The OPEN Project

The OPEN Project

Outcomes of Patient’s Evidence with Novel,
Do-it-Yourself Artificial Pancreas Technology

Project logbook - Day 1532: Mission complete.

In 2019, the patient-led research project set out to gather real-world evidence on the clinical and quality of life outcomes of open-source automated insulin delivery (AID) systems to better understand their impact on the lives of people with diabetes.

Over 40 researchers joined us on this mission with a consortium consisting of 8 international members, including University College Dublin, Charité - Universitätsmedizin Berlin, #dedoc° Diabetes Online Community, The Australian Centre for Behavioural Research in Diabetes, University of Copenhagen, Steno Diabetes Center Copenhagen, Stanford University, Diabetes Center Berne and King's College London.

We are grateful to everyone who contributed to the success of the OPEN project and our Closing Conference back in February 2023.

Huge congratulations to the OPEN team, who has been working tirelessly to make this project a reality.

Watch the Closing Conference

Watch the entire recording of the conference with it an in-depth summary of our work here.

This project has received funding from the European Commission's Horizon 2020 Research and Innovation Programme under the Marie Sklodowska–Curie Action Research and Innovation Staff Exchange (RISE) grant agreement number 823902.

About DIYAPS

Digital innovations in healthcare up until recently have typically followed a ‘top-down’ pathway, with device manufacturers leading the design and production of technology-enabled solutions, and patients involved only as users of the end product. However, this is now being disrupted by the increasing influence and popularity of more ‘bottom-up’ and patient-led open-source initiatives. A leading example is the growing movement of people with diabetes who create their own “Do-it-yourself Artificial Pancreas Systems” (‘DIYAPS’) through remote control of medical devices with an open-source algorithm. The DIYAPS community has created a safety-focused reference algorithm and an implementation of a hybrid closed-loop system. Based on the prediction of glucose levels, the system constantly and automatically adjusts basal insulin levels, in order to keep glucose levels inside a safe range, night and day.

DIYAP systems are not CE-marked or FDA-approved systems or devices. They are not being manufactured or sold anywhere in the world. However, the algorithm and source code are made open-source and are publicly available.

Further information is available here:
OpenAPS: www.openaps.org
AndroidAPS: www.androidaps.org
Loop: www.loopdocs.org

4 Years of OPEN: A Recap

Click image to zoom in

Work Package 1
Clinical Outcomes and Professional Guidance

Work Package 1 explored physical health-related changes that open-source AID users experience, such as the changes in hemoglobin A1c (HbA1c) and time of glucose spent “in range”, and on practical guidance for healthcare professionals who want to support people with diabetes using open-source AID.

In the DIWHY survey, we found that after people started using open-source AID, HbA1c levels were significantly reduced, and time in range was increased. These changes were found in people of different genders and ages, ranging from 2 to 77 years, and users of all types of open-source AID systems (OpenAPS, AndroidAPS, and Loop).

In an international consensus statement, the OPEN team worked with a group of 44 medical and legal experts on the topic of open-source AID from around the world. In the statement, we provided an overview of the state-of-the-art research on open-source AID, a description of the different technologies, discussed safety aspects, but also availability and access issues of diabetes technology, and the ethical and legal considerations for these systems from an international perspective. Furthermore, we provided much-needed guidance for healthcare professionals that will enable them to support people with diabetes who choose open-source AID for their treatment. Lastly, we provided recommendations for key stakeholders that are involved in diabetes technologies, including developers, regulators, and industry, with respect to interoperability, transparency, and data ownership.

The consensus was published by Lancet Diabetes & Endocrinology as part of the “100 Years of Insulin” special issue on World Diabetes Day 2021. The full-text version of the article is available here upon free registration or on PubMed Central).

Click image to zoom in

Work Package 2
Person-Reported Outcomes & Lived Experience

With the OPEN survey, the project has collected the largest and most exhaustive dataset of Quality of Life (QoL) outcomes for people using open-source AID systems (see Knoll et al and Schipp et al). Open-source AID users indicate high quality of life scores measured in the survey, particularly in relation to well-being, treatment satisfaction, and diabetes distress.

The qualitative analysis of data acquired in WP2 in the form of photo-elicitation and autoethnography (“a diabetes diary”) provides deep insights into the impact that open-source AID has on the user. It illustrates the transformative potential of open-source AID while highlighting the many daily challenges and successes that are part of life with open-source AID. It also spotlights the motivations for people using open-source AID, and the resources they have drawn on to do so.

Aside from underlining the massive transformative potential of open-source AID for the individual user, the data also point to the importance, both practically and emotionally, of the #WeAreNotWaiting community. Many open-source AID users say that becoming part of that community has had a profound impact on their lives with diabetes and beyond. In many ways, it is as important as the open-source AID technology itself.

Finally, our findings point to the challenges that people using open-source AID may confront in their interactions with healthcare providers and services. Although many users report positive experiences and how using open-source AID has changed the dynamic of their interactions with HCPs in a constructive way, there are still more who highlight the scepticism and lack of support by healthcare stakeholders as negative experiences resulting from their opting to use open-source AID.

Although we could not find any significant differences in average HbA1c and time in range between people of different genders, the many testimonials of users highlighting the menstrual cycle and other female health-related implications led us to further look into the topic. In a series of interviews, 12 users of open-source AID shared their observations, thoughts, and strategies with us on how they manage diabetes throughout the menstrual cycle and menopause. Many of them stated that the differences in insulin needs only become obvious to them after they started using AID, and that professional guidance both from diabetes and OBGYN specialists is limited. We highly recommend future research to address these topics, in data analytics as well as in user experience research. The full-text version of this article is available here.

Click image to zoom in

Work Package 3
Technical Aspects and Improvements

Work Package 3 made extensive use of two of the now-largest diabetes datasets in the world: the OpenAPS Data Commons and the data contributed to the OPEN project, which could be paired with the survey data. We sought to make it possible to pair survey data with donated device data via a pseudonymous user ID, which is something that is not commonly done, and created an open-source research tool called the “Gateway” (available open-source here for other researchers to use). More details about the Gateway are in this paper.

We also developed methods for de-duplicating data from participants who contributed to the two datasets anonymously, to avoid replicating analyses on individuals who had already been analyzed as part of previous work. This enables us and other researchers to leverage diabetes data from multiple datasets. This method can also be used to enable researchers to use diabetes data from clinical research and real-world datasets. We are excited to see these methods used in the future to expand the amount of open data that is used for research, particularly for training machine learning and deep learning algorithms to further improve AID algorithms.

Ultimately, the goal of our work was to address how machine learning might be used to learn from and then propose improvements to open-source AID. We used several machine learning and deep learning models (methods and results available here) to extend glucose prediction or forecast horizons up to 48 hours in the future, whilst maintaining accuracy and evaluating the ‘cost’ of computing power to do so. In the future, we are excited to see others build on this work to evaluate new prediction models alongside computing costs to better evaluate the trade-offs of using these models in real-world systems; possibly test and apply these additional forecast methods to open-source AID; and continue to leverage the open source code for glycemic variability analysis to continue to improve “edge cases” like hypoglycemia reduction during and after exercise, improving individual mealtime outcomes, reduce the work involved in everyday diabetes, and more.

Click image to zoom in

Work Package 4
Barriers to Uptake

The main aim of WP4 was to examine the barriers to the wider diffusion of open-source AID technology. With this in mind, we conducted a survey of members of the DIY community who were not (yet) users to examine what they perceived as the most common barriers to building a system. The main barriers identified towards uptake included: sourcing the necessary components, such as loopable insulin pumps or continuous glucose monitoring sensors, lack of confidence in one's own technology-related knowledge and skills, perceived time and energy required to build a system, and fear of losing healthcare. Some of these barriers may be overcome over time through the peer support of the community as well as greater acceptance of open-source innovation among healthcare professionals. However, other barriers, such as sourcing components, are more structural in nature and reflect long-standing inequalities in access to diabetes technology.

These findings also echoed in another study conducted by WP4 that examined the costs associated with diabetes supplies across different counties. For this, we were fortunate to collaborate with the patient-led NGO ‘T1International’ on their global Out-of-Pocket Cost Survey. Worldwide, one out of every four respondents reported having underused their insulin at least once within the last year due to high cost. In Ghana, for example, the majority reported rationing test strips every week.

Meet the people behind OPEN

Shane O'Donnell

– Project Coordinator, UCD –
Postdoctoral Research Fellow and Social Scientist,
University College Dublin
Hanne Ballhausen

– Project Manager, #dedoc° –
BSc International Business,
MSc Management and Corporate Sustainability
Bastian Hauck

- Industry Partner -
CEO, Dedoc Labs GmbH
Founder, #dedoc°
Tahar Kechadi

Professor of Computer Science,
University College Dublin
Katarina Braune

Medical Doctor and Researcher in Pediatric Endocrinology,
Charité – Universitätsmedizin Berlin
Christine Knoll

Doctoral Candidate in Medicine,
Charité – Universitätsmedizin Berlin
Mandy Wäldchen

BSc Biology,
MSc Biomedicine
Bryan Cleal

Senior Researcher and Social Anthropologist,
Steno Diabetes Center
Henriette Langstrup

Associate Professor of Science and Technology Studies,
Center for Medical and Technology Studies,
Department of Public Health,
University of Copenhagen
Timothy Skinner

Professor in Health Psychology,
University of Copenhagen
Jonathan Garfinkel

Ph.D. Candidate in Medical Humanities, Poet, Journalist, Novelist, Playwright
Jasmine Schipp

BSc Psychology,
Ph.D. Candidate and Research Assistant,
Australian Centre for Behavioural Research in Diabetes
Yanbing Chen

PhD Occupational Health and Sport Management,
Postdoc Workplace COVID-19 Prevention, School of Public Health,
University College Dublin
Drew Cooper

BSc Biology,
MSc Medical Neuroscience,
Data Scientist
Dana Lewis

Founder,
OpenAPS
Adrian Tappe

Developer,
AndroidAPS
Tebbe Ubben

Developer,
AndroidAPS
Greta Ehlers

MSc Strategic Communication
Digital Channel Manager at DCB
Saskia Wolf

Volljuristin*,
Founder of #LooperInBerlin
* with Second Legal State Exam (Germany)

Consortium

Frequently Asked Questions (FAQ)

Thank you to everyone who contributed data to the OPEN project. While the OPEN survey is now officially closed, this does not mean you do not have to withdraw your consent or delete your data from the OPEN Humans platform. In fact, we would be thrilled to continue having it available for future research on open-source AID, and it can help facilitate any questions you might have in this regard. Your identity will remain anonymous and all data sets remain safely stored, and will only be shared with researchers upon request if their goals and ethics align with the spirit of the #WeAreNotWaiting community. Open Humans is a not-for-profit organization with a policy that, firstly, all findings and outcomes generated from donated data are shared open-source and thus help benefit everyone in the community, and secondly, participants will be notified and supported in case of “incidental findings” (anything unusual that might occur in your data).
For people who have not shared their data on Open Humans for the purpose of open-source AID research yet but wish to do so, you can join the OpenAPS Data Commons on Open Humans, an ongoing collection of device data that is being used widely in diabetes research–by the OPEN team and others.
If you would like to work with any of the OPEN data, please email us at hello@open-diabetes.eu. We would be happy to help accommodate any of your requests and potentially collaborate with you in the future.