With the first wave of the COVID-19 pandemic tapering off at least in parts of the world, governments in many countries are working towards providing a framework of infection surveillance and infection control measures that hopefully allow economies to restart and lift most of the lockdown measures that were implemented in March and April. The minimum goal would be to move from an undifferentiated national lockdown to a more flexible, more regional, more data-based approach that is chiefly about cluster containment and does less harm to the economy and society.
Whether or not – and if so, how – digital solutions can contribute to this pandemic transition phase is among the most interesting questions that the COVID-19 pandemic poses from a digital health point of view, and is explored in further detail in the HIMSS Insights eBook. It is also a difficult one, first because there are very different ‘ground control’ requirements in infectious disease prevention and surveillance that could be covered or partially covered by digital solutions, and second because there is a certain tension between what might be desirable in terms of public health and what is necessary to protect an individual’s privacy.
Infectious disease surveillance by social media monitoring?
So let us take a closer look. Most importantly, there are very different types of COVID-19 related public health apps. At least some of the recent COVID-19-related privacy discussions come down to misunderstandings of the purpose of certain types of applications. And some politicians would have been well advised not to associate different public health goals when pushing for COVID-19 apps.
The first type of digital solution that has been mooted to help contain COVID-19 is in fact much older than the current pandemic. These are infectious disease surveillance applications. They don’t aim to manage individual human beings with or without infections. Rather they are early warning tools, geared at detecting increases in disease activity on a population level.
There are different approaches towards this goal. A famous one that looked very promising in the beginning but proved error-prone later on was Google Flu Trends (GFT), launched in 2008 and abandoned in 2015. GFT tried to predict influenza prevalence based on search engine queries and social media content. It turned out that, used as a stand-alone tool, GFT was considerably less accurate than conventional influenza surveillance. Nevertheless, social media surveillance is among the approaches that many “conventional” surveillance systems have been adding to their tool kit in recent years. Therefore, while GFT is dead, its vision is alive and continues to be researched.
Activity tracking as an additional tool
The COVID-19 pandemic somewhat coincided with a different, more sophisticated concept of digital disease surveillance, one that doesn’t use social media data but wearable data, specifically fitness tracker data. Not anticipating COVID-19, scientists such as Jennifer Radin, Eric Topol, and Steven Steinhubl from Scripps Research in La Jolla, California, published a research study in January 2020 that analysed de-identified sensor data from 200,000 Fitbit customers.
Participants wore the device for at least 60 days and used it regularly over a two year period between March 2016 and March 2018. Data analysed included resting heart rate and sleep measures. The scientist could show that in US states with a high number of participants, analysing abnormalities in the Fitbit dataset significantly improved the prediction of influenza-like illnesses – by 6% to 33% – compared to a baseline model that used data from the Centre for Disease Control and Prevention (CDC) only.
The spatial resolution problem
In March and April 2020, researchers from the German CDC equivalent Robert Koch-Institut (RKI) together with the Berlin-based digital health startup, Thryve, decided to apply the Scripps-approach to the COVID-19 pandemic. They published an app called ‘Corona-Datenspende’ (corona data donation app) that collects data from different types of smartwatches and fitness bracelets in order to look for early warning signs that could make a second or third COVID-19 wave visible before infection rates go up and before laboratory statistics start looking suspicious.
According to RKI, within six weeks, more than half a million wearable users have downloaded the app. If the user gives their consent to the data donations, the company algorithms can access step counts, resting heart rate, and movement-related data and analyse whether there are regional patterns suggesting clusters of infected people that would feature an increase in resting heart rate and/or a decrease in movement.
This type of analytics is obviously not specific to COVID-19. However, it could help pinpoint regions that require more stringent COVID-19-testing. While convincing in theory, there is a spatial resolution problem in reality. With half a million downloads, it means that, at the moment, only one in 160 citizens in Germany is using the app: “Provided we had only 1,000 new infections per day, the number of data donors would be too small to make regional clusters visible,” says RKI researcher Dirk Brockmann. There are limits for early recognition as long as the number of data donors or wearable users is low.
Digital contact tracing: The Holy Grail?
Another type of digital health application that many think could be among the keys to unlock the doors back to a somewhat normal life in times of a pandemic are contact tracing applications. Nearly all countries that were heavily hit by SarsCoV2 have started contact tracing app development programmes or launched such apps already. These apps inform people about whether and when they might have had contact with someone who later tested positive for SarsCoV2. The goal is to warn people who might have been infected as early as possible.
Some of these contacts will be identified by conventional, manual contact tracing anyway. For them, the app should generate a warning that will typically arrive two to three days prior to the phone call of the health authority responsible for the manual contact tracing. Others, who may not have been identified through manual contact tracing at all, would receive a warning. “With conventional contact tracing, we lose many people because we cannot inform them, for example fellow passengers on a bus or on a train,” say Gottfried Ludewig, head of digitisation and innovation at the German Ministry of Health. Ludewig is in charge of the German SarsCoV2 tracing app ‘Corona-Warn App’ that was made available for app store download in the middle of June.
There are problems with digital contact tracing apps though. A number of evaluations have shown that many of them don’t really work from a public health point of view. Among the difficult issues is adoption. Coronavirus tracing apps tend to be voluntary. In Norway, one of the forerunners, only one fifth of the population was actively using the tracing app ‘Smittestopp’ at the end of May. There were also severe privacy concerns that led to the project being cancelled in mid June. In Singapore, among the poster child countries for digital healthcare and no stranger to citizen surveillance, the local tracing app called ‘TraceTogether’ was adopted by only 25% of the population in May, two months after its launch. This low uptake, again, is considered by local experts to be related to privacy concerns.
Iceland, too, is among the countries that started digital contact tracing early. The adoption rate there is 40%, more than anywhere else, but still not that impressive in a country with a population of only 340,000. Iceland is also among the countries that openly admit the app had not been as helpful as anticipated. In an interview with the Technology Review, Gestur Pálmason from the Icelandic Police Service said that the ‘Rakning C-19’ app was useful in a few cases but far away from being a game-changer.
Is privacy a success factor or a risk factor?
Around the globe, there are different strategies to improve digital contact tracing. One school of thought says the applications should be mandatory. This is the road that was recently taken in India, a country with high numbers of infections in some parts of the country and a fairly draconian lockdown that left many people stranded away from home. In India, people who want to travel to or enter their workplaces have to download the government’s ’Aarogya Setu’ tracing app.
India’s tracing app, like the one in Iceland, uses GPS technology. The app had come under scrutiny in early May, when the French IT expert Robert Baptiste, a ‘friendly hacker’ who calls himself Elliot Alderson on Twitter, claimed that he was able to identify infected people based on the location data that the ‘Aarogya Setu’ app is collecting and which are stored on a central server. This led to discussions in India about whether the app should be made open source for better transparency, and whether mandatory use should be abandoned or at least restricted to certain areas.
Next to India, Singapore is another country that is toying with the idea of making digital contact tracing mandatory. In the earlier phase of the pandemic, leaders of Singapore’s Coronavirus Taskforce have repeatedly said that they wanted everybody to use the tracing app once the technology was ready. But this didn’t quite work. By early June, the new goal was to replace the ‘TraceTogether’ app by a Bluetooth-based wearable that would be mandatory to wear and serves the same purpose. The advantage, according to the Singapore government, is that a separate wearable can also be used by people without a smartphone. Furthermore, a separate wearable might be more socially acceptable, since tracking it might be considered less intrusive than smartphone surveillance.
The more trust, the more voluntary adoption
In fact, many countries outside of India and Singapore are trying to boost adoption rates of contact tracing apps or contact tracing wearables not by making them mandatory, but by applying unusually high transparency and privacy standards, and by using technology that makes the applications less prone to fraudulent use. While, for example, the contact tracing apps in South Korea and Iceland use the GPS signal, and thus could, in principle, be misused for location tracking, most of the newer apps that are being developed and implemented now are based on Bluetooth signals.
Bluetooth is not particularly suitable for location tracking. It also provides a more accurate depiction of contacts, since truly close contact can be identified more accurately, and time of exposure to an infected contact can also be quantified better. Advanced pseudonymisation technology and decentralising large parts of the applications are other technological approaches that aim to increase trust. One example is the UK, that had originally planned to inroduce a centralised contact tracing app infrastructure. In the midlle of June, the government suddenly announced that it would instead back the contact tracing model favoured by Apple and Google and use their interface for the national contact tracing app that, so far, is not yet available.
Germany is also using the Apple and Google interfaces for its contact tracing app. There was a representative survey made by the Nürnberg-Institut für Marktentscheidungen among 1,500 people in Germany shortly before the German healthcare system started rolling-out its Bluetooth-based, largely decentralised and fully open-source contact tracing app, developed by SAP and Deutsche Telekom, in the middle of June. The survey asked whether people would be willing to install contact tracing apps of various types.
A hypothetical app that only stores data on the mobile device and nowhere else, that is completely voluntary and evaluated continuously by independent researchers, received the highest approval rate, an impressive 69%. Contact tracing cannot, by definition, be fully decentralised. But the results indicate that transparency and privacy will very likely increase the willingness of citizens to install a contact tracing app on their mobile devices – and thus increase the likelihood, that such an application will make a real public health difference.
This is also true for other digital contact tracing that do not rely on peer-to-peer recognition. Singapore again: The city has rolled-out its ‘SafeEntry’ application in late April. This is a QR-code based check-in system which replaces hand-written lists of people visiting certain location. Citizens entering hotels, supermarkets, hospitals and offices are required register upon arrival and departure.
Beyond Excel: How to manage infected patients and contact persons
Finally, successful coronavirus ground control is also about managing infected people and their contact persons. This has nothing to do with digital contact tracing per se. It is rather more about process management tools for local government health agencies that, during certain phases of the COVID-19 pandemic at least, have to deal as efficiently as possible with hundreds of cases and contact persons in parallel. Each of these patients or contact persons has to be accompanied for 14 days typically. In many countries, the method of choice for healthcare workers or COVID-19 tracing staff is to use Excel sheets. This works reasonably well in, say, meningitis outbreaks with a limited number of infected children and a limited number of contact persons. In the case of SarsCoV2, it is different: Excel-based administration of infected persons and contacts can quickly become a process management nightmare.
Interestingly, dedicated outbreak software has been in existence for years already, but it is typically not used in industrialised countries. A prime example is a software called ‘SORMAS’ that was developed in Europe and installed in hundreds of local health centres in the Ebola regions in Africa, covering a population of more than 140 million citizens in several countries. ‘SORMAS’ provides digital pathways for managing patients and contact persons more efficiently than can be done with spreadsheet software. Recent add-on apps also allow communication with infected persons and contacts digitally, replacing the need to do daily phone calls and ask for symptom development.
The last category of digital applications for infectious disease control is quarantine surveillance tools. These are in use in only a few countries now, since in many parts of the world, they are perceived as some kind of electronic ankle bracelets and as such somewhat too intrusive in a public health context.
The prime example of this app is the South Korean ‘Self-Quarantining Safety Protection’ app that visitors to South Korea have to install for 14 days when entering the country. Kerala, a province in Southern India that has contained COVID-19 with relative success, used a similar tool, the app ‘Covid Safety’ to enforce quarantining in one heavily affected region, Kasargod, and later in another city, Kochi. This app informs the local police when a person quarantined is more than 50 metres away from home. Interestingly, though, the app is voluntary, but seemed to work nevertheless. Speaking to the Times of India, Kochi City Police Commissioner Vijay Sakhare said: “People in quarantine should realize that a mistake could put the lives of our elderly and others in society at risk.”