key: cord-1054619-am3u7z76 authors: Waller, J.; Rubin, G. J.; Potts, H. W. W.; Mottershaw, A.; Marteau, T. M. title: Immunity Passports for SARS-CoV-2: an online experimental study of the impact of antibody test terminology on perceived risk and behaviour date: 2020-05-10 journal: nan DOI: 10.1101/2020.05.06.20093401 sha: 3b2dce36dbde9481453170d7f89721a59ed9d6dc doc_id: 1054619 cord_uid: am3u7z76 Objective: To assess the impact of describing an antibody-positive test result using the terms Immunity and Passport or Certificate, alone or in combination, on perceived risk of becoming infected with SARS-CoV-2 and intention to continue protective behaviours. Design: 2 by 3 experimental design. Setting: Online with data collected between 28th April and 1st May 2020. Participants: 1,204 adults registered with a UK research panel. Intervention: Participants were randomised to receive one of six descriptions of an antibody test and results showing SARS-CoV-2 antibodies, differing in the terms used to describe the type of test (Immunity vs Antibody) and the test result (Passport vs Certificate vs Test). Main outcome measures: The primary outcome was the proportion of participants perceiving no risk of becoming infected with SARS-CoV-2 given an antibody positive test result. Other outcomes include intended changes to frequency of hand washing and physical distancing. Results: When using the term Immunity (vs Antibody), 19.1% of participants [95% CI: 16.1 to 22.5] (vs 9.8% [95% CI: 7.5 to 12.4]) perceived no risk of catching coronavirus at some point in the future given an antibody-positive test result (AOR: 2.91 [95% CI: 1.52 to 5.55]). Using the terms Passport or Certificate, as opposed to Test, had no significant effect (AOR: 1.24 [95% CI: 0.62 to 2.48] and AOR: 0.96 [95% CI: 0.47 to 1.99] respectively). There was no significant interaction between the effects of the test and result terminology. Across groups, perceiving no risk of infection was associated with an intention to wash hands less frequently (AOR: 2.32 [95% CI: 1.25 to 4.28]) but there was no significant association with intended avoidance of physical contact with others outside of the home (AOR: 1.37 [95% CI: 0.93 to 2.03]). Conclusions: Using the term Immunity (vs Antibody) to describe antibody tests for SARS-CoV-2 increases the proportion of people believing that an antibody-positive result means they have no risk of catching coronavirus in the future, a perception that may be associated with less frequent hand washing. The way antibody testing is described may have implications for the likely impact of testing on transmission rates. At the height of the first wave of the COVID-19 pandemic, about a third of the world's population is estimated to have been in lockdown, with all but essential workers largely confined to home (1) . Without an effective treatment or vaccine, testing for infection combined with contact tracing and isolation will be central to effective strategies to ease populations out of lockdown while keeping the basic reproduction number (R0) below one (2) . Testing for antibodies to SARS-CoV-2 is a possible complement to testing for active infection to identify those who have developed antibodies to the virus and so may be able to return to work and other activities without significantly increasing transmission rates (3) . These tests have been variously described in the media as Immunity Passports (4, 5) , Immunity Certificates (6,7) Immunity Cards (8) and Release Certificates (9) . Unfortunately, the use of these terms implies a certainty unmatched by current evidence about antibody tests (10) . Uncertainties inherent in tests for antibodies to SARS-CoV-2 include the extent and duration of immunity conferred (11) . They also include the uncertainties inherent in any test regarding the proportion of those who would be correctly identified. This depends upon the test performance -its sensitivity and specificity -as well as the population prevalence of the tested condition (12) . Given these uncertainties, those who receive a test result indicating the presence of antibodies will have a residual risk of becoming infected by SARS-CoV-2 in the future. Understanding that there is this residual risk -albeit one that is difficult to quantify at present -will be important to minimise transmission that could arise from those receiving "antibody positive" test results. If people testing positive perceive that they have no risk of becoming infected by the virus, they may ignore any future symptoms of infection and facilitate transmission if they fail to self-isolate appropriately. Such a perception may also overgeneralise to a belief that they are unable to transmit infection through contact with contaminated surfaces. Regardless of antibody status, all individuals can indirectly transmit the virus between surfaces by touch. Hand washing or sanitising therefore need to remain frequent. Evidence from other testing programmes suggests that interpreting a low risk result to mean no risk can be reduced by verbal and numerical expressions of residual risk when presenting test results (13, 14) . But even before testing programmes are in place, the terms commonly used to describe these tests -Immunity Passport or Certificates -may inadvertently be fuelling a misplaced sense of certainty about their results. It is unknown whether describing these tests as being for immunity -as opposed to antibodies -or their results as passports or certificates increases misunderstanding of the residual risk inherent in an antibody-positive . CC-BY-NC-ND 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 May 10, 2020. . https://doi.org/10.1101/2020.05.06.20093401 doi: medRxiv preprint test result and thereby reducing adherence to protective behaviours and increasing risk of transmission (15) . This study was designed to test two hypotheses: describing a test indicating the presence of antibodies using the term Immunity (vs Antibody), and describing test results as Passports or Certificates (vs Test), increases the likelihood that those with this test result erroneously perceive they have no risk of becoming infected in the future with coronavirus. Ethical approval for this study was granted by the King's College London Research Ethics Committee (reference: MRA-19/20-18685). The protocol was preregistered on the Open Science Framework https://osf.io/tjwz8/ Study 2 The statistical analysis plan was pre-specified and uploaded to the Open Science Framework prior to receipt of the data https://osf.io/tjwz8/ Study 2 An initial study with similar methods was conducted https://osf.io/tjwz8/ Study 1 but, due to an error, the intervention was not correctly programmed. This study is therefore not reported. The study was an online experiment using a 2 × 3 factorial design, with participants randomised, with an equal allocation ratio, to one of six groups varying in the description of an antibody test and a result showing the presence of antibodies. These descriptions differed only in the term used for what was being tested (Immunity vs Antibody) and the term used for the test result (Passport vs Certificate vs Test). A quota sample of 1,204 adults was recruited via Predictiv, the Behavioural Insights Team's online experimentation platform (https://www.bi.team/bi-ventures/predictiv/) comprising 500,000 adults in the UK. Quotas were based on age, gender and UK region to achieve a sample broadly representative of the UK population. 1373 clicked on the link to enter the study of whom 1214 subsequently completed the study. Ten were excluded for failing to meet quality checks. Participants were reimbursed in points (equivalent to £1) which could be redeemed in cash, gift vouchers or charitable donations. Participants did not know the topic of the study prior to participation. Due to the rapid nature of this research, the public was not involved in the development of the study. The sample size was chosen pragmatically without reference to a specific power calculation. We are fitting a full model with an interaction. Conservatively, we then had an 80% chance of detecting, at a 5% significance level, an increase in the primary outcome measure from 50% in a baseline group to 64% in another group. Participants were randomised to groups by random number generation. A random number between 1 and 6 was generated for every participant upon entry to the study to determine which description they saw, with each of the six numbers corresponding to one of the six description. As this is based on true randomness, the number of participants within each group can vary due to chance. The intervention comprised a description of antibody testing and test results indicating the presence of antibodies [see Box 1 for one example and S1 for wording of all six descriptions]. These differed across six groups in test name of results indicating the presence of antibodies. All descriptions included the information that the result would mean a lower risk of future infection and transmission, and that people with this result could return to work earlier. Wording of the items used for each measure is shown in Supplementary Materials (S2). Scientists are developing tests to see who has already had coronavirus. No test is 100% effective. This means that those who test 'positive' would have: • Lower risk of catching coronavirus in the future, and therefore also • Lower risk of passing it on to others Those who test 'positive' would get an immunity passport. They could return to work early. . CC-BY-NC-ND 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 May 10, 2020. . https://doi.org/10.1101/2020.05.06.20093401 doi: medRxiv preprint Primary outcome Proportion of participants perceiving an antibody-positive test result to mean no risk of catching coronavirus in the future, assessed in response to a question with four response options. Perceived likelihood of catching coronavirus in the future, assessed on a visual analogue scale from 0% to 100%. Intention to engage in handwashing less or more frequently than now, given an antibodypositive test result: assessed in response to a question with five response options. Intention to avoid physical contact with others outside the home more or less frequently than now, given an antibody-positive test result: assessed in response to a question with five response options. Interest in undergoing the test if offered today: assessed in response to a question with four response options. Demographic characteristics: age, gender, level of education and geographical region of residence. Employment status, planned to be included, was omitted due to a technical error. A detailed statistical analysis plan is available on the Open Science Framework, specified prior to receipt of the data https://osf.io/tjwz8/ Study 2. Binary logistic regression was used to assess the impact of test type (immunity/antibody) and result type (passport/certificate/test) on the odds of believing the antibody test result means there is no risk of future infection. An interaction term was included in the model (16) . The analysis was repeated adjusting for age (including a quadratic function to model a non-linear relationship), gender, education and region based on prior results showing these are predictors of risk beliefs. Binary logistic regressions were run (as above) for the secondary outcomes: intention to wash hands less, intention to engage less in social distancing and intention to undergo the test. Unadjusted and adjusted odds ratios and 95% confidence intervals are reported. Logistic regression was run to assess the extent to which intentions to engage in less frequent handwashing or social distancing measures is predicted by perceiving the test result to mean no risk of being infected in the future by coronavirus. As only a very small proportion of participants gave a 'zero' response on the sliding scale of future risk, we used a linear regression model to examine this outcome, rather than a binary (zero vs. other) logistic regression as pre-specified in the analysis plan. . CC-BY-NC-ND 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 May 10, 2020. . https://doi.org/10.1101/2020.05.06.20093401 doi: medRxiv preprint Data were collected using an online survey platform, Predictiv. Upon entry to the study, participants were informed that they were to be asked some questions about coronavirus and that it would take about five minutes to complete. Participants were then shown one of six brief descriptions of an antibody test for coronavirus (see S1 for full text for each of the six descriptions). They were then asked five questions, assessing the primary and secondary outcomes. Participants' demographic characteristics were accessed from the survey platform. The sample comprised 606 women and 598 men with a median age of 36 years. Around a quarter had some graduate-level education (24.2%) and there was good representation of all UK regions (see Table 1 ). Distribution of sample characteristics by exposure group is shown in Table 1 . Responses to the five outcome questions for the whole sample and by experimental group are shown in Table 2 Perceived level of future risk (on a scale of 0 to 100%) showed a complex, trimodal distribution. The median was 35% with an interquartile range from 18% to 51%. Only 5% of respondents put their risk at 0%. Overall, 63% put their risk below 50%. 10% put their risk at 50%, which was the modal response. 24% put their risk at greater than 50%, but below 100%. 3% of respondents put their risk at 100%: that is, they said they were certain to contract the virus. . CC-BY-NC-ND 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 May 10, 2020. Proportion answering 'no risk' in each sub-group % (95% CI) (n=1204) Odds ratio (95% confidence interval) Mutually adjusted (n=1204) Figure 1 ]. There was no significant effect of result type and no significant interaction. We analysed the continuous measures of future perceived risk of infection using a linear model (ANOVA) with two levels for test type, three levels for result type, and an interaction term. Overall, there was no significant effect: F5,1198 = 1.46, p = 0.20, adjusted R2 < 1%. We repeated the analysis adjusting for demographic factors as covariates. Overall, there was a significant effect: F13,1165 = 1.88, p = 0.03, adjusted R2 = 1%. This was because of a significant effect of age: as age increased, perceived risk decreased. There remained no significant effect of the experimental variables. Logistic regression analyses examining the impact of Test type, Result type and their interaction on intentions to wash hands and avoid physical contact less frequently and on willingness to have the test are shown in Supplementary tables 1 and 2. Neither Test type, Result type nor their interaction were significantly associated with these behavioural outcomes. Logistic regression analyses were used to examine belief that the result meant 'no risk' as a predictor of intention to wash hands and avoid physical contact less frequently, given a positive result (see Figure 2 and Supplementary . CC-BY-NC-ND 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 May 10, 2020. Using the term Immunity -as opposed to Antibody -to describe antibody tests for SARS-CoV-2 doubled the proportion who erroneously perceived they would have no risk of becoming infected with the virus in the future if they were given an antibody-positive test result, from 9.8% for Antibody to 19 These was no significant association with intended frequency of avoiding physical contact with others outside of the home. Interest in undergoing the test was high -with 85.2% saying they would probably or definitely have it if offered -and was unaffected by the terms used to describe the tests. This study was designed to test two hypotheses, providing strong support for the first, that describing a test indicating the presence of antibodies using the term Immunity (vs Antibody) increases the likelihood that those with this test result erroneously perceive they have no risk of becoming infected in the future with coronavirus. This likely reflects a certainty about risk of future infection implicit in lay understandings of the term immunity that is not implied by the term antibody (17) . Qualitative studies could explore this and other potential mechanisms for the effect observed. The results of this study did not support the second hypothesis that describing test results as Passports or Certificates increases the likelihood that those with this test result erroneously perceive they have no risk of becoming infected in the future with coronavirus. This does not mean that these terms are unproblematic however, only that they did not influence the specific perceptions that we explored. Qualitative studies are warranted to understand the broader meanings these terms have in the context of testing for antibodies for SARS-CoV-2 and other contexts. Responses on the sliding scale of future risk showed a high variability and were largely unexplained by the experimental intervention or other variables measured. This may point to considerable uncertainty in the public as to how to interpret test results. It also likely reflects the well-described tendency of people to use a 50% response to indicate uncertainty rather than a true judgement of probability (18) . We also saw that about a quarter of respondents on the first question stated their risk was "average" or "higher". This may point to considerable uncertainty in the public as to how to interpret test results Use of the top end of the scale is hard to interpret but may either reflect a failure to read the information carefully and . CC-BY-NC-ND 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 May 10, 2020. . https://doi.org/10.1101/2020.05.06.20093401 doi: medRxiv preprint therefore a misunderstanding of the meaning of the result, or participants' using information beyond the experiment to assess their risk and not adequately considering the hypothetical test result when making their response. While we found no evidence for a direct effect on protective behaviours of the terms used to describe antibody tests results, there was indirect evidence that perceiving no risk of future infection might reduce frequency of handwashing. This finding is tentative, given it is based on behavioural intentions in response to a hypothetical antibody-positive result. Nonetheless the potential for antibody testing to increase viral transmission must be considered alongside the potential benefits the tests might have in allowing the easing of lockdown restrictions. Clear communication about the ongoing need for handwashing, in particular, will be essential and raising public awareness of the main mechanisms through which SARS-CoV-2 is transmittedthrough air and surfacesmight help improve adherence. This, in addition to acknowledgment of the imperfect nature of the tests, will give the public a more accurate representation of the meaning and implications of an antibody test result and a better understanding of how to reduce the risk of transmission. Such communications need to emphasise that transmission can occur through contact regardless of antibody status. Such communications also need to be rigorously evaluated to ensure their effectiveness at communicating these points both to those undergoing antibody tests as well as to general populations that are now having to learn to live with SARS-CoV-2. This study provides the first experimental evidence for the potentially adverse impact on risk perceptions and protective behaviours of commonly used terms to describe SARS-CoV-2 antibody tests and their results. As such, it provides timely evidence to inform policy and research to mitigate these effects to realise the potential benefits of such tests. The study has several limitations. First, participants were responding to a hypothetical test and asked to imagine that they had received a test result that had detected antibodies. Findings from such studies can generalise to clinical settings (19, 20) but some caution is warranted. Second, the protective behaviours of handwashing and physical distancing were measured using single items assessing behavioural intentions following a hypothetical test result. Third, the sample size was insufficient to detect effect sizes that could be important at a population level. It is possible, for example, that the use of the terms Certificate or Passport might impact on risk perception, but the current study lacked the power to detect this. Fourth, while quotas were used to achieve a sample broadly representative of the UK population, research panels are not representative of the general population (21, 22) . We found no evidence that the impact of the interventions in this study was modified by demographic characteristics of the participants, providing some reassurance about the generalisability of results across age groups, gender, educational level and geographical region of the UK. . CC-BY-NC-ND 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 May 10, 2020. The results of this study have several implications for research and policy. The effectiveness of antibody tests for SARS-CoV-2 will depend not only on the extent and duration of any immunity conferred and the performance of a test, but also upon a good understanding of the meaning of tests results among those offered them. First, the use of the term Immunity should be avoided in phrases to describe antibody tests, whether described as Passports, Certificates or Tests. Second, research is needed to evaluate different ways of informing those offered tests and receiving tests results to minimise the proportion erroneously perceiving an antibody-positive test result to mean no risk of becoming infected with the virus. It should also focus on maximising understanding that -regardless of antibody-status -anyone can indirectly transmit the virus by touching a contaminated surface and infecting the next surface they touch. Hand washing or sanitising therefore need to remain frequent. Research is also needed with those undergoing actual tests, powered to detect effects judged meaningful in the context of a population-based testing programme and involving measures of actual behaviour. Interest in SARS-CoV-2 antibody testing is high -across many countries, employers and populations. While such testing could contribute to wider strategies to ease lock-down restrictions, their use may have an adverse impact on transmission-related behaviour. This appears to vary with the way the tests are described. Using the term Immunity (vs Antibody) to describe antibody tests increases the proportion of people believing that an antibodypositive result means they have no future risk of coronavirus, a perception that may be associated with less frequent handwashing and hence increased risk of transmission. . CC-BY-NC-ND 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 May 10, 2020. 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 May 10, 2020. Scientists are developing tests to see who has already had coronavirus. No test is 100% effective. This means that those who test 'positive' would have: • Lower risk of catching coronavirus in the future -and therefore also • Lower risk of passing it on to others Those who test 'positive' would get an immunity passport. They could return to work early. Scientists are developing tests to see who has already had coronavirus. No test is 100% effective. This means that those who test 'positive' would have: • Lower risk of catching coronavirus in the future -and therefore also • Lower risk of passing it on to others Those who test 'positive' would get an immunity certificate. They could return to work early. . CC-BY-NC-ND 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 May 10, 2020. Scientists are developing tests to see who has already had coronavirus. No test is 100% effective. This means that those who test 'positive' would have: • Lower risk of catching coronavirus in the future -and therefore also • Lower risk of passing it on to others Those who test 'positive' would get a result showing immunity. They could return to work early. Scientists are developing tests to see who has already had coronavirus. No test is 100% effective. This means that those who test 'positive' would have: • Lower risk of catching coronavirus in the future -and therefore also • Lower risk of passing it on to others Those who test 'positive' would get an antibody passport. They could return to work early. Scientists are developing tests to see who has already had coronavirus. No test is 100% effective. This means that those who test 'positive' would have: • Lower risk of catching coronavirus in the future -and therefore also • Lower risk of passing it on to others Those who test 'positive' would get an antibody certificate. They could return to work early. . CC-BY-NC-ND 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 May 10, 2020. Scientists are developing tests to see who has already had coronavirus. No test is 100% effective. This means that those who test 'positive' would have: • Lower risk of catching coronavirus in the future -and therefore also • Lower risk of passing it on to others Those who test 'positive' would get a result showing antibodies. They could return to work early. . CC-BY-NC-ND 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 May 10, 2020. Oxford COVID-19 Government Response Tracker, Blavatnik School of Government World Health Organization. Critical preparedness, readiness and response actions for COVID-19. World Health Organization Disease Control, Civil Liberties, and Mass Testing -Calibrating Restrictions during the Covid-19 Pandemic Immunity passports' could speed up return to work after Covid-19. The Guardian Delta's CEO said he would support an 'immunity passport' program or other steps to jumpstart travel as the airline reports its first quarterly loss in more than 5 years. Business Insider People could be given coronavirus 'immunity certificates' to leave lockdown early. The Independent Mass Coronavirus Antibody Tests Have Serious Limits. Bloomberg.com [Internet Coronavirus immunity cards for Americans are 'being discussed'. POLITICO Chile to push ahead with coronavirus 'release certificates' despite WHO warning. Reuters Evaluation of antibody testing for SARS-Cov-2 using ELISA and lateral flow immunoassays. MedRxiv Prepr What policy makers need to know about COVID-19 protective immunity. The Lancet Understanding sensitivity and specificity with the right side of the brain Numbers or words? A randomized controlled trial of presenting screen negative results to pregnant women Women's understanding of a "normal smear test result": experimental questionnaire based study Immunity passports' in the context of COVID-19 Factorial versus multi-arm multi-stage designs for clinical trials with multiple treatments Fuzzy' virus: indeterminate influenza biology, diagnosis and surveillance in the risk ontologies of the general public in time of pandemics Verbal and Numerical Expressions of Probability: 'It's a Fifty-Fifty Chance'. Organ Behav Hum Decis Process The impact of genetic testing for Crohn's disease, risk magnitude and graphical format on motivation to stop smoking: an experimental analogue study Effect of communicating DNA based risk assessments for Crohn's disease on smoking cessation: randomised controlled trial Office for National Statistics We thank Steve Reicher for comments on an earlier draft of the study protocol . CC-BY-NC-ND 4.0 International license It is made available under a All authors have completed the Unified Competing Interest form (available on request from the corresponding author) and declare: no support from any organisation for the submitted work; and no financial relationships with any organisations that might have an interest in the submitted work in the previous three years. HWWP declares consultancy fees from Babylon Health; all authors declare no other relationships or activities that could appear to have influenced the submitted work. The authors affirm that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as originally planned have been explained. Anonymised data will be made available upon reasonable request. The study was conceptualised by TMM, JW & GJR. AM completed data collection. JW & HP analysed the data. All authors contributed to, and approved, the final manuscript.