people reported going out in the last week for various reasons, and the percentage of people who reported wearing a face covering frequently or very frequently. For each data point, we calculated 95% confidence intervals.
For analyses investigating hand washing and wearing a face covering, phrasing of the question changed on 26 October 2020 (wave 31). We compared data from 28 September to 11 November 2020 (waves 29 to 32; last two waves of the original compared to first two waves of the updated question wording; no repeat respondents) using a χ2-test, to investigate whether responses for the different question wordings were comparable.
We used generalised estimating equations (GEEs) to investigate whether engagement with protective behaviours differed by survey wave, accounting for repeat respondents where possible. For binary outcomes (hand washing, wearing a face covering), we used logistic regression analyses, reporting odds ratios (ORs). For count data (total outings), we used negative binomial regressions, reporting incidence rate ratios (IRRs). We also investigated whether total outings differed between lockdown periods. As participants reported on their behaviour in the previous 7 days, we selected waves where the reporting period was solely contained under lockdown restrictions (first lockdown: 30 March to 6 May 2020, waves 10 to 15; second lockdown: 16 November to 2 December 2020, waves 33 to 35; third lockdown: 25 January to 24 February 2021, waves 42 to 44).