🎉 JOGL is soon launching a new version. All the users of the v1 will be migrated to the new version. In the time being, we do not allow the creation of new users on this platform.
Quantified Flu banner
Project
18
Members

Status:
Active/Ongoing
Linked to group(s)/challenge(s):

Quantified Flu

About reviewed project
Can physiological parameters tracked by our wearables predict when we’re getting sick? We're building a citizen science project for this!




Introduction

The Problem

Quantified Flu is a collective project in which we explore how to use our wearable devices and symptom self-tracking to create individual insight, to try to predict & understand when we're getting sick. While there are increasing numbers of people that have wearables such as Fitbits, Apple Watches etc, there is so far very little known about whether the sensor data from those devices could be used for gaining a better understanding of infections.


While the current interest in this is spurred by COVID-19, the results of this will relate also to the flu and colds in more general: We expect to see physiological changes – e.g. in resting heart rate, blood oxygen saturation, body temperature, etc. – in all of those cases. By doing a collaborative citizen science project around this we hope to harness the collective intelligence of all participants to quickly get a better understanding of our collective data!



Our Solution

We are planning to do a collaborative community science project in which we will try to use both retrospective data and ongoing data collections to understand wearable device sensor data before, during and after infections:


  • For our retrospective data we will annotate our already existing, historical wearable data with the dates when we've fallen sick to see how the physiological signals change
  • For an ongoing data collection we will collect data each day: Are we falling sick right now? Having the flu? A cold? Maybe even COVID19? We'll annotate how we're feeling, which symptoms we have etc.


We are using Open Humans for the data storage and management of the wearable device data, as it already implements GDPR-compliant data access & storage for various wearables (Fitbit, Oura Ring, Withings, Google Fit,…) and can easily be extended to access more wearables. 


For the collection of the symptoms and additional metadata we have designed a small prototype website that accesses the individual wearable data and allows people to annotate their sickness events (when did they happen, what might their symptoms be? What kind of sickness did they have?). For each data set people can opt-in into sharing their own recordings, to enable a growing community of contributors to analyze the data.


State of the Project & Timeline

The project has started sometime beginning/mid of March. It has already launched a fully functioning prototype website that has over 100 volunteers contributing their time and data. Since the initial launch we have been refining the symptom collection forms, started to add further wearable data sources, and refined the data visualizations. Further optimizations are on our roadmap, which you can see below.


The next steps are:


  • Integrating Google Fit data imports into Quantified Flu: 1-2 working days
  • Adding Garmin device data import: ~5 working days
  • Creating a full public data API: ~2.5 working days
  • Adding new data visualizations (e.g. of symptom reports, relative to wearable data, relative to background frequencies of other people's reports, …): ~2.5 working days


Project Implementation

Our initial, community-generated hypothesis was: Can we use wearable devices to predict when we are falling sick, even before we are consciously aware of symptoms? But beyond this our more broad hypothesis is that we can collectively learn a lot about how our physiological signals respond to infections and which symptoms we display when having an infection.


Prior research has shown how influenza infections could be observed through the signals recorded by wearable devices as long as one looks at them on a population-level. Spurred by the COVID-19 pandemic, research teams around the globe are now trying to use such wearable data for predicting infections on a more individual level. Teams working on this can be found at UCSF, UCSD, Stanford, Scripps, the Robert-Koch-Institute (the German CDC equivalent) and in many more places. 


All of these approaches follow a traditional research scheme, meaning: 

  • participants are only asked to provide raw data for researchers to use
  • there is no ongoing communication between participants and academic researchers 
  • The data is collected through closed-source tools which are not accessible to participants
  • Neither the individual data nor the aggregated data are being fed back to the community or individual participants


Based on the large interest of the Quantified Self & Open Humans communities in doing this kind of research together, we are trying to see how useful these data – wearable data & symptom tracking data – can be on a personal level. In particular, we are interested to explore longitudinal symptom self-tracking as a personal data source (with or without wearables – how to do it better, how to discover individual value). While highly relevant to the current COVID-19 pandemic, this has lasting value beyond.


To enable such a community-driven inquiry, we're adopting a strong open source/science spirit from the get go, including the full commitment to making data accessible to individual participants and the larger community.


Methodology

To enable individuals to collect their own wearable device data we will be utilizing the Open Humans ecosystem, which offers support for self-research, tools for personal data access & aggregation, and data analysis. It already provides integrations for wearable data that comes from Fitbit, Oura, Withings and Google Fit devices. As part of this project we will extend the support to Apple Watches, Garmin and potentially other data sources. Contributors to this project can store their aggregated data privately in Open Humans and then consent to this data being used more widely, e.g. in this project. The complete Open Humans ecosystem with all of its data import pipelines is open source and can be found at https://github.com/OpenHumans/open-humans 


For the on-going, manual collection of symptom reports we created the quantifiedflu.org website, which allows people to easily add their daily symptom recordings once per day. Users can set up their own reminder time slots in which they will get a notification email to that end. The website also can pull in the wearable device data to perform retrospective analyses and provide data visualizations for individuals. Last but not least the website gives all contributors the choice to make data public at any point to allow others to reuse this data. Data made public that way can be aggregated and re-used through APIs which are currently in active development. As for the Open Humans platform, all the code for Quantified Flu is open source at https://github.com/openhumans/quantified-flu


Expected Outcomes

Given that we got over 100 contributors in the first few days of the website being live and without dedicated marketing efforts, we expect at least a couple of hundred contributors to participate in the longitudinal collection of symptom reports alongside their wearable sensor data and publish those data in public for maximum reuse. Together we will gain valuable insights into how to best perform individual-centric, longitudinal symptom tracking, as compared to the currently on-going academic-centric projects.


We also expect the contributor community to be the first ones to take full advantage of them, as they have already demonstrated a strong interest in this. Furthermore we do expect this data to be extremely useful for all the academic teams mentioned earlier, who will be able to compare their own cohorts to this one.


Regulations

As this project is not wet-lab based, bio safety/security procedures etc. are not applicable. The main regulation applicable for this work is the EU's GDPR around the storage and use of personal data. By using Open Humans as the backend we can easily comply with the relevant regulations of the GDPR.


Impact, Issues & Risks

We expect this project to have impact in multiple dimensions, it will

  • create usable & open source symptom tracking workflows that can be used self-research
  • lead to a better understanding of the value & utility of personal symptom tracking for self-research
  • empower people to do research on a topic that currently is on everyone's mind and make them feel less helpless


Team Experience

  • Dr. Bastian Greshake Tzovaras is an independent research fellow at the Center for Research & Interdisciplinarity (CRI) in Paris and has worked in the field of participatory research for nearly 10 years. Bastian has a PhD in Bioinformatics and co-founded openSNP.org, a crowdsourced open data repository for personal genomes, in 2011. Since 2017 he is also the Director of Research for the Open Humans Foundation. He serves on the board of the Open Bioinformatics Foundation and is a member of the open science advisory board for Wikimedia Germany. Last but not least, he's also a big quantified self nerd and enjoys visualizing and analyzing his own data. 
  • Dr. Mad Price Ball is a co-founder & Executive Director of the Open Humans Foundation (USA) and has a decade of experience in citizen science and participatory research in health and wellness. They are also skilled at Django/Python, and lead ongoing development of the Open Humans platform. As a PhD student and postdoc in the lab of George Church, they served as Director of Research managing operations of the Personal Genome Project (2010-2016), and also serve on the editorial board of WikiJournal of Science and as a founding Director of MyData Global (nonprofit advocacy for human-centric personal data).
  • Ariadni Karolina Alexiou is a senior software engineer with experience in building robust data collection systems and data visualizations, in particular for time series and geospatial data. She has worked in the industry for 8 years and has a Master’s degree in Information Systems from the Swiss Federal Institute of Technology, and has collaborated with Open Humans as a contractor in the past to integrate data from Google Fit and Github.


Additionally, we have a large group of contributors that are involved with providing their expertise around data analyses, data visualization, software engineering etc. A complete list can be found on https://quantifiedflu.org/about/


Funding

So far our project has been mainly funded through the Open Humans Foundation and the Center for Research & Interdisciplinarity. This funding has included paying for the time of Mad Price Ball & Bastian Greshake Tzovaras to kickstart the project (both have pivoted to work on this project ± full time), as well as all the infrastructure costs for hosting the project (Amazon S3, Heroku, and related cloud services). Both organizations are dedicated to continuing their support.

To become more attractive for volunteer contributors and data-donating participants we will need to do additional development to support more data sources, which currently can not be met by the existing support or volunteer contributions. Consequently we are looking to set aside some paid development time for them.


The development tasks that would be needed this front (and expected costs of doing them) are:

  • Integrating Google Fit to attract all Android Wear users (a significant part of relevant work that has already been carried out can be re-used) (200 EUR)
  • Integrate the APIs of Garmin to support their devices (1000 EUR)


These two will allow us to drastically broaden the scope of our project, by supporting all of the most common wearable devices. In order to enable a larger number of volunteer contributors to join into the development process a few extra steps are needed:


  • Add an API that exposes all public data provided by people, to allow easy data reuse (500 EUR)
  • Add example data visualizations to encourage and enable more contributions (500 EUR)


Ideally we would get the full 2200 EUR to implement all of the tasks above within the next few weeks, at latest Mid-May. Smaller funding amounts (or increased volunteer contributions!) would enable us to get individual bits of this work done as well.

Additional information
  • Short Name: #QuantifiedFlu
  • Created on: March 19, 2020
  • Last update: July 12, 2021
  • Grant information: Received €2,200.00€ from the OpenCOVID19 Grant Round 2 on 04/19/2020
Keywords
Webdesign
Crowdsourcing
Community management
Data science
Time series analysis
+ 2
3Good Health and Well-being
4Quality Education
9Industry, Innovation, and Infrastructure
17Partnership for the Goals