Citizen scientists: the power of people data to manage public health
Image credit: Dreamstime
When the World Health Organization declared Covid-19 a global pandemic in March 2020, very little was known about the raging infection. Testing capability was limited and data sparse. Yet one citizen science project managed to gather and process real-time information and insights faster than most, propelling the power of people-led science combined with technology to the fore.
As the UK entered its first-ever national lockdown, Tim Spector, a professor of genetic epidemiology at King’s College London, while cycling home through London, had what turned out to be a game-changing idea. To garner on-the-ground information, he decided to survey the more than 14,000 participants of Twins UK, a study he set up in the 1990s, about their experiences of coronavirus.
Seeing the potential of simple surveys to gather data at pace, Spector enlisted the team and technology behind ZOE, a precision nutrition app, of which he is the scientific founder, to launch the survey nationwide. Within days the Covid Symptom Study app was born. Twenty-four hours later it had more than a million users, and by July four million, making it the largest citizen science project ever.
The project, which a year later still has over a million loyal contributors, is now supporting the government to understand the efficacy of its vaccine strategy. Most notably, however, it has become a model of how citizen science, with the right technology, can transform public health and medical research approaches, reducing both costs and timeframes.
The Covid Symptom Study app asks people to log daily whether they feel ‘normal’ or have one of 19 symptoms. Test results are also recorded and by the end of April the app had collected around 60,000, including 10,000 positive results. Using machine learning and artificial intelligence, researchers at King’s College London looked at symptoms that clustered with the positive test results, against those clustered with negative results and trained an algorithm, with an 80 per cent predictive ability, to determine whether a person has Covid-19 depending on their symptoms. Researchers extrapolate this data across the population to create an estimated daily nationwide and local figure for people with symptomatic Covid-19.
This data was particularly useful when there was limited community testing during the first lockdown, which, according to Spector, enabled them to determine where outbreaks were occurring up to 10 days before the government did (which incidentally rattled the government and saw them nearly shut the study down).
Through their work, the researchers were also among the first to identify lack of smell and taste as a key predictor of the disease, publishing a paper in Nature.
“I’ve been doing studies for 30 years and never got a result in less than six months,” says Spector. “Yet, with the app, we included a question on loss of smell and taste, and we had the answer in a week. It’s an incredibly fast and agile tool.”
Overall, data from the app has informed over 315 scientific papers, assisted the government’s testing programme by sending out thousands of tests to app users, as well as home stool and blood collection kits for research into long Covid.
With public funding, the researchers are now monitoring vaccine efficacy and side effects, something the government is keen to understand, Spector says, given its “bold” approach allowing a 12-week window between both shots. The first paper, published in The Lancet, found up to 74 per cent protection with a single shot and low side-effects.
“We are directly impacting policy; if our daily reports had shown 30 per cent efficacy and lots of post-vaccine infections, the government probably would have changed its 12-week policy,” says Spector.
The study, he adds, will show faster than anything else the emergence of vaccine-resistant variants and identify groups that need a booster.
‘I’ve been doing studies for 30 years and never got a result in less than six months. Yet, with the app, we included a question on loss of smell and taste, and we had the answer in a week. It’s an incredibly fast and agile tool.’
The dictionary defines citizen science as “scientific work undertaken by members of the general public, often in collaboration with or under the direction of professional scientists and institutions”.
The practice has typically been used for tracking nature, such as migration and population levels. Though not completely new to medical science – biobank studies that collect samples from healthy people are essentially citizen science in their approach – however, technology is providing new ways to run studies.
In March, researchers at the Centre for Psychedelic Research at Imperial College London published the results of its ‘self-blinding citizen science’ study, which they say is the largest placebo-controlled trial on psychedelics to date, comprising 190 people (the second had 86).
The participants, who were already micro-dosing with LSD, implemented their own placebo-control measures at home, following PDF instructions. This required them to buy and prepare two batches of gel capsules – one filled with a micro-dose and empty ones – putting them into zip-bags, and undertaking a specially devised shuffling process, that involved generating random numbers with QR codes. Through this process the researchers knew what the capsules contained but the micro-dosers didn’t.
Throughout the month-long study participants were asked to guess what they were taking – a placebo or micro-dose – and from the findings the researchers determined that what mattered most about people’s experience was not what they were actually consuming but what they thought they were taking.
Balázs Szigeti, lead author and a research associate at the Centre for Psychedelic Research at Imperial College London, says the study demonstrates the power of the self-blinding method, which has never been tried before.
“Previously, the only way to do placebo-controlled studies was through extremely expensive clinical trials, but we’ve shown anything that can be put into a capsule can be tested for a fraction of the price,” he says.
The same study in a clinical environment would cost between one and five million dollars, according to Szigeti, whereas this study cost $10,000 (£7,100).
He recognises the shortcomings of such a trial. However, he says the idea is not to replace clinical studies, but to screen ideas before committing to the full cost.
Szigeti and his colleagues are planning another micro-dosing self-blinding trial, including heart-monitoring devices and patients with ongoing mental health diagnoses. It will run on an app called My Delica. He is also planning another with CBD (cannabidiol, a component of cannabis).
Increasingly researchers are keen to exploit the low-cost and fast results citizen science projects can yield. To help manage its workload, global independent health network Cochrane, which provides systematic reviews and other synthesised research evidence to inform health decision-making, developed the Cochrane Crowd, which asks volunteers to characterise data and research. The results are fed into an algorithm to classify a record with 99 per cent accuracy.
“It’s disruptive,” says Anna Noel-Storr, information specialist and project manager at Cochrane Crowd. “It has allowed us to scale; through this approach we can now handle more records than ever before.”
For another study, researchers at Imperial in December asked people to download the free DreamLab app, from which they could share the computer power of their phone to create a ‘virtual supercomputer’. They used this power to processes billions of calculations, using experimental methods to find new drugs and superfoods to fight Covid. “Tens of thousands” have participated, according to the project lead, Dr Kirill Veselkov, from the Department of Surgery and Cancer at Imperial, reducing the work timeframe by two-three times.
However, Veselkov says one of the main benefits is communicating with a highly engaged community who want to learn and contribute. “This is a particularly exciting part of citizen science projects: our work is usually published in a paper and read only by experts in the field, but this reaches thousands of people,” he says.
It’s obvious there are limitations and challenges to citizen science; people who interact with apps, for example, are not reflective of the entire population, participation is uncontrolled, and people may misunderstand or not answer questions properly or honestly. These factors can create a lot of noise in the data.
Sebastien Ourselin, professor of healthcare engineering at King’s College London, whose team of 12 manage the data collected from the ZOE app, says they had to find ways to clean the data. For this they built software, mostly from scratch, and used algorithm methods such as inverse probability weighting that estimated and redressed the balance.
“What makes it work is when you start to reach a critical mass, when you have a million daily users, you start to reach a certain level of signal which is robust enough and the biases become less apparent,” explains Ourselin.
However, though not a replacement for clinical studies, Szigeti says testing medicines and other healthcare approaches with citizen science adds a different value.
“A patient looking for medication for a condition is not necessarily interested in whether the medicine works in a highly controlled clinical study but whether it is going to work for them in the context of their daily life. Citizen science can answer that question by trialling it in a less controlled environment. It could complement clinical trials,” he says.
Nicholas Timpson, professor of genetic epidemiology at Bristol University, who runs The Children of the 90s longitudinal study and has used data from the Covid Symptom Study, says now is an inflection point. “I want the visceral understanding the public now have of studying disease replicated outside Covid-19; we need to harness this opportunity to get people to understand the value of their data. If we get that right, it’ll transform how we do population science in the future,” he concludes.
The Covid Symptom Study, which Spector believes has maintained its loyal following due to the constant feeding back of timely information and his personal updates, is funded for another year.
After that, Ourselin hopes it will continue to study and support long-Covid sufferers, perhaps keeping the cohort together as a biobank for other relevant studies. They are working with HealthData Research UK to do this, with all data to be open access.
“I think the app is just the tip of the iceberg. Citizen science offers many, many opportunities,” says Ourselin. “Beyond the science, I do strongly believe many people actually felt better because they were contributing to the fight against Covid-19 and without this study they would not have had that feeling.”
The future of personalised healthcare?
If personalised healthcare is the future, then so is citizen science-type health monitoring, according to its proponents.
Sebastien Ourselin of King’s College London believes regularly reporting personal health information could become part of governments’ public healthcare strategy, helping to manage seasonal influenza and other diseases.
“I think our study demonstrates, at a proof-of-concept scale, that citizen science can really provide proactive information for minimal cost,” he says. The Covid Symptom Study costs roughly £1 per contributor.
Through a national healthcare app, people could be asked to take part in regular surveys to monitor seasonal flu infections and forecast hospitalisations at a local level, he explains. “This would help improve flu management every winter and allow public health agencies to warn people about strong strains, so they can take preventative measures,” he says. “People, in the UK especially, are very willing to engage.”
His colleague Tim Spector agrees that personalised health monitoring is the future. That’s what the ZOE app he co-founded was originally created for; since 2018 it has run a remote gut health study, called PREDICT, the findings of which form the basis of a precision nutrition service, already launched in the US, which provides personalised meal plans based on results of home health-test kits.
“People will begin to do blood and stool tests at home, filling in online questionnaires, using glucose and fat monitors, scan foods for their health score; everything will be personalised through handheld and web tools, all linked to wearables,” says Spector.
However, because of a preoccupation with security, there are doubts that the NHS currently has the right approach to benefit from this innovation. “ZOE or Google can do things that my hospital or university couldn’t do and that’s madness... why would you trust a start-up company more than an established medical university?” he asks.
Despite eventually working with the Covid Symptom Study app, the government has approached it nervously. The Department for Education resisted plans to run a study monitoring school infection rates, telling schools not to take part. Similarly, the Department of Health would not allow the team to put up posters in vaccine centres advertising the app’s vaccine trial, even though Spector says the Chief Medical Officer and Chief Scientific Officer supported the idea. This, he says, could have brought in “10 million more users’’ to improve the study.
Another prohibiting factor could be an apparent lack of trust in government-led tools, probably exacerbated by the failure of its test, track and trace app.
“What’s clear is many people like our app because they don’t trust government data or the BBC any more. They feel it’s often distorted,” says Spector. “I think there’s going to be a clash – our app is like a giant pilot project of all future health apps.”
Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.