Digital tooling must assist human contact tracing not replace it
All over the world a frantic quest for a coronavirus app has set off. In the past period, governments concentrated on decreasing the spread of the virus by putting societies in lockdown and scaling up hospital capacity. The drastic lockdowns have indeed brought down the magical number R0, the virus reproduction number: the average number of cases generated by one infected person. Very cautiously, governments are now relaxing the lockdown regulations little by little.
Governments all over the world believe that a coronavirus contact tracing app is a necessary tool for the ‘normal functioning’ of societies in the coming period, so long there is no vaccine to prevent people from getting ill. Look at this impressive list of initiatives in diverse countries. What are the reasons we need this app, according to our governments?
To slow down and stop the spreading of viruses, public health workers perform ‘contact tracing’. If a person gets a contagious disease, ‘contact tracers’ start an investigation to find the source of the contagion and the persons the patient could have infected. This is a very laborious process where it is quite normal that a small team spends several hours or even days tracing the contacts of one individual infected person. In the coronavirus pandemic, this means one needs an enormous capacity to carry this out. Some countries like Germany indeed have scaled up their contact tracing capacity, other countries like the Netherlands have not (yet) done that. The coronavirus contact tracing app is seen as a necessary instrument to assist the human contact tracing.
Convergence to a decentralized Bluetooth solution
The mobile phone is an obvious candidate for digital contact tracing. It has several technological options to support this, which can be used separately or in conjunction with each other. The most promising technologies are GPS, Bluetooth or QR-code scanning. It appears that the winner of this technology context will be the Bluetooth option.
Currently, Apple and Google are in a joint effort to come up with platform independent Bluetooth support. This Bluetooth framework is proposed to be ready in a couple of weeks. Respect for privacy is a very important design criterium: Bluetooth digital contacts stay on the phone, and only after being positively diagnosed, a Bluetooth id is copied to a central register where other phones can identify whether a digital contact has occurred.
Effectiveness
So far, so good? There is one caveat: Bluetooth has not been designed for proximity tracking and is not very good at it. Apple and Google are also stating this explicitly: Bluetooth proximity checking ‘only provides a rough estimate of distance’. To put it bluntly, we do not have a clue if this digital solution will be very effective.
The key question is what the correspondence is between a digital contact and a viral infection. There are numerous aspects in real life that impact Bluetooth proximity tracking, so it is very questionable if it is a dependable predictive factor for someone being infected. And, a viral contamination is also possible via contact surfaces where humans do not have to be near each other at all.
There are other problems. Bluetooth technology and the new Bluetooth framework need the newer smartphones, they just do not work on older technologies. And of course, a considerable amount of people in western societies do not have or use smartphones. Read here for an extensive and critical report on societal and technological barriers for effective use of a contact-tracing app.
But on the other hand, if some of the viral spread is reduced, why not.
Solutionism
Fundamental criticism on such technological solutions comes from Yevgeny Morozov, who calls this kind of technological innovations ‘solutionism’: instead of solving the root of problems, our governments and citizens resort to quick digital fixes. Morozov sketched out this pattern in his book ‘To save everything, click here’ and this coronavirus app is a typical example of this happening. He argues that the coronavirus pandemic raises a lot of fundamental questions that we should engage with, and develop structural solutions, and not (only) apply sticking plasters. But:
“when our lives are at stake, abstract promises of political emancipation are less reassuring than the promise of an app that tells you when it’s safe to leave your house.”
With all the energy centered on the coronavirus app, we tend to forget that it originated with a practical problem, namely that human contact tracing is too laborious and too slow a process. Let’s look at this contact tracing process in some more detail.
Manual contact tracing
When someone gets ill from a contagious, dangerous illness, a public health organization gets in contact with the infected person. A healthcare professional interviews the infected person (and possibly his or her relatives) to find out two things: firstly, what is the source of the contagion and secondly, which persons could be contaminated by this person? The contact tracers focus on a time slice relevant for the investigation. An obstacle in this investigation is the fallible memory of humans, and this amongst others makes it an intensive process. All kinds of memory assisting tools like agendas are used to come up with all the contacts the person had in that period.
After finding the group(s) of people that could be infected, these people must be contacted. When these persons are known it is not very difficult, but it takes a lot more time to track down unknown people: a real ‘investigation’ is necessary for getting in contact with these people and inform them that they are possibly infected.
Apart from this primary goal of contact tracing, an important secondary goal is to assist and help the people involved. Of course, people are very worried about their health and their families. So, the contact tracer has an important role to mentally assist and educate people concerned.
Time is key in this contact tracing. All the time that an infected person has not been traced, he or she may infect other people. So, it is paramount that, as quickly as possible, all the people that run risk are tracked down. A smartphone doing the warning in a split second, how ideal is that?
The digital contact tracing variant
The digital Bluetooth contact tracing goes roughly speaking like this. People have installed a coronavirus contact tracing app on their phone. When two people are nearby each other — according to some pre-installed parameters — a digital contact is registered on the phone. By the way, people’s privacy can be protected by registering only the minimal data needed and even by permanently changing Bluetooth ids.
When a person is tested positively, a public health agency changes the status of the app signaling that the person is infected. Then, the Bluetooth ids of that phone are copied to a high secure central register. Repeatedly, contact tracings apps are checking in this central register whether it contains a Bluetooth id which is also stored in its local register. If that is the case, that person is signaled as possibly infected by the virus.
The instruction to the person could be to quarantine immediately together with its household, or being extra careful looking out for symptoms and staying away from other people. In addition, when enough tests are available, the person can be instructed to have a corona virus test as quickly as possible.
We need a human in the loop
This solution is indeed a telling example of the digital solutionism approach. We have a problem with the manual laborious human source and contact tracing process, so we adopt an app: a quick technogical fix. On the other hand, this solution has one thing in favor of the human approach. It is much faster. In an instant, a lot of people can be reached and signaled that they run a risk of infection. Moreover, one can argue that the privacy of people is better safeguarded than in the human process, where human health professionals have access to heaps of sensitive and private information.
What are the down sides? To start with, the automatic process is an inhuman one. Via your smartphone you get a clinical signal that you could be infected with the corona virus. No clue when that could have occurred, nor what the probability of an infection really was. You and your family need to take immediate action. Contrast this with the human contact-tracing process where based on particular circumstances of the contact, the human health professional is able to give different instructions. A bystander in a grocery can be advised to look out for symptoms, the person with whom you have talked with in the park, can be instructed to quarantine immediately.
Personally, it gives me the shudders to have an app that can do that automatically, and I would most likely decide not to install such an app on my phone. Not because of the privacy nor the security aspects, but because it is a process where medical health professionals are not involved or only as a secondary agent.
For the contact tracing process as a whole, this scenario seems to me undesirable as well. Human contact tracers even have to do the same contact tracing, namely to find out whether contacts have occurred that are missed by the contact-tracing app. As a consequence, the work load for the contact tracers has not been brought down a lot.
A key question is how big the error rate is: how many people have been infected without a registered digital contact (you can call these ‘false negatives’), and how many there will be with a registered digital contact but without being infected (‘false positives’).
Having separate digital and human contact tracing process, it will be impossible to tie all ends togeter. And either way, the exact way a virus has been transmitted cannot be detected anymore.
Jason Bay from the Singapore Trace Together initiative makes this point strongly.
If you ask me whether any Bluetooth contact tracing system deployed or under development, anywhere in the world, is ready to replace manual contact tracing, I will say without qualification that the answer is, No. Not now and, even with the benefit of AI/ML and — God forbid — blockchain (throw whatever buzzword you want), not for the foreseeable future.
For another plea for this argument read this opinion where the authors write:
Ultimately, the best technological interventions to fight COVID-19 will be the ones designed in collaboration with contact tracers to enable them to do their best work and integrated into the planning processes of these programs.
Digital contact tracing assisting to human contact tracing
It is very important that we are not going to replace or bypass the laborious human contact tracing by a digital app, but instead develop tools that assist and speed up this process. The Singapore Trace Together initiative did just that. Every digital contact is uploaded to the public health contact tracing agency where contact tracers may analyze and filter the contacts and personally inform people that are possibly infected. Very recently Austrailia has launched their COVIDsafe app that is based on the Trace Together software.
The adoption rate of the Singapore Trace Together initiative reached twenty percent within one month. Although that is a considerable adoption, its adoption should be higher to be more effective. Already one million Australians have downloaded the COVIDsafe app, but we have to wait and see what the total adoption will be.
Perhaps, one of the factors that is preventing larger adoption of the Trace Together app is the privacy issue. To make it possible that human contact tracers get in touch with people, telephone numbers are registered in the system. In most western societies citizens will not accept this, and therefore adoption rates probably will be even lower than the twenty percent of Singapore.
For matters of privacy, a solution has to adhere to the European privacy regulations, the GDPR — General Data Protection Regulation. In essence, this regulation prescribes that using and registering personal data, it must be ‘fit for purpose’ and it must be done in a secure way. Depending on the sensitivity of this information the regulations are more or less strict.
The completely decentral anonymous contact tracing technology has the best guarantees for securing the privacy of the person. An intermediate role for human contact tracers seems impossible in this scenario so it appears that privacy concerns are in conflict with a more prominent role for a medical authority.
Nonetheless, I am personally much more at ease with a solution where a public health authority has a central role in the contact tracing, than with an anonymous solution completely guaranteeing my privacy. Of course, I also want that the app minimizes personal data, is secure and has been set up in an open, transparent manner.
Other solutions
Already, it seems that the whole contact tracing app has become a tunnel vision that prevents endeavors for supporting human contact tracing in other ways. For example, ‘just’ scaling up the capacity could be a worthwhile endeavor as well. The interview process can be assisted with automatic interview questionnaires. Other digital solutions are possible than digital contact tracing e.g. concentrating on geographic movement of the smartphone holder. One could focus on anonymous infection maps to analyze the spreading of the infection and identify areas where there seems to be an increase of corona victims. For an extensive overview of testing and contact-tracing, look at this article in Thomas Pueyo's serie ‘Learning how to dance’.
However, to come back to our digital contact tracing app. Speaking personally, if I am convinced that a contact tracing app assists our medical health professionals, that it has a positive effect on our fight against the virus and that the solution is GDPR proof. Yes, I would use such an app.
