Comparative Analysis of Digital Contact-Tracing Technologies for Informing Public Health Policies ^†

Abstract

Contact tracing is the cornerstone of epidemic control of infectious diseases, especially in the era of COVID-19. This labor-intensive task calls for the use of digital technology to help identify individuals who have potentially been exposed to the infection to deliver necessary interventions and treatment. Mobile applications based on different technologies and system architectures have been developed and widely used in concert with public health policies and regulations. Three main types of digital contact-tracing technologies, namely Bluetooth low energy (BLE), location tracking, and check-in, were adopted in contact-tracing apps and implemented with a centralized or decentralized system architecture to protect privacy and facilitate spatiotemporal co-occurrence matching. Passive data collection methods, including BLE and location tracking, could be contrasted with check-in-based apps, which require users’ effort to record their whereabouts. Persuasive computing with a centralized system for collecting contact-tracing data from users might require additional legislation for authorization and privacy protection. Technology options should therefore be taken into account when designing enforceable public health policies on the use of contact-tracing apps. At the same time, public health policies also inform the information system’s design. This paper aims to delineate and contrast current technologies and system architectures used for developing contact-tracing apps and examine the intertwined relationship between the design and implementation of public health policies and the design of digital contact-tracing systems.

1. Introduction

To control the Coronavirus disease 2019 (COVID-19) pandemic, an unprecedented global effort has been in place since its first detection in Wuhan, China. As it is transmitted primarily through respiratory droplets and contact routes, minimizing person-to-person contact in social settings is an effective and important means of reducing transmissions [1]. Collectively referred to as “non-pharmaceutical interventions”, strategies involving non-biomedical products to help prevent the spread of infections, such as social distancing, hand washing, wearing facemasks, restricting mass gatherings, lockdown, and contact tracing, have been implemented in different places worldwide, especially when effective vaccines were not available [2]. When there are no lockdowns or restrictions on domestic travel, citizens may be free to visit multiple places for work, study, daily activities, and even entertainment. During the incubation period, which can be as long as 18 days [3], an infected individual may already have visited plenty of places; hence, direct contacts might have been exposed to the infection, so early identification, quarantine, and testing of these individuals would be necessary to contain further transmission [4]. Labor-intensive manual efforts by healthcare workers were required to help consolidate one’s travel and contact history. However, not everyone, both the patients and the potentially exposed individuals, can recall his or her whereabouts with spatiotemporal niceties. It could also be difficult for people to keep a diary of their daily itinerary. Tools for assisting contact tracing are therefore needed. To this end, health authorities in various jurisdictions have developed and promoted the use of contact-tracing apps based on different technical frameworks. Although app-based contact tracing provides an additional reduction in transmission, as these apps often collect data relating to users and several apps require registration with phone numbers, controversies about privacy and data security have never ended. In this paper, technologies used by contact-tracing apps were evaluated and juxtaposed with a comparative analysis framework.

2. Related Work

Previous works have been extensively discussed the privacy issues relating to contact-tracing apps, with a particular focus on applicable privacy laws, data collection statements, data sharing policies, and compulsory use. Data relevant to one’s whereabouts over an extended period would play an important role in epidemic control, but there is a tradeoff between the extent of privacy sacrificed by individuals and the effectiveness of contact tracing [5]. Legal tools to protect users’ privacy have always been in place in most jurisdictions. With the introduction of contact-tracing apps, the compatibility of the two required investigation. Existing data protection laws in the European Union and the United States were evaluated and contrasted to predict the challenges faced by implementing a contact-tracing app [6]. In Australia, in order to protect data collection and use from contact-tracing apps, new legislations were introduced to provide trust to app users so as to increase uptake and maximize its effectiveness in controlling the spread of infection [7]. Du et al. identified data protection laws in different jurisdictions and argued that global pandemic control could be jeopardized by the limited interoperability between jurisdictions as constrained by domestic data protection legislation [8]. Even if one jurisdiction could successfully contain the epidemic, with the extraneous introduction of pathogens should international travel resume, the risk of local transmission could re-emerge. This phenomenon is similar to dengue virus transmission in a non-endemic region surrounded by endemic places [9]. Although the perceived severity of infection might not be associated with the contact-tracing app uptake [10], it could be difficult to encourage a population with a low perceived risk to continue to comply with non-pharmaceutical interventions as they might consider it no longer necessary, aside from concerns about privacy [11]. Barriers to adoption in the general public population were identified in different countries. Difficult-to-read data collection policies could be a reason why the general public is reluctant to use these apps [12]. Digital illiteracy was often cited as a hindrance in using general and contact-tracing apps [11,13]. People with lower incomes might not have an updated mobile phone that is compatible with the latest version of the operating system to support contact-tracing apps [14]. A higher uptake rate was shown in Germany in the older population, as well as among those who had attained a higher level of education and had higher household incomes [15]. On the other hand, with an aspiration to “return to normal” and to stop local transmissions, people were willing to use contact-tracing apps [11]. Monetary incentives could be a booster for one-third of respondents who did not use a contact-tracing app [13]. Adopters’ satisfaction in using contact-tracing apps is the next paramount issue to consider for their continued use. However, most apps received negative reviews and ratings in app stores [16,17], often accompanied by complaints about battery consumption and the perceived effectiveness of the app [17]. These highlighted technical challenges and logistical considerations when implementing policies on the use of contact-tracing apps [18], both of which will be further discussed in this paper.

3. Digital Contact-Tracing Technologies

While several communication technologies could be used for digital contact tracing [19], three major ones were used in the actual implementation of contact-tracing apps [20], which are therefore included in this paper: Bluetooth low energy (BLE), location tracking, and check-in using quick response (QR) codes or barcodes (Figure 1).

3.1. BLE

Being the most prevalent technology adopted in the implementation of contact-tracing apps [21], devices with BLE broadcast their identifiers to nearby counterparts, who would log all received anonymized identifiers for the later detection of close contact with a patient. To protect privacy, the identifiers would be re-generated randomly from time to time, while a list of all identifiers ever generated in the device would be stored. Based on BLE, a privacy-preserving exposure notification system was thus jointly developed by Apple and Google and released to health authorities for developing their contact-tracing apps [22]. Other frameworks with different implementations, such as DP-3T and BlueTrace, were also developed [22].

With the strength of signals measured in the Received Signal Strength Indicator, the distance between devices could be estimated, although the precision could be low [23]. Bringing two pieces of information—contact history and signal strength—together, contact detection algorithms could be developed to define contact and recommend follow-up actions, which is the main purpose of contact tracing. However, such definitions varied across jurisdictions. In Taiwan, contact was defined as being within 2 m for at least 2 min [24]. In Australia and Ireland, at least 15 min of contact were needed, but the former requires close contact of less than 1.5 m, while the latter requires contact of under 2 m [25]. In Europe, BLE contact-tracing apps in Switzerland, Germany, and Italy adopted the definition of close contact with a distance shorter than 2 m for 10 (Germany) or 15 (Switzerland and Italy) min [26]. However, the correlation between signal strength and physical distance was far from accurate, even after calibrating the device models [26]. The algorithms adopted in Swiss and German contact-tracing apps did not result in any exposure notifications, although the devices were positioned in line with the definitions. It was also identified that there was a 50:50 chance to have triggered a correct exposure notification, similar to a coin flip event.

3.2. Location Tracking

A few contact-tracing apps opted for location-based technologies, such as a global positioning system (GPS), to identify people who had visited venues where a patient had also been by constantly collecting the user’s location data. However, a GPS’s error in measuring an exact position could be as wide as 7 to 13 m in urban environments [27]. Rather than detecting exact physical contact with another individual (or device), location-based apps aimed to identify occasions where the patients and other users coincided in terms of time and space. They also could not differentiate people on different floors, nor could they tell if the people were separated by a wall partition [28]. The draining of the mobile phone battery was also another issue identified in the pilot study. This can be alleviated by reducing the scanning frequency, at the expense of lower accuracy [29].

3.3. Check-In

A more accurate location history could be logged and stored on the device when an app user actively checks in and out by scanning a QR code or barcode posted at the entrance of a building or part of a building such as restaurants. In effect, this serves a similar purpose to recording one’s whereabouts in a diary. Yet, the app facilitates the logging process and allows for comparison with patients’ location histories. Similar to location-based apps, so long as an individual has been to the same place at the same or similar time, an exposure notification would be issued. This is the primary technology adopted in the contact-tracing app in Hong Kong, where the detection rules for venues and taxis were different: direct contact was defined as being in the same venue for at least 1 s, whereas indirect contact in a taxi was defined as taking the same taxi within 24 h after the patient left [30]. The implementation in the United Kingdom (UK) adopted a wider definition—a notice would be sent if the user were present at the same venue on the same date as the patient did [31].

3.4. Comparison of Digital Contact-Tracing Technologies

BLE and location-based apps are passive, which means that they send and collect data without users’ extra effort, whereas check-in-based app users should actively scan the QR codes or barcodes whenever they arrive at a venue. While the former are hassle-free solutions for forgetful users, the latter requires additional reminders or policies to encourage users to scan the codes when entering and to complete the log when leaving. In an ideal scenario where all app users check in and out appropriately with a high uptake rate, check-in-based apps offer a greater granularity of contact history compared with the low accuracy of location or distance estimation with other technologies. However, BLE is advantageous in identifying close contacts, the signal of which is stronger regardless of the definition of closeness. The most location-based and check-in-based apps can do is identify people who have been to the same place or part of a building for the latter one. It does not imply that the app users who have visited the same location have a physical encounter with each other. Therefore, data collected using different technologies also carry distinct implications for the policy evaluation process. It is almost certain that contacts detected with BLE could be defined as close contacts due to its technological constraints, whereas check-in apps rely on spatiotemporal co-occurrence, so linkages between individuals would require the design of an appropriate algorithm [32].

4. System Architectures of Contact-Tracing Systems

Logging data per se is insufficient for identifying potentially infected individuals. An appropriate system architecture should be designed for comparing contact histories between patients and other app users. Intuitively, the more data a third party possesses, the more concerns about an individual’s privacy will arise. In view of potential trust and legal issues, different types of system architectures were adopted [33]. A centralized approach was developed in which the centralized server collects data uploaded constantly by app users and performs matching once a positive user is identified for notifying potential contacts. This often requires the registration of personal information, including, but not limited to, mobile phone numbers, so that the health authority or the delegated agency can contact exposed individuals for necessary follow-up actions. If no data, no matter whether it is personal or not, were transferred to a third party or a centralized server, privacy could be protected while necessary notifications could still be received by the user. This decentralized approach could be achieved by constantly cross-checking with locally stored data after downloading patients’ contact histories, which were stored on a centralized server, without uploading their data.

The system architecture is fundamental to contact-tracing app design and implementation. When choosing a system architecture, relevant legislation on the power of data collection and data privacy and protection in the jurisdiction should be considered. A decentralized approach does not require the collection of personal data, which could be more trusted and acceptable by potential users and further reinforced by additional protection by enacting new laws [7]. Constantly uploading data to a centralized server could drain device batteries as well. From the point of view of the service provider, a larger server storage capacity and processing power would be required, incurring extra financial resources for deployment. The data collected in the centralized approach could, on the other hand, be used to monitor outbreak clusters in real time and in retrospect [34].

5. Contact-Tracing Apps in Practice

For a public health policy or regulation to be justly enforceable with the use of contact-tracing apps, additional system requirements in their design and implementation are always necessary. In Hong Kong, where a check-in-based contact-tracing app is in use in line with the dynamic COVID-zero policy implemented following China [35], a compulsory testing notice could be issued under local regulation if a patient had visited certain venues [36]. In April 2021, after the identification of two cases with unknown sources, such a notice was issued to anyone who had stayed in a one-million-square-feet four-story shopping mall for more than two hours in the preceding two weeks, although the two cases only visited two floors [37]. In contrast, in July, only selected floors and shops in a shopping mall that was visited by a patient were included in the compulsory testing notice [38]. It may seem unnecessary for people who did not have close contact with patients to be tested. This highlights the granularity of check-in-based contact-tracing technologies, which limit how a reasonably enforceable policy can be established. To affect the least irrelevant people, BLE could be used to help identify true potentially exposed individuals, even if a weaker signal threshold was set to determine close contacts. If the other two technologies were used, the best they could do is limit the location down to the unit of a building or part of a building, so more people would need to be tested. Thus, more resources would be required to test a larger amount of people.

While the use of the contact-tracing app was not mandatory, it serves the purpose of notifying people who should be compulsorily tested. There is a legal risk for non-users who were present in the specified place but did not undergo testing, even if they were unaware of the notice [39]. On the other hand, it could be labor-intensive or even impractical for law enforcement units to identify and track down all visitors by checking CCTVs. The uptake of these voluntary measures depended on one’s perception of their appropriateness as an antiepidemic policy [40], privacy concerns, and trust in the government’s management of the digital contact-tracing app [41]. Therefore, minimizing data collection with a decentralized approach and transparent data management policies would be some of the important elements to be taken into consideration when planning and designing a successful digital contact-tracing app.

6. Conclusions

In this paper, intertwining technological issues and public health policies are discussed, highlighting the decisions on technology and system architecture adopted that confine the scope of policy and regulation enforcement. When designing and developing contact-tracing apps, input from relevant public health policymakers is warranted to ensure the policies are enforceable and compatible with the system. With different system designs, the resources required for their implementation and subsequent policy execution could vary. Bearing in mind that the uptake of contact-tracing apps is paramount for maximizing their effectiveness in identifying patients to curb transmissions, trust building with app users and appropriate, proportionate complementary measures are vital.