Contact tracing tools for pandemics

Dutch version
French version

Monday, April 6, 2020.

The potential quality of data

The potential impact of an app is determined by the quality of data that the app collects. Data quality will depend, among other factors, on the traceability of users, on their data submission (if applicable) habits, and on the availability, state, and (correct) usage of the device. Mobile phone-based solutions for contact tracing rely on two fundamental yet questionable assumptions. First, the proximity of two mobile phones is deemed a proxy for virus transmission. Virus transmission, however, does not depend only on proximity; it depends mainly on the nature of the interaction (e.g. kissing has different implications than sitting back-to-back on a train) and also on the infectiousness of the infected person, the stage of the disease, etc.

Second, a mobile phone is deemed as a proxy for a person. The person-phone relationship or correlation, however, is shaped by values and practices. In some countries, people tend to have a one-on-one corresponding relationship with their mobile phones, keeping them on their body most of the time, making people generally traceable. Notable exceptions to this, however, are people who limit mobile phone use (e.g. in the presence of children) and minority groups (e.g. migrants, refugees, homeless, ethnic minorities, poor people or the elderly) whose mobile phone practices are not well-documented. In other countries, this person-phone correspondence has been shown to be limited by SIM sharing, phone sharing, and multiple SIM for example. Furthermore, cloud connectivity can be erratic and phone locations might only be tracked to the level of granularity of the phone towers. In such contexts, the accuracy of tracking people when relying on GPS is limited. The big data pulled from cellular towers during the Ebola containment efforts of 2014, for example, did not sufficiently reflect where infected people actually were3.

For mobile phone-based solutions that require data input from participants, numerous factors can affect data quality. Data input is most valuable for containment efforts when it is submitted in real time (i.e. input is provided as the observation or action is actually happening). In contrast, the submission of data through batching (i.e. input is provided later at the convenience of the participant) may lower the quality of data collected, as participants may forget to enter the data, or they may be subject to recall bias. If the app relies on participants’ willingness to report symptoms, diagnosis, or other factors such as their movements, then the data collected may not be accurate, because self-reporting might not be done properly or it may be highly subject to response bias (where people tend to submit socially acceptable answers when they, for example, anticipate positive reinforcement or a reward, or when they fear potential stigma).

Quality of data is also affected when some vulnerable populations may be less likely to have a smartphone or to keep it sufficiently charged all the time. Even the spreading of the epidemic itself might impact app usage. For example, illness can affect the quality and timing of how participants submit data. Also, as individuals come to know more people who are infected (especially loved ones), they may be more likely to use the app and to provide data.

The government’s approach and public trust

During containment efforts, the government’s approach and the public trust in its health officials and programs will be key to enhancing the levels of app acceptability and usage, and thus key to the app’s ability to trace contacts. To make a meaningful impact on public health, the app should be tracking a sufficient proportion of the population and at the same time effectively inviting contacts to get tested.

Governments can aim to increase public uptake by openly supporting, urging, and/or enforcing app usage. They may also offer mutual, value-added benefits, such as education about the new disease, health monitoring, or telehealth. Governments will need to consider the purpose of their measures following the identification of an index case and to adapt their strategies to certain levels of public awareness (about the virus and its transmission) and levels of literacy (especially among different sectors of the population). Education level, for example, may correlate with specific actions that can determine the success or failure of contact tracing apps.

Although in a mitigation situation the impact of an app for contact tracing on the course of the epidemic decreases (as people’s movements are already restricted), governments still may be able to use the app to estimate how well public health regulations are followed. However, this comes with risks, related to phone practices (e.g. people may have it turned off, leave it at home, or choose not to respond) and also to the legal and ethical dilemmas associated with obtaining such personal data. People who fear privacy breaches or aim to hide specific behavior may install the app on a second/alternative phone to circumvent regulations or find other ways of tampering with the device.

A government’s approach to contract tracing apps varies widely according to local culture, the political and governance system, and the nature of the application itself. In the case of China – who has a history of using surveillance software⁴ – app usage for the coronavirus outbreak was obligatory, and it enabled government surveillance, a practice that would be severely scrutinized in most Western countries. Individuals were each assigned a QR code based on their testing results, and locations also had a QR code to track individuals’ movements⁵. In critical times, transparent communication is key to avoid conspiracy theories or anti-app movements and to build trust in government, which was shaken at times in the COVID-19 case of China.

Potential shifts in human behavior

The use of apps (voluntary or obligatory) may have unintended social consequences. These should be considered in advance, as they could leave a lasting impact on society, even after the pandemic recedes. For example, a contact tracing app could create a shift in the perception of responsibility. Whose responsibility is it to change the course of the epidemic? Changing behavior to follow containment or mitigation measures requires accepting one’s own responsibility. When contact tracing is used, responsibility may be perceived to shift from the individual to the government. Another concern is that it might be difficult to roll back the use of such an app, or it might get used for other purposes.

While on the one hand, peer pressure to use an app could erode individual choice and agency, on the other hand, an app could enhance participants’ perceptions of their control over disease prevention. If a participant has been in close contact with an infected person and the app indicates this, then the participant might use such information to proactively warn their network.

People receiving a warning via the app could react in one of many ways. Will they come forward for testing? Will they aggressively demand to get tested? Will they panic? Will they keep silent out of fear of being quarantined or stigmatized? Will they discriminate against specific people they have been in contact with? What are the short- and long-term effects for vulnerable populations? Will tests be available when people need testing? In the case that a person’s profession involves many close contacts daily (e.g. cashier at a supermarket), how can they discriminate between less risky contact versus other riskier contact? And if this cannot be determined, what will be the effect of multiple warnings from the app? Behavioral aspects will also depend on the person’s literacy and understanding of the scope, features, and aims of the app.

In some cases, the usage of an app can create a false sense of security and safety, potentially leading to more infections. Those not receiving a warning to get tested might interpret this as “I am not infected” and consequently, they may undertake riskier behavior or mistakenly fail to interpret specific symptoms and thus delay or reverse their health-seeking behavior.

The ethical and privacy issues

In times of crisis and urgency, ethical and privacy issues may be overlooked, leading to unintended consequences that are later difficult or impossible to undo. How should data be handled, and are there safeguards in place to protect participants? Given the sensitive nature of contact tracing, data management principles and practices (e.g. where the data is stored and for how long, how secure is the system, who has access, is the data anonymized) must be not only well planned but also transparent. Most apps promise users that data will be deleted within a particular time frame, but their other privacy practices may vary. Attitudes towards privacy may also vary with culture, institutional structure, and governance. Individual freedom and privacy will need to be weighed against collective responsibility. Lessons can be learned from previously launched apps that have faced the same issues.

Contract tracing apps and their use of big data appear to expand the scope of traditional concerns about anonymity, as the data collected might be detailed enough for other parties - such as the government, hackers, or even other app users - to identify or track specific individuals. If the data will also be shared later for research purposes, what level of granularity of data will researchers receive? Would they be able to track infections (which may then facilitate triangulation of data and thus, the identification of individuals)?

Other unique issues also arise from the current situation. One can imagine that when following social distancing rules, an app user might unintentionally identify people (e.g. if they have been close to only one other person recently).

Some of these concerns can be solved technically (e.g. when Bluetooth is used only to record timing of close proximity between two mobile phones with the app). Other concerns have to be modeled using a systems approach to see how sensitive contact tracing may be to changes in the population of participants or in their behaviors. Such models have to be developed or run for each app separately, in order to take unique features of the app into account.

Conclusion

• These considerations need to be discussed broadly before big data approaches are endorsed or implemented. This weighing of arguments will be influenced by the severity of the epidemic. Governments and individuals might be more willing to take risks when more is at stake.
• What works in China or South Korea will not necessarily work in other countries.
• A thorough deliberative process will arguably hinder rapid launching of the numerous innovative, contact tracing apps currently being developed for the COVID-19 pandemic emergency. However, it could also minimize the potential risks for the current situation and enable quicker decision-making for future pandemics.  
• The use of app data for epidemiological research purposes is different from the use of app data to control the outbreak via sharing information with contacted persons.
• Our aim here is to highlight the key factors and their implications that should be considered before deciding to deploy contact tracing apps. One proposal is to roll out apps in phases, with the first phase being a pilot that precedes roll-out to an entire population. The pilot can be limited to a select group of users who have been engaged specifically to help in the contact tracing design process. This may alleviate some of the concerns outlined here.

References

1. Europe eyes smartphone location data to slow virus spread [Internet] 2020. www.pbs.org/newshour/health/europe-eyes-smartphone-location-data-to-slow-virus-spread. Accessed March 2020.
2. Researchers explore contact tracing app for Coronavirus pandemic [Internet] 2020. www.med-technews.com/news/researchers-explore-contact-tracing-app-to-contain-coronavir. Accessed March 2020.
3. Erikson SL. Cell phones≠ self and other problems with big data detection and containment during epidemics. Medical anthropology quarterly 2018; 32(3):315-39.
4. How China Uses High-Tech Surveillance to Subdue Minorities [Internet] 2019. https://www.nytimes.com/2019/05/22/world/asia/china-surveillance-xinjiang.html. Accessed March 2020.
5. Pan, X. Application of personal-oriented digital technology in preventing transmission of COVID-19, China. Ir J Med Sci (2020). https://doi.org/10.1007/s11845-020-02215-5 (see also https://www.newscientist.com/article/mg24532703-600-china-is-using-mass-surveillance-tech-to-fight-new-coronavirus-spread/)
6. Mobile Consumer Survey [Internet] 2019. https://www2.deloitte.com/be/en/pages/technology-media-and-telecommunications/topics/mobile-consumer-survey-2019/introduction.html. Accessed March 2020.
7. Overview of information available to support the development of medical countermeasures and interventions against COVID-19 [Internet] 2020. https://rega.kuleuven.be/if/pdf_corona.