‘Immunity Passports’ for SARS-CoV-2: an online experimental study of the impact of antibody test terminology on perceived risk and behaviour

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 severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and protective behaviours. Design 2×3 experimental design. Setting Online. Participants 1204 adults from 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 describing the type of test (Immunity vs Antibody) and the test result (Passport vs Certificate vs Test). Main outcome measures Primary outcome: proportion of participants perceiving no risk of infection 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 given an antibody-positive test result (adjusted OR (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)); there was no significant association with intended avoidance of physical contact (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. Trial registration number Open Science Framework: https://osf.io/tjwz8/files/

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 (R 0 ) 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 over-generalise 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 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.

Design
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).

Participants
A quota sample of 1,204 adults was recruited via Predictiv, the Behavioural Insights Team's online experimentation platform (https://www.bi.team/biventures/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.

Patient and public involvement
Due to the rapid nature of this research, the public was not involved in the development of the study.

Power
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.

Randomisation
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.

Intervention
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

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.

Secondary outcomes
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 antibody-positive test result: assessed in response to a question with five response options.

Immunity Passport
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. Interest in undergoing the test if offered today: assessed in response to a question with four response options.

Other measures
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.

Statistical analyses
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. 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

Procedure
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.

Sample characteristics
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.

Descriptive statistics Primary outcome
Responses to the five outcome questions for the whole sample and by experimental group are shown in Table 2. Overall, 14.5% of respondents [95% CI: 12.5 to 16.6] interpreted the test result as meaning they had no risk of future infection. Over half (57.8% [95% CI: 55.0 to 60.1]) correctly interpreted the test result as meaning that their future risk of coronavirus was 'lower'.

Secondary outcomes
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%.

Discussion
Using the term Immunity -as opposed to Antibody -to describe antibody tests for  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y  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 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 transmitted -through air and surfaces -might 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.

Strengths and Limitations
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. 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.

Implications for research and policy
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

Conclusion
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 lockdown 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 antibody-positive 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.

Competing interests
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.

Transparency declaration
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.

Data sharing statement
Anonymised data will be made available upon reasonable request.

Author contribution statement
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.

Immunity Passport
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.

Immunity Certificate
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.

Immunity Test
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.

Antibody Passport
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.

Antibody Certificate
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.

Antibody Test
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. how likely is it that you will get coronavirus at some point in the future?
Please answer on a scale from 0 % to 100% where 0% means no chance and 100% means certain.

III Imagine that you were given [an immunity passport/an immunity certificate/a result showing immunity/an antibody passport/an antibody certificate/a result showing antibodies]
would you wash your hands with soap and water or sanitiser:

Strengths and limitations of this study
 This study provides the first experimental evidence for the potentially adverse impact on risk perceptions and protective behaviours of using the term Immunity -as opposed to Antibody -to describe SARS-CoV-2 antibody tests and their results.  It provides timely evidence to inform policy and research to mitigate these effects to realise the potential benefits of such tests.  Participants were responding to a hypothetical test and asked to imagine that they had received a test result that had detected antibodies.  Protective behaviours of handwashing and physical distancing were measured using single items assessing behavioural intentions.  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 (R 0 ) 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 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 test result and thereby reducing adherence to protective behaviours and increasing risk of transmission (15 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.

Design
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).

Participants
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 (e.g. the unique participant identifier was missing, or the participant had entered the survey more than once). 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. All participants had provided opt-in consent to take part in surveys when they signed up to the online panel. No identifiable data were collected as part of the survey. All data collected by Predictiv are processed in line with their published privacy policy (http://www.predictiv.co.uk/privacy-policy.html).

Patient and public involvement
Due to the rapid nature of this research, the public was not involved in the development of the study.

Power
The sample size was chosen pragmatically without reference to a specific power calculation. We fitted a full linear model with two levels for test type (immunity/antibody), three levels for result type (passport/certificate/test) and an interaction term. 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. The power calculation was done using http://powerandsamplesize.com/, taking 50% as the baseline to provide a conservative estimate of power.

Randomisation
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 (see Table 1 for numbers in each group).   Table 2.

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.

Immunity Passport
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. Intention to engage in handwashing less or more frequently than now, given an antibody-positive 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.

Other measures
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.

Statistical analyses
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. 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.

Procedure
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.

Sample characteristics
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.

Descriptive statistics
Primary outcome Responses to the five outcome questions for the whole sample and by experimental group are shown in Table 2 Figure 1]. There was no significant effect of result type and no significant interaction.

Between-group differences in secondary outcomes
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: F 5,1198 = 1.46, p = 0.20, adjusted R 2 < 1%. We repeated the analysis adjusting for demographic factors as covariates. Overall, there was a significant effect: F 13,1165 = 1.88, p = 0.03, adjusted R 2 = 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.

Association between test result meaning 'no risk' and behavioural intentions
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

Discussion
Using the term Immunity -as opposed to Antibody -to describe antibody tests for 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.
They also showed only broad consistency with responses to the main outcome measure; some participants who believed the result meant 'no risk' then rated their future risk as 100%, and vice versa. This may point to considerable uncertainty in the public as to how to interpret test results. Some participants may have been responding solely on the information provided; others may have been using other information about coronavirus and immunity to make judgements about their own risk. 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 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 transmitted -through air and surfaces -might 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.
It was notable that 85% of respondents said that they would probably or definitely have an antibody test if they were offered it. This suggests high interest but should be treated with caution not least because fewer undergo than express interest in undergoing tests (19).

Strengths and Limitations
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. The extent to which people responded using the information provided in the study is unknown. Some will likely have used other information about SARS-CoV-2 and their own experiences to answer the questions about risk. This is a limitation of all hypothetical scenario-based studies of this type but findings from such studies can generalise to clinical settings (20,21) although some caution is warranted. As antibody testing is not yet routinely being carried out in the UK, the impact of the wording used in the current study amongst those undergoing testing cannot be assessed. All scenarios used the term 'positive' so we were unable to examine any possible impact of the use of this term. In addition, due to the rapidity with which the current study was conducted, piloting of the materials in general population samples was not carried out. It is therefore possible that some of the information provided may have been misinterpreted.
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 (22,23). 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.

Implications for research and policy
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. This has implications for the presentation of antibody testing by policy-makers and the media,

Conclusion
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 lockdown 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 antibody-positive 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.

Transparency declaration
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.

Data sharing statement
Anonymised data will be made available upon reasonable request.

Immunity Passport
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.

Immunity Certificate
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.

Immunity Test
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.

Antibody Passport
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.

Antibody Certificate
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.

Strengths and limitations of this study
 This study provides the first experimental evidence for the potentially adverse impact on risk perceptions and protective behaviours of using the term Immunity -as opposed to Antibody -to describe SARS-CoV-2 antibody tests and their results.  It provides timely evidence to inform policy and research to mitigate these effects to realise the potential benefits of such tests.  Participants were responding to a hypothetical test and asked to imagine that they had received a test result that had detected antibodies.  Protective behaviours of handwashing and physical distancing were measured using single items assessing behavioural intentions.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  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 (R 0 ) 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). But whether these terms actually encourage a misplaced sense of certainty even before testing is widely available is unknown and the focus of the current study.
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.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  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 over-generalise 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 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.

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.

Design
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).

Participants
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 (e.g. the unique participant identifier was missing, or the participant had entered the survey more than once). 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. All participants had provided opt-in consent to take part in surveys when they signed up to the online panel. No identifiable data were collected as part of the survey. All data collected by Predictiv are processed in line with their published privacy policy (http://www.predictiv.co.uk/privacy-policy.html).

Patient and public involvement
Due to the rapid nature of this research, the public was not involved in the development of the study.

Power
The sample size was chosen pragmatically without reference to a specific power calculation. We fitted a full linear model with two levels for test type (immunity/antibody), three levels for result type (passport/certificate/test) and an interaction term. 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. The power calculation was done using http://powerandsamplesize.com/, taking 50% as the baseline to provide a conservative estimate of power.

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.

Immunity Passport
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. Intention to engage in handwashing less or more frequently than now, given an antibody-positive 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.

Other measures
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.

Statistical analyses
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 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.

Procedure
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. 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.

Descriptive statistics
Primary outcome Responses to the five outcome questions for the whole sample and by experimental group are shown in Table 2 Figure 1]. There was no significant effect of result type and no significant interaction.

Association between test result meaning 'no risk' and behavioural intentions
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

Discussion
Using the term Immunity -as opposed to Antibody -to describe antibody tests for  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 (18,19). Some participants may have been responding solely on the information provided; others may have been using other information about coronavirus and immunity to make judgements about their own risk. 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 (20). We  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  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 participants' not reading the information carefully and therefore misunderstanding the meaning of the result being communicated in the information, or their using information beyond the experiment to assess their risk. Participants likely considered many factors in making sense of the hypothetical result, such as their actual risk of contracting the virus based on their behaviour and likely exposure.
Only 7% of the sample responded that their risk would be higher on the primary outcome measure. Reasons for this are unclear.
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 transmitted -through air and surfaces -might 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.
It was notable that 85% of respondents said that they would probably or definitely have an antibody test if they were offered it. This suggests high interest but should be treated with caution not least because fewer undergo than express interest in undergoing tests (21).

Strengths and Limitations
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.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y 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. The extent to which people responded using the information provided in the study is unknown. Some will likely have used other information about SARS-CoV-2 and their own experiences to answer the questions about risk. This is a limitation of all hypothetical scenario-based studies of this type but findings from such studies can generalise to clinical settings (22,23) although some caution is warranted. As antibody testing is not yet routinely being carried out in the UK, the impact of the wording used in the current study amongst those undergoing testing cannot be assessed. All scenarios used the term 'positive' so we were unable to examine any possible impact of the use of this term. In addition, due to the rapidity with which the current study was conducted, piloting of the materials in general population samples was not carried out. It is therefore possible that some of the information provided may have been misinterpreted.
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 (24,25). 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.

Conclusion
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 lockdown 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 antibody-positive 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.    Anonymised data will be made available upon reasonable request.

Author contribution statement
The study was conceptualised by TMM, JW & GJR. AM completed data collection.

Immunity Passport
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.

Immunity Certificate
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.

Immunity Test
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.

Antibody Passport
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.

Antibody Certificate
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.

Antibody Test
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.