Data Ethics Club: Oura Wants Its Wearables to Revolutionize Pregnancy Studies#

Article Summary#

Oura is a Finnish healthcare company that sell wearable rings which collect biometric data and monitor the user’s health. The rings are “continually collecting data on 50+ health and wellness metrics” of the wearer, generating detailed reports on sleep, stress, exercise, and more. The cost of an Oura ring starts at $299 with monthly $5.99 subscriptions for access to the full suits of data.

Oura is partnering with the Scripps Research Digital Trials Center to further pregnancy research by analysing retrospective biometric data of 10,000 users who have been pregnant. Shyamal Patel, Oura’s Senior Vice President of Science claims that conducting studies in this way means that “we can actually do science in the real world, instead of doing science in this sort of constrained, clinical lab type of setting”.

A retrospective study means that researchers can use data from very large databases, and research can be conducted much faster than the several years a pregnancy study would usually take. The article states that “researchers can shorten research timelines by simply determining which members wore their rings while pregnant and then seeking their consent to participate in the study.” To give consent, participants are directed from the Oura app to the Scripps’ MyDataHelps platform where there is information about the study design, consent form, and questionnaire. The total time to participate is estimated to be “less than 20 to 30 minutes.”

The article goes on to discuss the ongoing federal funding cuts in the USA and injustices in access to healthcare. Black women, for example, have been historically under serviced and are disproportionately impacted by pregnancy complications. Dr. Robin Wallace, a medical advisor for Planned Parenthood, emphasises the importance of advancing maternal health care and suggests that wearables could be covered by a patient’s insurance as part of maternal health care bundles.

Discussion Summary#

What are the downsides of retrospective studies?#

From the large pool of data that Oura holds, we assume that the study will selectively pick out what is deemed as relevant. As the data has not been specifically collected for the study, it will need to be retroactively selected or shaped in a way that can be appropriately analysed, which could be expensive. To select data, those conducting the study must first be clear about exactly what data has been gathered in the first place. Gathering data first and finding a purpose later means that there will be things that have been just put in for the sake of it, hoping that something useful will come out of it later.

Contrasting retrospective studies, prospective studies follow participants over time and gather data for pre-defined purposes. Prospective studies are generally deemed less susceptible to bias or confounding variables and are organised in ways that make it possible to target certain aspects and tie things together. Retrospective data may contain more variation in peoples’ behaviour, as particular variables have not been isolated, which could be good or bad.

The way that the data has been collected will affect the study’s findings. We wondered how intrusive the data collection is, and if it is representative of actual behaviour. As the ring is small, people may forget they’re wearing it and behave fairly naturally. Alternatively, people may be aware that they are wearing it and being tracked and adjust their behaviour accordingly. Oura rings are aimed at helping people to eat more healthily, walk further, and prioritise sleep more. Yet, these behaviour modifications will influence the data that has been collected and how representative the sample is of the wider population. Some types of data will be missed by focusing specifically on the ring, for example, the ring collects heartrate and temperature but perhaps not always diet or mood. Retrospectivity also means it is more difficult to design the study in ways that facilitate recruiting a representative sample from the target population.

Amongst the millions of Oura users, the company estimates that there are thousands who have been pregnant while wearing the ring; we wondered how the company plans on identifying those people. Asking users that have explicitly noted on the app or who the company predicts have been pregnant has the potential to cause distress, as it is a sensitive topic and people may have had traumatic experiences. Reminding people of this via what is a very personal platform, in an automated and detached way, could be upsetting. Predicting who has been pregnant also raises privacy and autonomy concerns (reminiscent of when Target predicted a teenager was pregnant before her family knew).

It is unclear whether, if there is no human interaction as part of the study, participants are provided with a level of care or the option to withdraw. We wondered if users would opt in if something traumatic has happened, and if this is discussed in the sign up page. Even if people feel do comfortable at one stage to sign up, the potential for distress especially with pregnancy is huge, and peoples’ level of vulnerability may shift over time as may their comfort in taking part in the study.

A retrospective study design raises questions around informed consent. It is unclear if it is possible to obtain properly informed consent if participants submitted data without knowing what the end use of it would be, or that it would be used for medical research. If users knew at the time data was being collected that it would later be used for research, their behaviour or approach to health may be different. We wondered if people could grant consent for specific timeframes or types of data, or if signing up for the study grants access to all their data. If people are able to select the timeframes, this would make the data more irregular and could cause issues for analysis. We also wondered how consent of the child is handled, and if the child is able to withdraw if they one day decide they don’t want their data involved.

Oura’s website is not very transparent regarding the company’s approach to data privacy. The privacy policy is slightly hidden in the footer, and otherwise there is just a vague statement on the “integrations” page that: “you are in control of your data at all times. You choose who it’s shared with, and when.”. The privacy policy states that consent for data collection and retention can be provided “through your actions, such as by adding sensitive personal data into your notes, or by adding health related tags”. It is not explicitly stated how consent is handled for medical research studies.

We wondered how easy it is to explain to potential participants how their data may be used, and how much agency people have over their data. It’s questionable whether consumers in general truly consent, as people don’t tend to read the terms and conditions of what happens to their data, but are nevertheless locked in after accepting them. Oura has a subscription service, and we wondered if you are still able to access your own historical data if you stop renewing the subscription. It is unclear how the study would work logistically, such as the geographical implications of data, how the data would be transferred with respect to GDPR (which states that you own your own data), and who would be able to see your data.

Research within private companies tends to be less clear about boundaries, although partnering with an institution may lead to more transparency. Retrospectivity also means that the study does not need to be registered anywhere, as the data has already been collected, so there is very little public information on the study. We would be interested to know what the process is that researchers go through to access Oura data.

Whilst there are questions surrounding informed consent and duty of care, we do appreciate that medical studies are difficult and there is a troubling lack of research on pregnancy. A separate point asks if we should be collecting the data at all, but given that it is already being gathered and stored, it makes sense to get more value out of it. Oura’s data could be a good starting point for feasibility and exploratory studies, investigating whether particular types of data could be used for particular tasks, and which types of data should be collected. There are whole fields of data science that have been created from data that was collected for other reasons, such as digital footprint data, and there has been some really interesting work to come out of these areas.

At some point, you have to accept the data that you have and be pragmatic. With retrospective data, instead of designing a study and then doing data collection, you end up fitting the study to the data - this is a different approach but still valuable. The advantage of collecting more data than is relevant for your specific question is that it gives the opportunity to ask other questions that come up in the future. Data may also be more authentic, as users know they are having data collected but don’t opt into the study until it has already been stored, so their behaviour isn’t directly influenced.

While the ring has been positioned as a ‘life style’ ring rather than one with a medical purpose, how might growing health research and study initiatives (like the article shared for discussion) challenge that positioning, and how may that affect user perception and interaction with the ring?#

The point at which it matters to call something “lifestyle” versus “medical” could be legal, as medical products will have more hoops to jump through and delay moving the product to market. Naming also ties into public consciousness, linking the product’s perception with particular themes like health and fitness rather than notifications.

Some of us were a bit uncomfortable with the transition from collecting data for “lifestyle” purposes to using it in medical research. Oura already has various pipelines and platforming in place, but we worried whether the device is good enough to be used for medical research if it is not a medical device. It is not clear how sophisticated the device is in medical terms and what the quality of data being collected is. Gathering “lifestyle” data and then using it for “health” misses the boat for designing scope. Data was collected for one purpose and now needs to be pushed into a different shaped box, which could cause information loss. Questions also arise regarding sensitive data storage when transitioning to medical research settings, where it is important to withhold personally identifying information.

Transitioning from “lifestyle” to medical could change the way that users interact with the ring. If it is associated as purely lifestyle, it may be taken less seriously than if users perceive it as a medical device. For example, if the ring signals that your heart rate is very high, users might be more concerned if they see it as a medical device than if it’s seen as a lifestyle product. The way that the information is presented is important, for example, if it is presented in an alarming way (like an alert) it may cause issues if users do not have the medical knowledge to interpret the information being provided to them. This could be problematic and cause undue concern if the product is not up to medical standards. Even if the model is 99% accurate, still 1% of the population will get an inaccurate result without being equipped with the knowledge to understand it. We wondered how much data the model needs about you before it is able to make predictions about what is “normal” for you.

Moving into the medical domain raises potential conflicts of interest. The key problem is the underlying commercialisation of the ring, as the data is likely to be used for making money. Oura’s underlying motivation is to make profit, and their duty is to their stakeholders. We wondered what exactly are the ways that Oura is benefitting from the study, and what their strategy is. It may be in their interest in the long run to appear to contribute to the public good but could complicate the way they are viewed by legislators. There could be legal implications if data was used for insurance purposes. Oura is open to market whim and may at some point be bought by another company, which would cause similar issues to the bankruptcy of 23andMe. Companies are often bought out specifically for access to data. We wondered what will happen to the data after it’s been “used”, and how accessibility to data and privacy is handled long-term.

Utilising data initially collected under “lifestyle” purposes for medical applications may give reason for additional ethics checks, possibly requiring review from some form of ethics committee. The data that Oura has access to is a very valuable resource, but not without potential for harm. It may be unlikely that a business would be willing to pay for the overhead of this, although commercial enterprises do have to pass ethics applications to run medical trials. Oura would have had to do some form of risk assessment while setting up the pregnancy study. We drew comparisons to pharmaceutical companies, where data is still withheld even after patents have been made public.

Oura rings are expensive (starting cost $299, with monthly costs for data access)! How will this affect data availability across demographics?#

Such high costs drastically shrink the spread of demographics that have access to and are likely to use the rings. Oura targets a certain kind of person who can afford it in the first place but also has the time and interest in managing their health. The study will not have a fair representation of pregnancy if it only includes people that already have a ring, raising issues with selection bias. However, we wondered if selection bias is also a common occurrence with clinical trials, where people who hold medical research in very high regard are more likely to participate or people who do not have access to medicine are more likely to participate to obtain access. It could be the case that some data, no matter how biased, is better than no data at all.

The ring and its findings could be used for the health of all people, and present many opportunities. It is possible to design studies around the available information, but the caveats must be stated and skewness in the sample clearly indicated. Studies with Oura data should be validated with other methods, such as comparing Oura step counters with a pedometer. A normal randomised control trial would involve recruiting participants, and then randomly providing 50% with the ring to see if it improved health outcomes, like the Jean Golding study into providing thermometers for infant’s bedrooms.

We wondered whether people with chronic health conditions wear Oura rings, and reflected on the idea of offering the ring to people who could be less likely to have it (e.g. those who are more financially insecure or have particular conditions) for data collection. There are already entrenched issues with health research for particular demographic groups, and the issues with understanding pregnancy are reflective of women’s health more generally. If the research was to be conducted fairly it would make sense to offer services for free or cheaper, or to compensate participants. Yet, the need for fair representation should be handled carefully alongside considerations of surveillance.

Using technology to bridge gaps in knowledge must be balanced with the remoteness of using an app without physical interventions. It is easy to fall back on technology as a shortcut to understanding marginalised groups, but it is also depersonalising and dehumanising. Everyone deserves access to real healthcare practitioners and it’s important that we don’t just plug gaps in medical research with commercial organisations. Research such as this study could be used as a supplement to research conducted with proper ethics guidelines but should not be a replacement.

If Oura were truly altruistic, we might expect them to do things simply “for the good of research”, such as giving rings away or making data a shared resource, e.g., by donating data to governmental institutions. Organisations like the NHS would probably benefit hugely from access to some of the health data collected by the private sector. Democratising data and healthcare will need to work out how to balance these trade-offs and catch up consent.

Attendees#

Datacentres in the UK