01. Mai 2020, Nicholas Martin
Among the countries hit by Covid-19, Taiwan stands out for several reasons. Most importantly, Taiwan has managed to keep the number of infections and deaths very low, with 429 cases (as of 30 April) and 6 deaths in total, or 18 cases and 0.3 deaths per million inhabitants. This compares to 161,539 cases and 6467 deaths in Germany (1928 cases and 77 deaths per million).
This fortunate outcome was not obvious. Taiwan’s dense human and economic links and geographical proximity to China potentially set the stage for a rapid spread of Covid-19. Over 400,000 Taiwanese citizens work in China; the total number of Taiwanese in China likely numbers over 1 million, out of a national population of about 23.8 million.
Taiwan’s fortunate outcome is interesting also because the country implemented neither a lockdown nor mass testing.
This blog entry does not attempt a full account of how Taiwan managed to control Covid-19.[1] The aim is more limited: to explore the role of digital tools in Taiwan’s response, and if possible draw some first conclusions about the implications of the Taiwanese experience for current German and European debates.
For two reasons, Taiwan presents an interesting case study in this regard. Firstly, it is a highly digitised society. For instance, some 79% of the population uses a smartphone, and Taiwanese companies are key players in global ICT supply chains. Secondly, Taiwan is a vibrant democracy with robust civil liberties, rule of law, and an active and contentious civil society. Taiwan received a score of 93 (out of 100) in the annual rankings by Freedom House for 2019, comparable to Germany (94/100).
Taiwan’s use of digital tools to contain Covid-19 can be divided into four areas; linking databases to identify potential Covid-19 cases, contact tracing, quarantine surveillance, and monitoring of mobility patterns.
Upon learning of the new Covid-19 virus, Taiwan moved very rapidly to identify and register individuals entering the country from risk areas (initially Wuhan, later extended step-wise to the rest of China, other Asian countries, and Europe and elsewhere). Upon disembarkation, people arriving in Taiwan from risk areas had to provide officials with information about their recent travel destinations, personal identity and their telephone numbers and address in Taiwan, and were obliged to quarantine for 14 days (see below).
Concurrently, the Taiwanese authorities linked the – ordinarily separate – National Health Insurance and Immigration and Customs databases. As Dr Anita Pei-chun Chan, an infectious disease specialist at Taiwan’s Center for Disease Control, explained in a telephone interview for this blog, this had several goals. By linking the databases, frontline medical personnel could be provided with immediate alerts if they were visited by patients recently returned from risk areas. This was to ensure that patients would immediately receive appropriate testing and care, and to protect frontline health workers.
Linking the databases also allowed the national Centre for Disease Control to re-analyse recent patient histories, to identify as-yet unrecognised cases of Covid-19. This was important, according to Dr Chan, because while the list of international risk areas was steadily expanding, it necessarily lagged realities on the ground, as Covid-19 gradually spread around the globe. By the time a country or region was designated “high risk”, Covid-19 could already have been spreading there for some time. But prior to designation, sick individuals returning from there and visiting Taiwanese hospitals would not be flagged in the way described above, meaning they might not be recognised as potential Covid-19 cases.
For reasons of data protection, only the last 14 days of travel information from the Immigration databases was made available to the health system. As the Internet entrepreneur and digital activist T.H. Schee explained in a telephone interview, linking these databases would ordinarily have required a court order. For the duration of the Covid-19 crisis, however, Taiwan’s (pre-existing) Communicable Disease Control Act apparently provided the necessary legal basis.
Institutionally, this data integration and analysis seems also to have critically depended on the fact that Taiwan has a single-payer national health insurance system with comprehensive, unified database(s). All Taiwanese nationals and most foreign residents are covered by the same national insurance. Patient records reportedly include “complete health history, underlying health conditions, and recent progression of symptoms, treatments, and hospitalization related to respiratory syndromes.” Because care providers must submit insurance claims within 24 hours, the national insurance database provides near-real-time information on all medical visits and treatment.
Taiwan devoted very substantial resources to tracing all people who had come into contact with confirmed carriers of Covid-19. Digital tools played an important role in this; however, for the most part they functioned as support tools, not substitutes, for trained human contact tracers. Only when human tracing was infeasible were digital tools used exclusively, and in those cases their value-add seems to have been uncertain.
According to Dr Chan, when a new case of Covid-19 is discovered, frontline healthcare workers initially interview the patient. The goal is to retrace their movements over the past 14 days, to identify the likely source of infection and all close contacts. However, people can rarely recall their movements and contacts reliably for longer than one or two days. Occasionally, she noted, there had also been problems with patients not wanting to disclose their movements or contacts. (Under Article 31 of Taiwan’s Communicable Disease Control Act, patients with designated communicable diseases are legally obliged to give medical personnel a full report on their contact and travel history.)
The contact tracers therefore next request a map of the patient’s movements in the past 14 days from the police, based on mobile phone cell-tower data. This is compared to the movements reconstructed through the patient interview. If inconsistencies remain that cannot be resolved through further patient interviews, CCTV data is requested from the police (where available). (Taiwan reportedly has a relatively dense network of CCTV cameras.)[2] Contacts thus identified are asked to self-quarantine or moved to hospitals.
In several cases the Taiwanese authorities resorted to mass contact tracing and alerting by means of cell-tower data (cell tower triangulation). To send and receive calls and messages, mobile phones automatically connect to the nearest cell tower, allowing their location at any point in time to be pinpointed (with an accuracy in the size of the mobile radio cell, which depends on the density of cell-tower coverage among other things).
Taiwan used this to help trace the movements of passengers from the Diamond Princess cruise ship (which had visited Taiwan and passengers of which had subsequently tested positive for the virus), and to alert locals who might have come into contact with them. Based on diverse data including tour itineraries as well as cell tower signals, the movements of Diamond Princess passengers during their shore leave were mapped. Then text messages were sent via cell-tower broadcasting to all mobile phones (~1 million in total) that had connected to the relevant cell towers at the same time as the cruise-ship passengers were in the vicinity.
A similar approach was pursued when sailors from the naval ship Pan Shi were found to carry Covid-19. The movements of the sailors were traced via cell-tower signals, and SMS messages then sent to 210,000 mobile phones that had logged into the same cell towers within 15 minutes of the sailors’ phones. As T.H. Schee notes, using SMS is more privacy-invasive than the cell-tower broadcasting relied on with the Diamond Princess. While the latter is rather like a radio broadcast to all nearby phones, sending SMS messages is more targeted and requires individual (targeted) phones’ numbers. It is unclear why SMS messaging was preferred to cell-tower broadcasts in the Pan Shi case.
It is notable that Taiwan made no use of smartphone contact tracing apps based on Bluetooth or similar. This contrasts with Singapore, where the government deployed the TraceTogether app by mid-March. While an app is now under discussion, T.H. Schee explained in our interview that since contact tracing with the means described had seemed to be working relatively well and the disease had so far been kept contained, an app was mostly believed to provide little extra value.
Indeed, it is unclear how useful even Singapore’s TraceTogether has been, since the voluntary app had only been downloaded by about 16% of the population as of mid-April – far from the 60% of the population that epidemiologists believe an app would need to be used by to have a meaningful effect (Economist).
Importantly, even the Singaporeans never envisioned TraceTogether as a replacement for human contact tracers. Jason Bay, who led development of TraceTogether, himself stresses that “contact tracing should remain a human-fronted process”. Very different physical situations with divergent epidemiological implications can produce similar or identical Bluetooth signals, that only trained human interviewers can disentangle.
Dr Chan, too, stressed that in Taiwan digital tools were never intended to replace human contact tracing. Indeed, she said that as of 25 April, the mass tracing/alerting via cell-tower signals had not actually led to the discovery of any additional Covid-19 cases.
Taiwan required people suspected of carrying Covid-19 or arriving from designated risk areas to quarantine for 14 days, under threat of fines up to NT$1 million (~€30,000). During this time they received a daily allowance of NT$1000 and sometimes reportedly also free access to online gaming and movie streaming services.
According to media reports, Taiwan has been thus quarantining up to 86,000 people at any one time. Quarantine adherence was strictly monitored through digital and “analogue” means. According to T.H. Schee, substantial police forces were deployed on the streets to help enforce adherence. Monitoring by neighbours, neighbourhood wardens, housing-compound staff and similar seems to have been important, too.
On the digital side, quarantinees had to provide their mobile phone number and the address at which they were quarantining to the authorities, who used these to track their location via cell tower signal triangulation. They were also phoned twice daily to check that they were with their phones. If the signal suggested they had broken quarantine (or if the signal failed) alerts were sent and police or local health officials came to check on.
The geofencing system was developed by the government in cooperation with Chunghwa Telecom, one of the island’s main network providers. Reportedly, only “the health minister” (it is unclear whether this referred to the minister personally or his ministry) has full access to the system, while local health authorities can only see the quarantine cases within their jurisdiction. At Chungwha, three persons are said to have access to the system (presumably for operational reasons). All personal data is to be deleted after the 14-day quarantine, and the entire system erased after the pandemic has passed. Government officials have also stated that audits will take place to ensure that no copies of the data are retained.
In practice, this so-called “digital fence” generated a certain volume of false positives. Cell-tower signals sometimes proved imprecise. Phones ran out of battery. In late March the government claimed that about 1% of alerts were false positives. It is unclear how many false alerts that corresponds to in total, but with almost 100,000 people under quarantine at different times and some individuals suffering multiple false alerts, the number might run into 4 digits. In one instance, T.H. Schee told me, police in the city of Taoyuan received 741 alerts within a two-day period, 740 of which were false alarms. This has led Chungwha to explore the development of an additional, voluntary, GPS-based app, though nothing seems to have been deployed yet, possibly also due to concerns over data protection.
More importantly, perhaps, while false alarms caused at least some quarantinees considerable annoyance, it appears that so far they have not led to penal consequences or heavy-handed police behaviour. At least, T.H. Schee – who has followed Taiwan’s digital Covid-19 response closely – knew of no such case when we spoke. Nor could the author find any media reports about such behaviour.
Taiwanese authorities also used cell-tower data as well as, possibly, the country’s electronic “eTag” highway toll-charging system to map aggregate population flows during Taiwan’s Spring Break (Tomb Sweeping Day) holiday. The aim was to prevent excessively large or dense accumulations of people, such as would heighten transmission risks. When such accumulations occurred, cellular broadcasting was used to alert people in the respective locales to the danger and ask them to adhere to social distancing and/or leave the locale.
An important reason for Taiwan’s rapid and effective response to Covid-19 was that it had the necessary institutional, legal and technological framework in place already. After the experience of SARS in 2003, Taiwan created an institutional structure to respond rapidly to pandemics. A Central Epidemic Command Center was set up to coordinate government responses, and the Communicable Disease Control Act updated to provide a legal framework. Subsequent legal challenges and Supreme Court Judgements clarified the constitutionality of this framework and the powers it conferred on the authorities, while leading to further amendments and safeguards for civil liberties.
Physical and human infrastructure also appears to have been rapidly put in place, or already existed. This includes trained contact-tracers, but also hand sanitisers and fever checks at most public and private commercial buildings, and the cellular broadcasting system (used hitherto mainly for earthquakes and typhoons).
To summarise, we can note that Taiwan made substantial use of digital tools to contain Covid-19, but deployed them as part of a larger, human-led and human-centred response. Digital tools were used to support, not substitute for human effort. This was no “digital-only” technological solutionism. Indeed, where the Taiwanese were forced to resort to technological means only (as with the mass contact-tracing/alerting) the results seem to have been less impressive than perhaps hoped.
Relatedly, Taiwan’s effective response was made possible by the fact that necessary institutions, legal frameworks and infrastructures had already been prepared. While the warning shots of SARS and later MERS, Ebola and H1N1 were heard around the globe, some countries evidently responded to them and prepared, and others did not.
Finally, this pre-existing legal framework also allowed Taiwan to, in effect, suspend significant data-protection rights and safeguards in order to ensure necessary information flows to contain Covid-19 – but to do so in a way that, while deep, was also narrow (relatively few individuals were affected), temporary, in line with the constitution, and came with safeguards. Thus, databases were temporarily linked and suspected Covid-19 patients had to accept detailed reconstructions of their movements and contacts, as well as strict quarantine orders.
In assessing these suspensions of civil liberties it is important not to fall back on orientalist tropes about “obedient Asians” or “Confucian collectivism”, as these are simply empirically wrong, and rejected by Taiwanese commentators themselves. Far from being “obedient”, the Taiwanese have created a vibrant, contentious democracy and civil society with strong social and protest movements. Accordingly, in assessing Taiwan’s response to Covid-19 from the standpoint of data protection and civil liberties, we should view and debate Taiwan’s actions as one possible, considered and measured response that a mature democratic society may take when faced with an unprecedented epidemiological threat.
Taiwan’s response can be seen as posing the question not only of how to weight the rights of individuals (e.g. to data protection) versus their responsibilities towards their fellow citizens (protecting their health and, ultimately, right to life), but – in as far as effective pandemic response may require rights curtailments – how we should choose between different possible menus of rights curtailments of rights in a pandemic. Taiwan’s decisions can be read as, in effect, trading relatively deep albeit temporary curtailments of the right to data protection of a fairly small number of individuals (those infected and quarantined) for the preservation of most other rights and freedoms – such as society-wide lockdowns would have curtailed – for the population at large. Specifically with regard to data protection, the Taiwanese polity can be read as having decided to significantly (but temporarily) curtail certain individuals’ right to decide how and by whom their data may be processed, while maintaining strong rules around purpose limitation, non-linkability, storage limitation and promise of audit.
How to choose between these kind of trade-offs is ultimately a political decision requiring broad societal debate, to be decided also in light of the specific political, institutional and infrastructural contexts. For instance, the question of whether Covid-19 infectees should have a right to decide for themselves whether to inform their contacts about their infection (whether through an app or by some other means) may be answered very differently if testing has to be restricted only to those in demonstrable contact with a carrier, or whether it is readily available to all. Similarly, whether safeguarding general rights of assembly (e.g. to demonstrate) are more or less important than minimising infringements of the right to data protection may be decidedly differently depending on the specific political circumstances and balances of social forces.
Taiwan used the years after SARS for these kinds of debates. Ours are only beginning.
~~~~~~~~~~
[1] For an overview see among other sources Wang, C. J., C. Y. Ng, and R. H. Brook, “Response to COVID-19 in Taiwan: Big Data Analytics, New Technology, and Proactive Testing”, Journal of the American Medical Association (JAMA), Vol. 323, No. 14, 2020, pp. 1341–1342.
[2] It is the author’s impression that the CCTV data is mostly analysed “by hand”, not by means of facial recognition algorithms, and that one tries to request the CCTV data in as targeted a manner as possible, i.e., just for those areas and times where there are gaps in the mobile and interview data – but this is was not clarified in the interviews and subsequent research.
Nicholas Martin is a senior researcher and project manager in the Competence Center for Emerging Technologies at Fraunhofer Institute for Systems and Innovation Research ISI in Karlsruhe