Testing immunity certificates: do the new antibody tests open the door to the creation of a ‘public health identity’?
The first in our new series of virtual events addressing health identity systems developing in response to COVID-19.
Four speakers from the fields of immunology, information systems and behavioural science discuss immunity certification: does the roll out of antibody testing open the door to the creation of a public health identity?
In this first in our new series of virtual events addressing health identity systems developing in response to COVID-19, ‘Towards a public health identity?‘, we explore whether the introduction of immunity certification could lead to a ‘public health identity’.
This post summarises the key points of discussion and debate from the webinar, which you can watch in full below:
Associate Director (Society, justice & public services)
Danny AltmanProfessor of Immunology, Imperial College London
Virginia BerridgeProfessor of History and Health Policy, London School of Hygiene and Tropical Medicine
Robert WestProfessor of Health Psychology, University College London
Edgar WhitleyAssociate Professor of Information Systems, London School of Economics
Systems for verifiably sharing private health data relevant to public health concerns, enabled by digital technology, are developing around the world. In China, there are now health code apps using a range of data sources, from location to quarantine status. Global companies are using private apps to categorise their staff: PwC are working with Vottun in Spain on a digital health passport, and EY are trialling Bizagi’s CoronaPass app.
In the UK, there are nascent plans for an ‘immunity certification’ system, spurred on by the successful development of an accurate COVID-19 antibody test. While there is little detail around these plans, Health Secretary Matt Hancock has confirmed the UK Government is actively considering how immunity certificates could be used to release some individuals from lockdown. This would create a form of ‘health streaming’ system, with citizens’ access to parts of society and possible employment, spaces, travel or interaction contingent on bringing personalised private health data into the public sphere.
The panel will consider:
- What we know about the current state of immunity, and whether it makes sense to link antibody testing to any kind of immunity certification?
- What social issues arise from immunity certification, and what lessons can be learnt from other public health crises?
- What issues we should consider if a form of ‘immunity certification’ is digitally delivered or appended onto an app ?
- Will any such system be a temporary response to the immediate crisis and the need to exit lockdown, or – once introduced – are public health identities here to stay?
What do we know about the science of immunity?
Depending where you are, including which country you are in, you will get access to many different antibody tests, from public health providers or privately available. Their reliability and validity with vary widely.
We don’t know what the requirements for protection are that could mitigate a second wave because the immune system is a thing that has thousands of different cell types in it, many types of antibody, many functions of antibody, and this test isn’t delving down into any of that detail. It’s just trying to count how many molecules of antibody that are there not what they are doing or whether they are working on the virus. It is a proxy which we hope might vaguely correlate with your chances of being protected next time around.
We might be able to get a certificate out of those results but it’s unclear how durable that certificate should be. Does it mean that you’re safe for six months, 12 months or 24 months? One of the things we definitely feel about coronavirus immunity is that it is rather fragile and not very durable. But it is too early in the day to be sure about this.
What if we got tested every month? Would that help with issues of reliability?
It would solve the issues related to people who can’t make any antibody but are immune through their T cells or whatever, bad tests, cross-reactivity with common colds and those sorts of things. However, it could help enormously on the question of the durability of immunity.
Some people would also not be able to pass any antibody test due to immune deficiencies or exposure to certain drugs. However, some people who are prevented from making antibodies due to genetic defects manage to recover perfectly well from this virus without any medical attention. It isn’t a correlative protection because you can have zero antibody on board by definition and still make a full recovery.
The take-home message is you can do an antibody test but the results are very, very variable. Antibodies do not equal protection – they may or may not be a good proxy for protection.
Is risk of infection the same as the risk of transmission?
There’s pretty good evidence that as soon as I’m making detectible antibodies, I’m almost certainly not shedding the virus anymore which could be some evidence for causal relationship between making antibody and neutralising virus. So being ill, making antibodies and shedding virus are different but related.
What criteria could and should we use to evaluate immunity certificates as a policy intervention?
I want to set the framework to help to understand all of the issues that one has to consider and this is a framework that we think applies to really any policy or intervention that you might be considering, and the purpose of it is twofold.
One is to provide a way of seeing all these evaluation criteria together because they interact with each other, and the other is to make sure that they all get considered, because one of the things that often happens when you’re considering policy is that we think of one thing and we think of another thing and then someone else thinks of another thing but we find it very hard to put it all together.
The criteria go under the acronym APEASE:
The extent to which this is acceptable to key stakeholders. Employers will have different interests from employees. People who are vulnerable may have different interests from others. There’s going to be a general issue of trust and the need for the thing to be perceived to be fair.
Can you roll out the kind of thing that you think you are rolling out? There is the issue of error correction. Mistakes will occur. As we’ve already heard. In fact, they may be quite considerable. What are we going to do about that? How will we build that in? Can it be rolled out as intended?
Does it achieve the desired objective? What is unclear here is what are the objectives? We know in vague terms that we want the economy to get moving again and people to be able to operate and people to go back to their lives, but these are competing objectives in some situations and one of the things that is very important here is to set benchmarks.
It is important that people understand the difference between sensitivity and specificity and predicted value of a positive and negative test result. I’ve tried to teach medical students this for about a decade at St George’s and I’m not sure I ever succeeded, so it may be even tougher to communicate the public.
Affordability is different from cost effectiveness. It doesn’t matter how cost effective something is in principle, if you can’t afford it within your budget, you can’t do it. It’s not just about whether it’s affordable to the public coffers, but also employers and people. And it’s not just about the affordability of the test, but also the affordability of the consequences of the tests. The question is: who pays?
- Spill-over effects
What are the unintended consequences? These can sometimes be good – they don’t have to be bad, but very often they are unwanted. The most obvious one here is the potential for social division
Is this going to increase or decrease inequalities? COVID-19 is already having an enormous affect exacerbating inequalities in society. There is a risk that something like this might do more.
What can we learn from historical examples of health-based categorisation?
In general, the idea of singling out groups of people in the population through some form of certification system for restrictive purposes, or for preferential treatment, has been fraught with difficulties. It’s tended to exacerbate privilege and divisions within society and here are three historical examples of that:
The American historian, Kathryn Olivarius, has written about the impact of Yellow Fever in New Orleans in the 19th Century. Only those who had had the disease, and could prove that they’d had it, were able to find a place to live, a job, a bank loan or a wife.
Newly arrived immigrants in the city struggled to get infected and accepted while the wealthy left the city during the outbreaks of Yellow Fever. The system was set up to emphasise privilege, but there was also confusion about how long immunity lasted and whether the tests were accurate.
Syphilis Marriage Certificates
In 1930s America, in many states you couldn’t get married without a premarital certificate to say that you were free of syphilis. You couldn’t get a marriage licence until you were free, according to the Wassermann test. The Wassermann test had a high number of false positives and so many people had treatment they didn’t need and experienced stigma based on inaccurate results.
Often it was only the bridegroom who was examined, and not the bride. People tended to move from states which required certification to those which didn’t in order to get married and it was possible to get around the restrictions if you had enough money. It was also part of the eugenics movement of the time.
Testing for HIV and what it meant was hugely controversial in the early years, when there were no effective treatments. Insurance companies took the issue of testing as a trigger for wider inquiries as well as to potentially refuse insurance.
In the early days, some people thought that being HIV positive was a good thing as they thought it meant that you were immune. But of course, very soon, understanding changed, and it became a very different thing to be HIV positive.
The historical conclusion about certifying people who either have, or don’t currently have, a particular disease is fraught with very difficult issues. In particular, it can lead to exclusion, stigma and the undermining of rights. Historically that has been for racial minorities, and for the poor, but can also be for other sections of society as well.
What issues might the digital implementation of immunity certification raise?
There are evolving and emerging frameworks that can at least help us to make some sense of the kinds of issues we are talking about when building digital identity systems.
It might be conceptually more useful to say we are starting to develop ‘attributes’ that we can link to a person, e.g. a person ‘has antibodies’, and that’s as far as we can go. Then we start thinking about how that attribute might be assured, how we might have confidence in that attribute, and what information that attribute contains, based on:
- Multiple tests being available, from different suppliers.
- More, or less good, administrations of those tests.
- Multiple scores for an individual in terms of how many antibodies they might have
- Different time periods since the test was undertaken
You then need to consider how you link that attribute of having antibodies to the person who was tested. It will be important to have the ability to link tests to the particular person and to be able to check that in a reliable way that can’t be faked, particularly if the incentive is to have the attribute of ‘has antibodies’ to re-enter employment, etc.
We also have to bear in mind that the ‘has antibodies’ attribute, is going to be relied upon by a variety of different parties. If you are looking to employ someone to work in a care home or checking whether someone can sit closer to you in public transport, there would be a whole series of different-relying parties who may also have different expectations and different requirements from the profiles from all of these different multiples that we are considering.
Therefore, we need to have a way of evaluating and determining those different kinds of profiles so that you might say, this set of combination of factors would not be useable for accessing this particular service, this particular employment. However, for lower risk environments, we might think that as long as you’ve got certain elements there, that’s absolutely fine.
If we start to build those kind of standardised profiles then you start having the opportunities for interoperability, so you are not limited to having to buy this test, done in this kind of a way, but rather the marketplace would allow different individuals to present their attribute of ‘has antibodies’ in different kinds of ways. There is also the opportunity for potentially cross-border interoperability as well.
Could immunity certification play a role in reducing inequality of the disease burden across occupations? What might the secondary impacts be? Could it exacerbate vulnerabilities?
I struggle to think how it might make things better in terms of inequalities, or to reduce them. My worry here is that something can become a serious proposition and it doesn’t matter how much evidence comes in that this really isn’t working, or isn’t going to work, it builds up a momentum. I think it’s slightly worrying that there are groups and countries that are seriously considering this at the moment when we don’t have something which would even do the job that you would hope that it would do, irrespective of everything else.
My hunch is that the best way to get a high level of antibodies is to be exposed an awful lot of times. So, I suspect that if you surveyed the population, the people building up the most antibody are the front-line workers who are exposed repeatedly. Although many of them still seem to be susceptible. So, there is a risk that I hive off a subset of society who are super-exposed, super at risk, but can get very well certificated because of that risk they face which would be a bizarre unforeseen consequence.
I can also see a well-intentioned, but clearly discriminatory consequence in terms of inequalities where, because we are concerned about those particular ethnic subcommunities that are more at risk if they do catch something, you check more often that they are OK and already have their antibodies before mixing in a particular location.
It ought to be a non-starter, but there is the commercial demand from companies and from digital ID providers who want to offer this kind of service. They are offering it because there is demand from relying parties. There are lots of companies that are desperately trying to find some way of restarting the economy.
Are there better way of balancing public health and liberties?
I think the problem here is that when you see something like this, that looks like a technical fix, it dominates people’s imaginations and there is an opportunity cost, because then people are not looking at other solutions. There are lots of different ways of tackling this issue – from a health point of view, from a public point of view and from an economic point of view.
For example, there are ways of building an economy that is much more resilient to this kind of infection by virtue of the way that it operates. An economy that relies less on mass gatherings, close contact etc.
Another one is getting effective and efficient track and trace, and that seems to me to be potentially a very viable option in the future. At the moment, it is not close to being there in the UK. But that looks very viable, with rapid response. Instead of focusing on waiting for a vaccine we need to be aggregating marginal incremental gains and moving ahead across a whole sweep of things.
With all the caveats I’ve already posed about reliable antibody testing, maybe we’re not ready for that and we’ve invested too much faith in that. We do have biomarkers of risk, and they are being old, being male, being obese or being a member of an ethnic minority. Think how unacceptable it would be to stratify your society on that basis.
Historically contact tracing was a key mode for controlling infectious disease for things like TB and STDs. I think that is a history which has tended to be a bit forgotten, located in local government with the medical officer of health, sending out contact tracers into the community. So that is another model that we could be looking at.
Syphilis testing in the States did actually result in a reduction in the levels of syphilis in communities. A punitive form of public health can bring about a “good” result. but it is the other issues that are involved with that which are also important.
Is there a risk of scope creep? Can we unwind a system we found acceptable in an emergency once the emergency is over?
At one level, the attribute ‘has antibodies’ could be no different to ‘is overweight’, or ‘has a predisposition to certain cancers’, or ‘has a propensity for other issues.’ So, there is nothing to stop the technology from releasing and sharing all of these attributes and that means that the framework in which you operate these kinds of systems would need to have strong regulatory controls. That might be a legal regime that says we’re only going to allow you to do this for COVID-19.
You can immediately see the kinds of pressures. We’re very concerned about people having heart attacks at work, so we want to at least be able to evaluate the likelihood that somebody driving a taxi is not susceptible to a heart attack. I suspect the key constraint on that would be the extent to which there’s genuine demand for all such theoretical potentialities. Realistically, if you need a taxi to get you somewhere quickly do you really care whether that driver has a slightly increased potential of having a heart attack, or do you assume that the general regime for driving tests, etc, covered that?
There may be something self-controlling about how many different things do you check, which things you account for? The potential in terms of the technology would not prevent it. It would need to be the wider ecosystem that would, could, or should, limit what other kinds of attributes you would want to do that.
Once you start doing those attributes, you could also add: ‘went to a public school’, ‘is a good skier’, all of these other kinds of proxies for lots of these assumptions that we’re making in society.
A lot of the criticism of the Government’s report about the disproportionate effects of COVID-19 on black and minority ethnic communities was also that it is not necessarily their ethnicity, but it could also be their income levels, housing arrangements, etc. We need to open up many of those conversations, and not just say, “Oh, because one measure suggests that, that that is the causal link to other kinds of concerns.”
We, as a population, are going to have to be much more savvy around how we understand and interpret information of this kind. It’s hard enough for epidemiologists so trying to get that into the public sphere and into insurance companies or various policy groups is going to be incredibly challenging.
We’re starting to move from seeing it as a sort of all-or-none thing to another part of the risk management that we do every day. In traffic, some days it’s more risky than others. If it is raining, then you behave differently from if it is dry.
What I could see happening is a system developing where we have place and time and condition and person-sensitive disease risk assessment, and that’s just built in to our overall risk management structures, and we think of it in terms of not as ‘safe’ or ‘unsafe’ but as a continuum of risk. I don’t see how we can avoid doing something like that.
Test, trace and trust: digital technologies and the COVID-19 response across the UK’s devolved nations
This is our third virtual event, exploring approaches of different nation states to using technology to control the spread of COVID-19.
Contact tracing: to centralise or not to centralise, that is the question
This is our second virtual event, exploring approaches of different nation states to using technology to control the spread of COVID-19.