Image: Caley Dewhurst (ODI)

Covid-19 symptom tracking research worknote #2

Tue Sep 15, 2020
$download_content = get_field('download_content');

What data are symptom tracking initiatives collecting, how are the organisations involved in these initiatives using it and are they increasing access to the data, beyond this initial usage?

Over the last couple of weeks, our Consultant James Maddison has been looking at the data that symptom tracking initiatives have been collecting, how the organisations involved in these initiatives are using it and whether they are increasing access to the data, beyond this initial usage. 

What data is being collected?

In order to identify the types of data that users are being asked to share with symptom tracking apps, I went through the process of using 24 of the apps or services and filling in their Covid-19 questionnaires. You might notice that this number is a lot smaller than the original 38 initiatives that we looked at; this is primarily due to some of the other initiatives not collecting the primary data, or me not being able to access the questionnaires, mainly due to a requirement to provide proof of residency (a phone number or postcode). I was also careful not to submit any false information to any of the questionnaires, so the research might not reflect questions in some apps if they operate using a decision tree process. 

Based on the 24 symptom tracking applications that I was able to access questionnaires for, it seems that symptom tracking data falls into 11 broad categories:

  • Personal and demographic information
  • Employment
  • Household
  • Location
  • Pre-existing health
  • Symptoms of illness
  • Covid-19 status
  • Travel
  • Interaction with others
  • Changes in lockdown
  • Mental health and mood

You can find the full list of questions available in an aggregated list here, as well as a breakdown of how many initiatives asked each question. 

Points of interest

Going through the process of comparing the types of data that different initiatives have been collecting has been fascinating, as there are lots of unexpected insights that have surfaced. I’ve picked out five areas that I find personally interesting to share in this worknote:

  • How do questions about wealth and education help address Covid-19 health concerns? 

A few of the symptom trackers ask questions about income, level of education, and one even asks how you feel you compare to others in your country of residence. While you might be able to guess at how this data is being used (wealth + location + Covid-19 test results might help healthcare investment decisions, for example), it is difficult to identify a clear reason as to how answers to some of these questions might help healthcare professionals, researchers or policy makers to make better decisions about public health.

  • How many ways can you cook an egg? 

A large majority of symptom trackers asked at least one question about the user’s age, which is not unexpected. What is interesting however is the number of different ways in which trackers asked this question. Some asked for a full date of birth, others asked for a specific age and many asked which age category the user falls into. And this is just an example of one category of question where this has happened. This leads to a larger question about standards; how can we compare similar data across different symptom trackers when the answers are given in different formats?

  • Sex and gender 

Early examinations of Covid-19 cases and deaths in multiple countries have found that men are more at risk of severe illness and death than women, so it seems fairly obvious as to why symptom trackers are collecting data about an individual’s assigned sex. However, a number of symptom trackers also ask for users to disclose their gender, and some ask for users to disclose their gender without asking questions about assigned sex. Hopefully, the questions about gender are being asked to address a specific research question, and not because gender and sex have been amalgamated into one category. 

  • Changes over time

A small number of the symptom trackers that were looked at ask questions about how lockdown has affected the user, and whether they have seen noticeable changes over time. This includes areas such as diet, weight, physical activity and sleep. These questions are nearly all asked by private sector led initiatives, which might suggest that symptom trackers are also being used as an opportunity for private sector companies to collect data, which can help inform their product or service offerings in future. 

  • Mental health 

Roughly a quarter of the symptom tracking applications ask at least one question about how the user is feeling, with the most common question being around feelings of anxiety. Given the potential effects that lockdown might be having on people’s mental health, and an uncertainty around how long and often people might be in lockdown, asking questions on this topic could be useful to mental health researchers to produce research insights that can be used to support and promote positive mental health and well being.

Outputs

Given the range of stakeholders involved in the symptom trackers examined, it is no surprise that there are a variety of outputs that are being produced. Outputs seem to fall into three rough categories:

  • Individual guidance – Once a user has filled out the questionnaire, they are presented with a set of recommendations. Many of the applications, such as the CDC Symptom Checker, provide simple recommendations such as to see a health specialist.
  • Research insights – Many of the initiatives are producing research insights. Some of these insights are shared at an individual level, like Your.MD’s tracker, which gives you a condition comparison after filling out the questionnaire, based on how you compare to other respondents. Other initiatives are sharing insights publicly, like the COVID Symptom Study, which is producing visualisations such as how Covid-19 cases in the UK are changing day by day, or how they have changed specifically in the lockdown period. A number of the initiatives which are producing insights are also sharing data specifically with health services, such as the NHS. Project Oasis, for example, is a data intermediary which collects data from third-party applications and shares it with the NHS to support the response to Covid-19 in the UK.  
  • Improvements to existing services – Some trackers, for example those that are attached to dedicated health companions, are collecting Covid-19 data to improve their existing services, such as the support they can offer to individual users. These services aim to support users to improve their general health, rather than to specifically address Covid-19. 

The level of detail required from users seems to differ greatly, depending on the outcome that the symptom trackers are trying to provide. For example, the Ink C-19 app doesn’t require anything other than your location and an indication of whether you are well, have symptoms, have been diagnosed as sick or have recovered from Covid-19. Once you have reported how you are feeling, a coloured circle representing your response appears on a map covering roughly a half-mile radius around your location, so viewers know that someone within that rough area has reported their status. In comparison, the Evergreen Life app – which usually operates as a general health and wellbeing app, but has now also integrated a Covid-19 check into its functionality – asks for more detailed information, such as your current symptoms, whether you’ve been tested, if you’ve been in contact with someone who has tested positive and whether you’d like the information you’ve submitted to be shared for NHS research purposes. This approach has yielded more detailed outcomes, such as visualisations of how the number of people reporting Covid symptoms has changed over time, or heatmaps of how many people in a particular area are following existing government guidelines. 

Could symptom trackers be more open?

The process of data discoverability proved to be fairly time consuming. A few of the symptom tracking apps that I looked at give users a high level overview of what they will be asked, but these descriptions rarely provide enough information to understand what data is being collected. In most cases, you have to complete the questionnaire in order to find out what questions you are going to be asked, and this has to be done in sequence, so you can’t skip ahead to look at the questions without filling in the previous page.

Although this approach might work for the organisations involved in these particular initiatives, I personally feel that this lack of transparency is quite a big barrier to collaboration. There could be researchers or health organisations with skills or resources to help these initiatives create greater impact, but if they can’t even find out what data is being collected by symptom tracking initiatives, then it could make it difficult for them to identify which organisations they might want to work with or support. The approach that Track Together has taken with publishing open datasets, data descriptions and calls for collaboration greatly increases the opportunity for organisations to identify a beneficial collaboration, and I would recommend that more organisations operating in this symptom tracking space consider taking a similar approach to working in the open.