Contributors to the not-for-profit application diligently submit their personal details – including location and age – before answering questions regarding their health status. This data has informed essential daily case estimates, local area graphs and UK-wide infection statistics. Findings from the study were among the first to confirm loss of smell (anosmia) and taste (ageusia) as a symptom of the virus, leading to an official change in NHS guidance.<\/p>\n
More than 4.6 million people have logged approximately 448 million daily health reports, 15 million tests and 1.5 million vaccines onto the platform so far. The project was awarded a \u00a32 million grant from the Department for Health and Social Care in August 2020.<\/p>\n
Wolf spoke with DataIQ about how ZOE adapted at pace to help tackle the most significant health crisis of the modern era. He highlighted the key role that users have played by generously submitting their data, and how continued public participation could revolutionise the way that health studies are conducted at scale.<\/p>\n
DataIQ (DIQ): What was ZOE\u2019s focus before the pandemic hit?<\/strong><\/p>\n Jonathan Wolf (JW): <\/strong>\u201cMyself, my co-founder George Hadjigeorgiou and our partner Professor Tim Spector founded ZOE a little over four years ago to tackle the problem of personalised nutrition. Tim heads up TwinsUK, a health research study that collects data on 14,000 twins worldwide. <\/strong><\/p>\n His research has shown that even identical twins can have very different outcomes with regards to susceptibility to disease based on a) what they eat, and b) the microbiome in their gut. He knew this was the case, but he didn\u2019t have the capability to conduct the study at scale or tailor dietary recommendations off the back of it.<\/p>\n That\u2019s where my background in machine learning and data science came in: I knew that if we collected enough information and provided enough tests for people to do at home, we could generate data from thousands if not millions of users. We could then give users guidance on what to eat to improve long-term health.<\/p>\n This vision became PREDICT<\/a> \u2013 the largest nutritional science study in the world. We were just finalising it as the pandemic hit.\u201d<\/p>\n DIQ: How did the idea to launch the Covid Symptom Study app come about?<\/strong><\/p>\n JW:<\/strong> \u201cWhen the pandemic first hit I clearly remember feeling that this is what it must have felt like for our grandparents when they went to war. One minute you\u2019re leading your normal life, then suddenly the toilet roll is going to run out, there\u2019s talk of food shortages and fears of the NHS collapsing. You have to do what you can in these situations to help out.<\/p>\n Tim gave me a call to say he\u2019d had an idea \u2013 which isn\u2019t out of the ordinary because Tim has at least one brilliant idea a day. He proposed that we tweak the app we\u2019d developed for the nutritional study to collect Covid symptom data. I thought it was a great idea, but instead of just focusing on twins, why not do it for everybody?<\/p>\n Given my data science background and what we\u2019d been doing with ZOE so far, we knew that if we could get enough people to log their information we could predict not just who had Covid, but caseload and infection hotspots too.\u201d<\/p>\n DIQ: The app was built and launched over one weekend \u2013 what did that process look like?<\/strong><\/p>\n JW:<\/strong> \u201cWe were in a race against time \u2013 back in March 2020 testing in the UK was limited and we were worried that the NHS could collapse like we\u2019d just seen in Northern Italy. It takes a couple of weeks for people to become hospitalised after being infected, and we felt we could help hospitals by tracking infections just as people fell ill to predict caseload in different areas.<\/p>\n We just thought, \u2018OK, lets try it.\u2019 Our engineering team basically didn\u2019t sleep as they built the framework. A team at King\u2019s including Tim and his colleague Dr Claire Stevens defined the survey questions, and I curated them into easy flow that could be done quickly to provide us with the core data. It needed to be simple because everyone had to be able to use it. It was built and launched in five days.\u201d<\/p>\n DIQ: You recruited a million users within a day of launching – how was the team able to move so quickly?<\/strong><\/p>\n JW: <\/strong>\u201cIf we hadn\u2019t already been doing the nutrition study we would never have been able to move as fast as we did. We knew the ins-and-outs of consent, GDPR and transparency because we\u2019d already been working with remote ethics sign-up. We used a lot of our existing backend capability and had a lot of prior experience with questionnaires. We used a new programme called React Native<\/a> to build the front end, which really sped things up.<\/p>\n We were in a pretty unique spot to be able to create a consumer-oriented data science app at scale. Most scientists work with much smaller sample sizes, and most data scientists aren\u2019t experts on the scientific method. Thanks to our association with King\u2019s, we were really in the right place.<\/p>\n When we launched, myself and George spammed our entire contact list asking them to fill out the survey. We stressed that users could really make a difference in the fight against Covid by sparing just a minute a day.<\/p>\n It went completely viral, no pun intended. We were seeing the figures jump by 100,000 people an hour \u2013 at that point it was clear we were doing something valuable.\u201d<\/p>\n DIQ: Getting users to freely submit data on a regular basis is the holy grail for anyone working in the industry – how have you achieved this so successfully at ZOE?<\/strong><\/p>\n JW:<\/strong> \u201cIt was perfect timing. Other than frontline workers everyone was locked down and feeling a bit useless. For many, participating in the study was a chance to do something that felt valuable. We\u2019re also very clear about the way we use the data, that we\u2019re a not-for-profit and our objective is to support the NHS. Importantly, the product itself is easy to use and accessible to a wide audience.<\/p>\n The study is a collaborative effort between us and our members and we\u2019ve worked hard to present it as such. The success of the project is down to every single person who has participated in the study and generously given us their data every day. We put a lot of effort into our communications back to the participants, including sharing insights with our <\/em>users. It helps that our information is the best available.\u201d<\/p>\n DIQ: The pandemic has led to heightened scrutiny of the way data is presented to the public \u2013 what\u2019s your approach to data visualisation?<\/strong><\/p>\n JW:<\/strong> \u201cWe\u2019ve had to bear in mind that we\u2019re communicating with members, rather than scientific journals, the NHS or scientists used to dealing with complicated numbers. It\u2019s one of our key missions at ZOE: taking cutting edge science and making it understandable for the general population.<\/p>\n Insights have to be understandable and accessible. As soon as we had our first models we built a simple chart showing the total number of cases, which has been very popular. We then used a third-party software to create a colour-coded map showing caseload in regions throughout the UK \u2013 allowing people to quickly understand where their local hotspots were.<\/p>\n When the data has changed our modelling has changed with it. For example, we had to create new models when the vaccination programme began because people weren\u2019t getting sick in the same way. If we have to readjust our data because we\u2019ve discovered something new, we\u2019re open about it. We\u2019ve got a small, hard-working team focused on making things understandable and transparent.\u201d<\/p>\n DIQ: What\u2019s next for ZOE?<\/strong><\/p>\n JW:<\/strong> \u201cThe entire team was working on the Covid study during the first few months of the pandemic \u2013 it just felt like the right thing to do. Later in 2020 we restarted our work on the nutrition product, and we\u2019ve since launched in the US and have had some really exciting feedback. What we have seen during the pandemic is that the quality of our diet has a dramatic impact on our likelihood of being hospitalised for Covid. Given this, we\u2019ve expanded our ZOE PREDICT<\/a> study to those at home, asking for opt-ins from our amazing community of members.<\/p>\n It\u2019s early days yet, but we\u2019re hoping that we can use the same approach used in the Covid study to understand how our day-to-day health affects our susceptibility to other critical diseases. I hope that this sparks a paradigm shift in the way that studies are conducted at scale. By utilising mobile phones, the internet and machine learning we can solve health problems in a way that\u2019s never been done before.\u201d<\/p>\n