 Hey everyone, glad to be here at CSV Conf again. So I want to talk today about how during the first year of the pandemic we implemented a way for people to use their variables to detect potentially if they have an infection and more importantly how we build something together to make this happen. I am the Director of Research for the Open Humans Foundation which helps people using and gathering their own data to learn from it and also a research fellow at the Center of Research and Digital Diplinarity here in Paris and let's start by the things that wearables can do. So there's a lot of them and they have increasingly been more clever in what they can actually track. Of course they can track your resting heart rate. Many of them now do your body temperature as well, your respiratory rate, your sleep duration and quality and interestingly coming out during the last year particularly a lot of them measure your oxygen saturation as well. And that's already a lot of physiological data that you can collect about yourselves. So even before the pandemic hit last year there was interest in using this for actually monitoring infections and here's a paper that was presented at the conference I think in 2019 that used wearables on a population level to surveil the influenza outbreaks and it was based on data from 2017-2018 flu seasons using I think around 3,000 participants and the metrics tracked through wearables were things like the sleep amount of step state and sleep quality. And for some of those variables as like shown here for the total number of steps walked people already behaved or started to behave differently one to two days before symptoms had any onset noticeably by the individuals. But the limitation of course here is that all of this is done on a population-wide surveillance level not for the individual use. So on average actually it turns out that you might see a difference before so you can see when the numbers rise up but it's not for making individual predictions. But of course population-wide monitoring is pretty cool and interesting. And then of course last year hit and particularly in March 2020 where all of a sudden everyone wanted to do similar things but for COVID. So at the UCSF and UCSD they collaborated with the people who make the URA ring wearable device there and also at the Scripps Research Institute in San Diego they started their own wearable data collection and also the healthcare innovation lab at Stanford did the same and last but not least even the German equivalent to the CDC started its own COVID data donation app to address the same same thing. So all of these came out and they all had pretty much the same approach as the earlier influencer approach I showed you it was about getting population level data to have like a population-wide monitoring thing. And there are some other issues as well because if you have academic research combined with wearables then you can easily end up having like some science by press release and here's just some small quote from one publication or not even publication the press release that came out saying that oh we can actually have a 90% accuracy in anticipating symptoms up to three days in advance. But they never actually published it and by now even the press release has disappeared from the university's website so oops maybe not really. But besides this there's other issues so there's no feedback for the participants there's no data sharing so no one can actually reuse it and there's definitely no idea of supporting individuals in learning and making sense of their own data which is what we are particularly interested in. So how can we use wearables for personal science to do this together? But first what's personal science? Personal science describes people using empirical methods to answer personal questions and here are some examples like if I'm a diabetic is this fiber really indigestible or will it raise my blood sugar? What is triggering my arrhythmia or does my transitional hormone therapy influence my mood? So all of these are questions people ask and actually last year just before the pandemic really started getting going with open humans and the quantified itself we started a project called the Keating Memorial to support people doing personal science collectively together and this was as the name implies in honor of Stephen Keating. Stephen who was a board member of the Open Humans Foundation passed away a year earlier because of a brain tumor in 2019 and since he was first diagnosed with it he became a very keen personal scientist and also data donation advocate. He collected a ton of data about himself, his own genome, his cancer genome and over 10 hour recording of his awake brain surgery which he shared and he believed that data could be useful and should be shared if possible and he reminded us that collecting this data doesn't need to be selfish but that it can be used for a greater good if we share data with others and use it collectively. So we just started doing this memorial which included doing weekly calls where people try to answer their own personal questions and the three examples I gave on the last slide actually were projects people worked on. But then of course the pandemic hit and people lost all time and motivation to work on these topics and instead had actually one larger global question which is we all have these wearables and can we use them to see when we are having an infection and to make sense of the data and maybe even use this data if we figured out how our personal body responds to having an infection on a physiological level to use it going forward. So collectively we started brainstorming during the calls and continued and select conversations to see how we could do this and as the first prototype we made for quantified flu we made one for doing retrospective investigations of physiological signals. So thanks to the infrastructure we already had in place with open humans we already had support for two wearables for this Fitbits and also the Aurora rings. So we could very easily make a retrospective analysis that people could make themselves so people would just punch in when they remembered when they fell sick for the last time and they would get such a nice graph is on the right side where you can just see in this case how the heart rate evolves over time around the the line of where you actually noticed falling sick and this is an example of my own data because I remembered falling sick of all times on New Year's Eve 2018 and yeah you can indeed see my heart rate already starts going up just a few days before I actually came down with something. So we made a couple of those and actually started doing like a very small scale data analysis on the end of six where we collectively wanted to see if we can also see the same signal overall and it turns out for the temperature and also the heart rate that there seems to be at least some kind of signal which is really nice but of course that's not super ideal because first of all you need to remember when you fell sick in the past if you want to make sense of this and that's really hard you can go back maybe to your calendar or to your chat messages and try to find out when you told someone you were falling sick but it's not really easy to do and of course you might want to track more specific symptoms instead of just knowing did I fall sick on a given date here so no and also continuous monitoring in general would be just much more useful so collectively our community decided to learn from the best and steal so we went to the academic efforts that were launched at the same time and we're like so what kind of symptoms are these studies actually trying to track and we started just remixing their symptom reports to unify them into one that seemed most useful to our participants so which symptoms and on which scale should they be tracked instead of just having binary symptom reports we settled for all of these symptoms on a five scale reporting and people would get daily email check-ins to say hey are you feeling sick today yes or no and if no you're basically done you just click a link that says i'm feeling okay and if you are sick you are getting taken to this form where you can just move the sliders around and you collect this data which is nice it's easy to collect but then how do you make sense of the data or how do you use it so again collectively some brainstorming of how could this data be visualized and we came up with this graph which converts the data that you provide both in terms of symptom tracking and your physiological data into one graph and the heat map on the top you see when you had symptoms and how severe they were over time in the middle in green you have like this commenting section because people can even write comments and below you can see your heart rate body temperature respiratory rate whichever thinks your variable tracks that you can actually use for this so that's really nice and like you can see it easily and can see when you are having like an elevated heart rate because the different gray backgrounds show the first and second standard deviation of the normal medium value of your heart rate so you can see if you have an elevated heart rate right now compared to what's normal for you personally instead of having to rely on some population approach and i think what's most interesting for this particular view actually is that there's a lot of value and you collecting data about yourself which academic studies would never actually collect and the comment here is that the person reports that oh my cuff actually it's not related i report having cuffs symptoms but it's not related to having an infection because well i've been smoking more since i'm now out of work and like a couple of comments earlier the person says that they were just they just lost their job because of the pandemic so okay there is now a pandemic related cuff but actually it's just from smoking at home because they're going to work any longer so this data gets collected by people they can use it for make sense out of it by themselves and also people can share the data so people can opt in and say i want to make this data available for all the other members of the community and basically everyone so if you want you can go to quantified flu.org and get all the data both in jason and of course as we are csv conf you can also download csvs of all the data which are the physiological signals and symptoms people report so that's the technical implementation on like the very rough level but i think what's really important to us is that this was not done by me or any of the other individuals who might somehow do this professionally but really by the community for the community all of this development was done by the community members of open humans and the quantified self and even just random people who ran into our slack and said i want to help so this included the initial ideation and the protocol design iterating the protocol over time implementing of the tool and different things like new wearable support and also testing the implementation of course and as we can see what we did is we just coded every single slack message into different categories whether it's about prototyping coming up with general ideas or even like the implementation of the prototypes and we can see that there's a very nice iterative development going on so after an initial release very early on there's a lot of renewed interest in discussing the protocol and prototyping more which leads to a new version of the prototype and so on and that's really nice because it's like many of the things i showed you were not completely just done by any individual person but collectively so the symptom tracking or which symptoms to track i said was done collectively the visualization was brainstormed collectively and then one one master student actually implemented it and we even had to completely random mobile developer who made an ios app to get hardware data out of apple watches because we had no mobile developer actually in the community so someone found us and was like oh yeah i'm willing to donate a bit of time to actually help you do it so besides this all being really cool what's the benefit of doing co-creation and what we see is that co-creation really improves the user fit if you think about how mobile health applications typically work and how much they are being used one can see in the academic literature that there is a really limited sustained use and the most extreme case is less than two percent of users ever return after using your app once and never provide data or get any use out of it and in contrast what we see looking at our usage numbers is that we have over 50 percent of our users have collected data for more than three months and actually to this day data is still being collected by our community and people are contributing so there's a big big benefit in doing it like this of course if you are an academic this iterative design is really problematic from an ethical oversight point of view because typically if you want to get institutional review board approval for your study then you need to completely tell them what the protocol is so there's no way of iterating rapidly and doing things differently and I think there's like a very interesting case to be made why actually this is a big issue if you would have done it in a way of getting IRB approval for every single point along the way we still wouldn't even have the first prototype most likely instead we had five maybe six different prototypes which evolved over time in a span of maybe three months which is fine if you are doing it outside academia if you are into academic publishing that's unfortunately a really big limitation and I think we need to find new ways for actually doing things together so to sum up what we learned to give a bit of time back is that first of all quantified flu it's not epidemiology but it's about generating personal insights and personal learning and making sense of your own data and if you want to try it you can go to quantified flu.org right now connect your wearable device start recording your data and in a few days you can see how your heart rate and so on evolve over time and see if there's any correlation and I think importantly we see that this co-creation approach and this community involvement leads to a really high fit to the user needs which leads in turn to a great continued user engagement and of course the data you can opt into making it publicly available and there's I think 50-60 data sets publicly available which you can play around with if you want and as I speak so much of the community we had a lot of people involved so besides me, Matt, Gary, Kate and many others were involved in the initial brainstorming and thinking about the project and helped us a lot in getting all the slack messages coded to actually understand how we co-created. Basil was instrumental in making the heatmap. Lukasz from Poland was our iOS developer. Carolina made the Garmin variables and Google Fit integrations. Konstantin helped getting the Fitbit support up and running for intraday data and Ilona and Melvin supported us in making all the visualizations for the collective data and of course there's over 200 users now who actually tried it out and please give it a try and before I wrap it up I just want to take 30 seconds to pitch some similar idea where we could need your support where we are currently crowdfunding. Clara, a student in our lab, is launching a really cool transgender health related project called the Transbiome which is about the idea that of course the cis woman's vaginal flora is quite well understood. There's a ton of studies on this. We know a lot about how it's involved in infections and the risk for cancer and all of this but in comparison I think there is less than a handful of studies for the microbial flora of the neovagina of trans women which is highly problematic because it means they cannot get gynecological care. So what Clara has designed is a community driven study which actually was designed by her as a trans woman with support of the trans community here in France in particular to collect data on the microbial diversity and to understand even what's the baseline picture because even that's unknown right now. And if you can help us out with just a few dollars to push us over the goal that would be much appreciated. So thanks so much and happy to take any questions. Thank you so much for the presentation. I think we have a few moments for a question or two if there's any in the chat and if there's not I would invite people. Oh looks like we have a question. So can we differentiate between co-creating and the pandemic pandemic conditions the participation rate sorry. Right yeah so with regard to the participation rates it's hard to say right now but we do know that there are all the academic studies which are going on right now which don't have the benefit of being co-created because they were done completely in an academic framework and the academic researchers set up the design and ran it. So once their data is out or at least the papers are out we can see how much participants they enrolled initially and for how many they got enough data in comparison and it's both pandemic related so we can hopefully make a comparison of that in the future. And I think that unfortunately that's all the time we have for questions but I invite everyone to continue the conversation in the Slack channel CSV6 Q&A. Thank you. Thanks.