In this large community-based survey we found strong evidence for a protective effect of wearing a face covering in reducing the risk of COVID-19 infection coming into the household, and strong evidence that being in crowded places was associated with an increased risk. We also found moderate evidence for a protective effect from social distancing and handwashing when arriving home. We found no evidence of a protective effect from the other non-pharmaceutical interventions under investigation, including those who reported handwashing before eating, cleaning things that might have virus on them, avoiding touching others’ pets, and other behaviours, such as taking alternative treatments.
Our findings on wearing a face covering, social distancing, and avoiding crowded places, are consistent with current evidence about airborne transmission of SARS-CoV-2,(10) and are already widely accepted as useful public health measures. Our finding of the strongest effects for wearing a face covering are consistent with other evidence on the effects of face masks.(11) However, most experts believe that fomite transmission plays a minimal role in transmission,(12, 13) and therefore our finding of an association between handwashing on arriving home and reduced COVID-19 infection is interesting. The increased odds of COVID-19 infection associated with many of the NPIs under investigation in this study was an unexpected and surprising finding. We have not been able to identify a plausible biological mechanism through which these NPIs could increase the risk of infection. These findings most likely result from bias or uncontrolled confounding. We believe the most likely cause is recall bias – participants who had an infection were more likely to perceive increased use of a NPI than those who did not. However, if this is the case then it is likely to affect all the NPIs that we asked about, so the reduced odds of infection associated with wearing a face covering, social distancing, handwashing when arriving home and avoiding crowds, are likely to be under-estimates of the true effects. We found approximately a one-third reduction in risk of COVID-19 for each increase in level of use of face coverings (never, sometimes, quite often, very often, always). This is a large effect, so if it is an underestimate, the true effect would be even larger.
A recent systematic review and meta-analysis of public health measures for preventing COVID-19,(14) which included 72 studies, reported a 53% reduction in risk of COVID-19 incidence associated with mask wearing (relative risk 0.47, 0.29 to 0.75) and a 25% reduction from physical distancing (RR 0.75, 0.59 to 0.95). Our estimates suggest slightly greater effect, ranging from 83–85% reduction (OR 0.17, 0.15 to 0.20 in unadjusted analysis and 0.15, 0.12 to 0.19 in analysis adjusted for other NPI, vaccination status, month of questionnaire completion, age group, gender, ethnicity and SES) for mask wearing and 43–73% reduction (OR 0.57, 0.41 to 0.80 in fully adjusted analysis and 0.27, 0.22 to 0.31 in unadjusted analysis) for social distancing. The same review found a non-significant 53% reduction (relative risk 0.47, 0.19 to 1.12) for handwashing.(14) This is comparable to our primary analysis estimates for handwashing on arriving home (OR 0.57, 0.46 to 0.73 (unadjusted) and 0.72, 0.47 to 1.11 (fully adjusted)). It is worth noting that our effect estimates for all of these NPIs showed evidence of a dose-response effect, with increases in effect for each increase in the frequency of use.
An ecological study looking at transmission rates and public health measures in 190 countries found that the largest reduction in time-varying effective reproduction number (Rt) was associated with social distancing (− 42.94%, − 44.24% to − 41.60%), while smaller, but still important, reductions were associated with mandatory use of face masks (− 15.14%, − 21.79% to − 7.93%) and quarantine policies (− 11.40%, − 13.66% to − 9.07%) .(15)
Our finding that handwashing when arriving home appears to be associated with reduced risk, but handwashing before meals is associated with increased (or more likely no reduction in risk), is interesting and suggests that fomite transmission is important when people travel outside of their home. Transmission of COVID-19 does frequently occur within households. Our analysis was not designed to look specifically at household transmission and so excluded people with a household contact, which is where any preventive effect of handwashing before meals is likely to occur.(16) However, our data would suggest that washing hands when arriving home is more likely to be effective at reducing the risk of COVID-19 coming into the household. We are not aware of any other studies that have compared the effects of handwashing at these different times.
A recent narrative review focusing on factors that influence engagement with NPIs found that women, more highly educated people, older people, married people, and those with worse self-rated health were more likely to engage with use of face masks.(17) Our analyses adjusted for most of these factors, although we did not adjust for marital status or ‘self-rated health’. More generally, there was widespread acceptance of the need to adopt NPIs. Perceived severity of the pandemic and personal risk were key factors influencing willingness to adhere. Interestingly, a narrative review conducted several years before the pandemic found greater perceived willingness to accept measures like handwashing and ‘respiratory hygiene’ than mask wearing and personal distancing.(18)
Strengths of our study include our broad and inclusive approach to recruitment, large number of participants, rigorous criteria for defining COVID-19 illness, and detailed data on sociodemographic and medical factors that could be controlled for in the model. Community testing for COVID-19 was virtually non-existent in the UK at the start of the pandemic, and only became widely available in the autumn of 2020. Therefore, it was not possible to base our case definition only on positive test results for our primary analysis. However, we did take a rigorous approach, using a respiratory illness with a positive swab test result for COVID-19 OR a respiratory illness during the pandemic that was associated with both fever AND loss of smell or taste for our primary analysis, and conducting a secondary analysis using only those with a positive test result as cases.
The main weakness of our study is that we used retrospective self-reported data. Almost all studies of public health measures are observational and use self-report data as it is very difficult to randomise people to follow (and continue to adhere to) different public health measures. We have already discussed the risk of recall bias associated with this approach, but we believe that recall bias is unlikely to explain the reduced risk of COVID-19 in those who described handwashing on arriving home, wearing a face mask and social distancing, or the difference in risk between handwashing on arriving home and handwashing before meals. Confounding is the other major risk associated with observational studies such as this. However, we were able to measure and adjust for all key known confounders and many other potential confounders. Although we cannot exclude some residual confounding, this is unlikely to explain the large effects observed in our study. The “other behaviours” in this analysis includes a heterogeneous group of behaviours, including exercise, nutritional supplements and herbal remedies. It is possible that some of these behaviours may be associated with reduced odds of COVID-19 when explored individually.