COVID-19 contact tracing apps were suggested as saviors. They sometimes delivered

Privacy fears didn't materialise, but bungling did

COVID Logfile IV As the SARS-CoV-2 virus and the COVID-19 disease it creates spread rapidly across the world in early 2020, governments that grasped the gravity of the situation wondered if technology could help them control the pandemic.

Some considered open-source intelligence gathered by the likes of Facebook and Google. Others wondered if records from ATMs or credit card payment terminals could be useful.

The intent was the same in both cases: governments felt that if they could learn where infected people had been, and who else may have encountered them, it would be possible to identify those at risk – and tell them they should isolate instead of continuing to mingle and maybe spreading the virus. Such data could also help contact-tracers, health professionals who were already adept at chasing diseases through society.

A little help from Bluetooth

In Singapore, the government hit on another idea – using the Bluetooth radios in mobile phones to track movements.

"Most people already carry a pocket contact tracer with them at all times: their mobile phones," the island nation's Government Technology Agency wrote on March 25, 2020.

Singapore put those sensors to work with an app called TraceTogether.

TraceTogether constantly sought out other Bluetooth-enabled devices that ran the app, then logged any encounters. Singapore's apps, and those that followed elsewhere, relied on apps sharing randomized keys via radio. Such Bluetooth encounters saw those keys exchanged, along with data such as signal strength as an indicator of proximity, plus date and duration of exposure.

Singapore built TraceTogether on top of BlueTrace, which made it possible to use Bluetooth reliably for contact tracing.

It required users to register and inform authorities if they contracted COVID-19. When that happened, TraceTogether would draw up a list of all other users who had encountered the infected person. Officials used that data to trace contacts and require them to isolate before they infected others.

The idea took off around the world – helped by the fact that Singapore open-sourced TraceTogether and BlueTrace and an advisor to the World Health Organization recommended the idea for wider adoption. Google and Apple weighed in with their approval, and eventually jointly delivered an API that allowed Android and iOS devices to provide Bluetooth-based decentralized, pro-privacy exposure notifications - an approach that emphasized warning individuals of their exposure to infected people, rather than health organisations controlling when notifications were sent..

By April 2020, various nations were building their own contact-tracing apps, with or without the decentralized API. Furious debate had erupted over whether user privacy and pandemic control were best achieved using a central repository of all data gathered by the apps so that governments could tell citizens to isolate, or the decentralized exposure notification model that gave users more control over how and when their data was shared and warnings issued.

What happened next differed markedly from country to country – and sometimes even state to state – and along the way showed that technology is not immune to politics and would not be a panacea.

Beacons of hope

Australia's COVIDSafe app – based on Singapore's TraceTogether – was launched by the federal government in April 2020, with much fanfare.

Australian prime minister Scott Morrison likened using the app to sunscreen – the kind of sensible precaution many Australians apply when they work or play outdoors – and said it was the key to loosening restrictions.

"We are now calling on all Australians to download the COVIDSafe app to help protect you, your family and your community from further spread of COVID-19," health minister Greg Hunt stated in a press release.

But the app was plagued with problems, as early analysis suggested it leaked information about whether users were actually running the software or merely had it installed. The app also changed session IDs so slowly it was unlikely to provide precise data about users' movements to contact tracers. Months after launch, the app still performed poorly on Apple devices.

Singapore had similar performance problems, but realised that some residents did not have smartphones. The nation therefore moved away from total reliance on phone-to-phone encounters and introduced a wearable Bluetooth device and 40,000 Bluetooth receivers in public places to detect them.

Amity in the UK

After giving up on an earlier design that didn't work properly, England and Wales' National Health Service COVID-19 app arrived in September 2020, employed the Google-Apple contact-tracing, and did rather well with six million downloads

A May 2021 analysis published in Nature found "approximately one case was averted for each case consenting to notification of their contacts", and suggested the app averted between 284,000 and 580,000 COVID-19 cases between September and December 2020.

Approximately 1.7 million exposure notifications were sent to users, and the Nature paper states the app's effectiveness was "similar" to that achieved with manually traced close contacts.

Things went less well in Australia. While COVIDSafe was installed over seven million times, by November 2020 just 735 users who tested positive consented to share their status. Just 2,579 potential close contacts were identified

Singapore did better. In November 2020 health minister Gan Kim Yong told parliament that TraceTogether had helped identify around 25,000 close contacts, of which 160 tested positive. "It has enabled the early identification and isolation of cases, faster than would have been possible with manual tracing," said Gan.

By January 2021, then education minister Lawrence Wong said digitization had sped contact tracing from two days to mere hours. He quoted the total cost of TraceTogether and SafeEntry at a combined SG$10 million ($7.4 million) and expressed a firm belief that the system worked.

"Results speak for themselves," said the minister, who credited the program's success to resident confidence in officials and technology. "Our ability to do all of this is not just because of the TT program, it's fundamentally because of the confidence and trust that Singaporeans have in our system, the way we handle the data and the way we go about doing the contact tracing," said Wong.

According to the Center for Strategic and International Studies, government assurances convinced Singaporeans that TraceTogether protected their privacy. That may have been an easy job given that Singaporean culture is generally accepting of surveillance. "The amount of data that TraceTogether collects is not unusual in Singapore, considering that the government already uses other surveillance methods, ranging from CCTV to drones," the Center blogged in February 2021.

California comes to the party

The USA did not implement a national app, but in December 2020, the State of California rolled out CA Notify – a mobile app for notifying people about possible COVID-19 exposure, but not sharing info with contact tracers as happened elsewhere.

To hear those involved with the project tell it, the app has been a success while also illuminating the complexities and limitations of using technology to address public health challenges.

Dr Christopher Longhurst, chief medical officer and chief digital officer at UC San Diego Health (UCSD), oversaw a September 2020 pilot test of the app. His team currently administers the app for California under a $4 million contract.

"We were testing all of our students, as well as employees, on campus, and then we were using that to launch the automated exposure notification. And to the best of our ability, we were hoping to be able to track it," Longhurst said in an interview. "What we learned is that it's so private, it's actually very hard to track."

State authorities judge the project to be a success. "Our own preliminary modelling suggests that CA Notify has prevented thousands of new COVID-19 cases and averted hundreds of deaths in California since its launch," the California Department of Public Health (CDPH) said in an emailed response to questions.

The California Department of Public Health (CDPH) said the app, available for both Android and iOS, has been downloaded and activated more than 16.3 million times and has produced more than 1.16 million exposure notifications over the past 14 months. That's in a state with about 24.6 million people aged 18 or older.

Results were good.

"We estimate that over 250,000 CA Notify users who tested positive for COVID-19 have used the code they received from the CA Notify system," the CDPH said. "Therefore, the system was able to anonymously notify those individuals that may have been exposed."

But CA Notify requires users who receive a positive COVID-19 diagnosis to actively share their status in order for contacts to be informed.

Many do not. The CDPH told us its best estimate is that 28 percent of codes are claimed and 12 percent are used – but it can't really say, because some notifications go to people who test positive but don't use its app.

The department does believe that as the number of COVID-19 cases in California increased, so too did the number of people sharing their positive test result with the CA Notify system. The CDPH admits, however, that it has limited data to analyze due to the app's privacy protections.

Longhurst said that one year into CA Notify, a symposium was held at which his group shared outcomes, based on the same public health models used in the UK, showing tens of thousands of avoided infections and dozens or hundreds of lives saved.

"Again, it's difficult because of the privacy preserving nature [of the app] to really track this down to individual data elements," he said. "We don't even have people's location."

Apps verses freedom

While California started to enjoy the benefits its app conferred, other nations turned theirs into political footballs.

In mid-2021, the NHS app came in for criticism as the volume of notifications and isolation recommendations was perceived by many as out of sync with policies that allowed – and even encouraged – greater freedom of movement.

The UK government, which had proudly released statistics about the app's adoption, now declined to release data about falling usage, or whether users persisted in using the app's location check-in functions while disabling contact-tracing functionality.

In May 2021 the UK developed a "COVID Pass" - a certificate denoting vaccination status, as up-to-date jabs became a requirement to enter some venues - and made it available in the main NHS app. But later in the year systems related to "COVID Pass" crashed as soon as the lockdown restrictions that made the pass more useful eased – again making the mix of COVID and apps, and their performance a political football.

Australians, meanwhile, just stopped caring about the app – in part because State Government apps facilitating QR code check-ins to almost any venue became mandatory and took on the role of informing citizens about encounters that required isolation.

QR code check-in data was also more useful than info from COVIDSafe, which had decided that residents of apartment buildings separated by bricks and mortar represented worrying encounters. Another supposed incident of concern involved people sitting in separate cars queuing for COVID-19 tests.

An Australian contact tracer who we spoke to on the condition of anonymity said they were never asked to use data drawn from the federal government's COVIDSafe but were often told that leads they were asked to chase came from QR-code check-ins to locations required by state government apps.

The app's poor performance became legendary and opposition parties started to use it as an example of government bungling and waste. They seem right to have done so because in February 2022, Australia's Health Department told the app had recorded 1.65 million digital handshakes and identified 2,829 "potential close contacts … from more than 37,600 encounters."

The department also offered us a case study of a state government using data from the app to "identify around 81 close contacts, including 17 contacts that were not identified by manual contact tracing." But at the time we were offered that data, the state in question had recorded almost 1.3 million cases of COVID-19.

A February 2022 analysis of the app's efficacy in the state of New South Wales also mentions 17 close contacts being found by a state government – but reveals they weren't properly followed up. By contrast, conventional contact tracing techniques found 23,500 cases across the state in the same period.

The analysis points out that contact-tracers weren't involved in the app's design, so it is little wonder the software's output was not useful.

Singapore, meanwhile, had also made venue check-ins mandatory.

Lessons for the next pandemic

The four nations we've considered have each taken different lessons from their experiences.

In March 2022, the Australian parliament's Select Committee on COVID-19 considered the app in a report [PDF] assessing the national response to the virus. By that time the app had cost over AU$9 million ($6.7 million), with almost a third of the spend going to Amazon Web Services.

That document says Australia's State governments gave up using data from the app in September 2021, the Federal government stopped updating the software a month later, and that the app's many flaws were ignored for months meaning it was never fit for purpose.

The Select Committee had just one recommendation about the app: "the Australian Government cease any further expenditure of public funds on the failed COVIDSafe application." It will be a brave politician that tries to re-introduce anything like COVIDSafe in Australia.

For England and Wales, the NHS app is still operating and has been funded until the end of 2022. But usage has plummeted, and its future role is uncertain. While the NHS' everyday app continues to evolve, the UK's government defunded another app called "Zoe" that tracked COVID-19 symptoms and allowed epidemiologists to predict the virus' spread. It is unclear if that signals an end to interest in contact-tracing apps in general.

Singapore has made TraceTogether and check-ins an ongoing feature of its society. In January 2021 the government amended TraceTogether's terms and conditions – to allow data sharing with the island nation's police under certain circumstances.

And in October 2021, minister of health Ong Ye Kung stated that the use of TraceTogether and SafeEntry would continue as it frees human contact tracers to work on cases at sensitive environments such as schools or aged care facilities. In January 2022, the minister said the tracking tools will continued to be used in settings like restaurants for the foreseeable future.

And in March this year, Singapore's government posted job advertisements for positions related to TraceTogether, sparking speculation that technology-enabled virus surveillance – like the virus – is going to be in Singapore forever.

In California, the outlook is less clear, in part because analysis of data drawn from CA Notify is undergoing peer review.

Longhurst said one lesson is that privacy fears have not been realized – as someone who has served for years as a CIO and now chief medical officer, he understands the privacy implications and said he sleeps easy at night with regard to CA Notify because none of the data stored is identifying.

Dr Susan Landau, Bridge Professor in Cybersecurity and Policy at the Fletcher School of Law and Diplomacy at Tufts University, wrote a book on such apps titled People Count: Contact-Tracing Apps and Public Health.

In a phone interview, Landau acknowledged there's value to these apps but also observed they have significant limitations and social consequences.

"When we first learned about [COVID-19] two years ago, we all had this image that you could stand on a supermarket line or could be on the subway or a bus, and easily get it from somebody standing next to you and you would never know that you've been exposed."

Apps were a response to that scenario, but Landau said COVID-19 is far more likely to spread between couples who share a bed than family members sharing a house.

Allowing that she's not an epidemiologist, Landau questioned how much exposure notifications help when the spread is so uncertain.

Landau also pointed ongoing debate about whether Bluetooth is an issue.

She referred to a study conducted in 2020 by Trinity College Dublin researchers who found exposure notifications apps performed poorly on a European commuter tram because the many metal components interfered with the signals.

UCSD's Longhurst does not think much of the study. The issue is not Bluetooth, he said, but the way the software is tuned to conform to CDC exposure criteria, which at the time were being within six feet (1.8 meters) for a period of 10 or 15 minutes.

"We are no longer trying to approximate like six feet for ten minutes," he said. "We're much more liberal now, which is what I want. I would rather have false alarms – give me some exposure notifications, I'll go test. That's not a big deal."

Landau, like Longhurst, praises the Google/Apple Exposure Notification API for being very privacy protective and suggests, as Longhurst has, that perhaps a bit less privacy would be helpful.

"If you think about the next respiratory pandemic, you might want some small percentage of the apps to collect location for the use of studies," she suggested. In any new pandemic, Landau also wants equity issues to be considered.

In England, she said, health authorities tested the apps on the Isle of Wight, where she said around 97 percent of the population is white, then conducted a second test that in the far more diverse London Borough of Newham.

Did it have to do with the population and distrust of the government?

"The uptake was much lower in Newham," said Landau. "Now was that entirely due to it being a second time that they were using the apps in Isle of Wight? Or did it have to do with the population and distrust of the government?

"It probably isn't the same kind of distrust of public health by minority populations in the UK, as there is in the United States. But this lack of testing the apps in different communities – and seeing uptake and understanding how that impacts public health – is, I think, a real failure."

Longhurst thinks this issue may be addressed in part through operating system-level integration of exposure notification technology. Apple completed the operating system-level integration of the contact tracing API in September 2020.

Baking these functions into the operating system means less is asked of the user – you can approve notifications with a single prompt rather than requiring someone to find the app and download it.

"From an equity standpoint, we're concerned about [the greater usage among iPhone users]," said Longhurst, noting that iPhone users on average tend to have higher socioeconomic status than Android users and that people in lower socioeconomic strata tend to be at higher risk from COVID-19 infection. "I would very much like to see Google build this into the Android operating system like Apple did with iOS."

Public health after all is not just a technological challenge. It has become a political one. ®

Singapore sections by Laura Dobberstein. California content by Tom Claburn.