Why #WeAreNotWaiting—Motivations and Self-Reported Outcomes Among Users of Open-source Automated Insulin Delivery Systems: Multinational Survey

Introduction

Background

Despite significant advances in health care, pharmaceuticals, and technological developments, type 1 diabetes remains a challenging chronic condition to manage, impacting life expectancy and diminishing quality of life [1-3]. Only a small proportion of people with type 1 diabetes achieve glycated hemoglobin (HbA_1c) levels below 7.0% (58 mmol/mol), as recommended by therapeutic guidelines to reduce the risk of long-term diabetes-related complications [4-6]. The complexity of diabetes self-management bears a high cognitive load and can cause distress in everyday life, with approximately 40% of people with type 1 diabetes reporting distress and/or depressive symptoms, particularly prevalent among adolescents and young adults [7-10].

In addition to optimizing glucose levels and variability, diabetes technologies have the potential to ease complex decision making and thereby reduce the cognitive and emotional burden of diabetes self-management. The latest advances in diabetes therapy combine sensors for continuous glucose monitoring and insulin pumps with computerized control algorithms, thereby enabling automated adjustments to insulin delivery in response to the user’s changing glucose levels. Automated insulin delivery (AID) systems, also known as artificial pancreas or (hybrid) closed-loop systems, are in various iterations of development and automaticity. Although a variety of commercial AID systems are under development, and some have recently become available in a limited number of countries, they are not universally available, accessible, or affordable.

To fill in the gap, open-source AID systems, also called Do-It-Yourself Artificial Pancreas Systems (DIYAPS), have been created by people with diabetes, in the web-based community behind the hashtag #WeAreNotWaiting, with instructions and codes for these systems available freely and widely via open-source platforms. Although anyone can access this, each user has to take responsibility to build their individual system and use it at their own risk. Initial observational studies have described significant improvements in glycemic outcomes in smaller cohorts of open-source AID users of all age groups, including children and adolescents whose caregivers build and maintain these systems on their behalf [11-15]. Further studies reported improved sleep quality and uninterrupted sleep, in particular, reduced burden of diabetes management, increased confidence in achieving diabetes management goals, increased energy, and reduced mood swings among open-source AID users [15]. An in-silico study of the AndroidAPS algorithm showed similar glycemic improvements and concluded that this algorithm is both safe and effective [16].

Despite the potential benefits of open-source AID systems, little is known about the reasons why people with diabetes initially chose to use this technology. It is important to determine the lessons to be learned from the #WeAreNotWaiting movement, especially for stakeholders involved in research and commercial product development and regulation, such as academia, industry, health care professionals, governance, and regulatory bodies.

Objectives

As part of the OPEN (Outcomes of Patients’ Evidence with Novel, Do-it-Yourself Artificial Pancreas Technology) Project, the aim of this study is to investigate motivational factors for building, using, and maintaining an open-source AID system among adults with diabetes (type 1, 2, and others) and caregivers of children and adolescents with diabetes, as well as their self-reported clinical outcomes, through a population-based survey [17].

Methods

Study Design and Participants

From November 2018 to March 2019, we conducted a web-based, cross-sectional survey titled DIWHY (Multimedia Appendices 1 and 2). The survey design was created by the patient-led OPEN consortium [17], in collaboration with open-source AID users, and piloted by a small number of them before the final release. The Checklist for Reporting Results of Internet E-Surveys was used to guide survey development [18]. The survey was approved by the Charité–Universitätsmedizin Berlin Ethics Committee (EA2/140/18). Participants were eligible if they were adults (aged >18 years), living with diabetes (type 1, 2, or other), or being caregivers of a child or an adolescent with diabetes using an open-source AID system.

Procedures

Participants were invited through public announcements on the OPEN Project website, in the Facebook groups Looped (>6000 members) and AndroidAPS users (>1800 members, November 2018), other regional subgroups on Facebook, and by public posts on Twitter using the hashtags #WeAreNotWaiting and #DIYAPS. All posts were organic, meaning there was no paid promotion or targeted advertising of posts on any platform. All participants gave their consent electronically. Participation was anonymous and voluntary; no financial or other compensation was provided. Participants were able to choose between 2 language options (English and German). There was a version for adults with diabetes and one for caregivers. Data were collected and managed using secure Research Electronic Data Capture electronic data capture tools hosted at Charité [19].

Measures

Initial questions focused on demographics, the type of open-source AID systems used, estimated commencement date, and 3 HbA_1c values each preinitiation and postinitiation of open-source AID (self-reported for adults; for caregivers, their child’s). In addition, participants were asked to provide their or their child’s average time in range (TIR; sensor glucose 70 mg/dL/4.0 mmol/L-180 mg/dL/10.0 mmol/L) before and after the commencement of open-source AID.

Subsequently, participants’ motivation to build an open-source AID was assessed with a single question: “What motivated you to build a Do-It-Yourself Artificial Pancreas system for yourself? Indicate your level of agreement with each statement.” A total of 14 fixed-choice statements followed to conclude the stem “I built a DIYAPS...” (eg, “...to achieve better glycemic control,” “...to improve my own sleep quality”). For each statement, a 5-point Likert-type scale was used (fully applies to does not apply at all). In addition, participants could indicate further motivational factors using free text.

Quantitative Analysis and Statistical Testing

To ensure the reporting of robust parameters regarding HbA_1c levels, entries with more than one missing HbA_1c value either before or after open-source AID implementation were not considered in the calculation of arithmetic means, SDs, and statistical tests related to HbA_1c. The reduction in the average HbA_1c levels before and after open-source AID implementation was assessed using the Wilcoxon signed-rank test (P value threshold of .05, paired: TRUE, and alternative hypothesis: greater). Entries not providing TIR values before and after open-source AID implementation were not considered for the computation of TIR-related descriptive statistics and testing for the increased TIR after open-source AID implementation (same statistical test as for HbA_1c, with alternative hypothesis set to lower). Quantitative analyses were conducted within the R programming framework (v4.0.2; R Core Team), and the ggplot2 package was used to generate figures.

Content Analysis

Content analysis was performed to analyze responses to open-ended questions [20]. A total of 3 researchers coded data and analyzed the responses thematically in 2 rounds. After the first round, which was open, inductive, and independent, 3 lists of codes were merged and combined into a final version. The second round of coding was deductive, and each of the coders assessed the content according to the final list of codes. The interrater reliability (percentage agreement for multiple raters) method was used to calculate the level of agreement between coders, and the final list of the most frequently discussed codes was generated [21]. Codes were then compared with assess the level of similarity, for example, an interrater reliability result of 100% indicated that all codes generated by individual coders matched.

Results

Characteristics of the Study Cohort

A total of 1125 individuals participated in the DIWHY survey. After excluding 25.6% (288/1125) incomplete responses, data from 897 individuals over 35 countries were analyzed. Detailed demographic characteristics are shown in Table 1. Participants were mostly from Europe (691/897, 77%), whereas 14% (125/897) were from North America, and 9% (78/897) were from other continents. Most adults (599/722, 82.9%) and caregivers (153/175, 87.4%) had a university degree or higher. Of the respondents, 26% (236/897) had a professional background in information technology and 19% (170/897) in biomedicine or health care. Furthermore, 82% (736/897) of the participants reported out-of-pocket expenses, with an average of US $530 and a maximum of US $1000 per year. In both groups, various types of open-source AID systems were used regularly, with Loop being the most popular system in North America and AndroidAPS being the most frequently used system in Europe. Otherwise, the geographical location and household income did not indicate any specific patterns.

Motivations to Commence Open-source AID Use

As shown in Figure 1, the most frequently endorsed motivations of adults as well as caregivers (as fully applies or largely applies) were to improve the overall glycemic control (476/509 adults, 93.5%; 95/100 caregivers, 95%), reduce the risk of acute (443/508 adults, 87.2%; 96/100 caregivers, 96%) and long-term complications (421/505 adults, 83.4%; 91/100 caregivers, 91%), put diabetes on auto-pilot mode and interact less frequently with diabetes technology (413/509 adults, 81.1%; 86/100 caregivers, 86%), increase their own or their child’s life expectancy (381/507 adults, 75.1%; 84/100 caregivers, 84%), and improve their own or their child’s sleep quality (364/508, adults 71.7%; 80/100 caregivers; 80%), because of the lack of commercially available closed-loop systems in their countries (359/507 adults, 70.8%; 79/99 caregivers, 80%) and unachieved therapy goals with the therapy options available to them (348/509 adults, 68.4%; 69/100 caregivers, 69%).

‎View this figure

Overall, the motivations of adults and caregivers of children and adolescents with diabetes were largely similar. As the most noticeable difference between the 2 groups, improvement in their own sleep quality (94/100, 94%) was a stronger motivation for caregivers compared with adults with respect to their partners or families (225/505, 44.6%). Curiosity (medical or technical interest) was endorsed more frequently by adults with diabetes (367/503, 73.0%) than by caregivers (45/97, 47%). Some believed that commercial systems did not suit their own or their child’s individual needs, more frequently reported by adults (316/498, 63.5%) than by caregivers (46/98, 47%). Out-of-pocket costs related to the use of commercially available systems (166/496 adults, 33.5%; 31/99 caregivers, 31%) played a subordinate role. Lack of adequate medical support (105/501 adults, 21.0%; 29/98 caregivers, 30%) or psychosocial support (94/501 adults, 19.0%; 22/99 caregivers, 22%) were less frequently endorsed as motivating factors, although caregivers more frequently indicated a lack of medical support.

Further Motivations

In addition to the 14 predefined items, participants could indicate further motivation in an open-text field. In total, 127 participants (103 adults and 24 caregivers) provided a free-text response. Textbox 1 provides a list, as well as representative quotes, of the respondents. In the independent coders’ selection of first-choice codes, there was an 83% interrater agreement between them (Multimedia Appendix 3).

Most of the indicated other motivations provided greater details about the 14 predefined statements. The most frequently mentioned motivations for all—adults and caregivers—were better management and reducing the disease burden. The first motivation appears consistent with several statements related to hypoglycemia and hyperglycemia and risk reduction, whereas the second motivation may correspond with to put diabetes management more on auto-pilot and interact less frequently with the therapy system. This aspect and sleep quality are understood as the quality of life gains. Of motivations not covered by the predefined responses, the most frequently mentioned was autonomy gain in both adults and children or adolescents, as indicated by the caregivers. All these aspects were associated with improvements in family life:

This is for my wife. She wants me to live forever, this is as close as I can do for her.

Psychosocial aspects, ranging from diabetes burnout and distress to a desire to improve athletic performance to increasing efficacy at work, were also identified as important motivating factors. The following comments illustrate the wide-ranging benefits experienced by many participants after adopting the technology:

Management of diabetes is helped by support, but it is very much a self-managed disease and requires 24/7 attention. Closed looping makes it just so much better, much of the time I can leave [AndroidAPS] to take care of basals by itself. Quality of life is so much better. I can sleep without worrying about not waking up because of a bad hypo. [...] I am so grateful to all the software developers who have freely given their expertise and time to make this possible.

As highlighted in the example above, an important role was played by the community spirit and peer support in social networks:

Something that also influenced me [...] was the support from the community, and the general feeling that the community gives. It feels like I am part of a big people-powered movement. It feels like a revolution.

Not only a Do-It-Yourself mindset and being early adopters of technology but also being motivated and empowered to improve one’s life were frequently mentioned:

Because it’s the most natural thing to do, after getting to know that it’s possible. Because I could.

Some motivations included other health-related aspects such as improving the management of existing diabetes-related complications and increasing safety by avoiding severe hypoglycemia. Other comorbidities, such as cancer, sexual health difficulties, or conditions requiring cortisone treatment, were also mentioned:

I have to take cortisone in different doses on a regular basis. This has made my diabetes management so difficult. The loop absorbs my BG fluctuations much better.

Women and caregivers of female children highlighted female health aspects such as hormone-related changes in insulin sensitivity, family planning, and pregnancy as reasons to commence open-source AID:

To have more insight as to why my BG was so volatile due to changing hormones (menopause).

For some, special features were only offered by open-source AID and not by commercial systems, that is, customizable targets and the option to bolus from a smartwatch. For caregivers, remote real-time access to their child’s data and the option to remotely control their child’s AID system have been frequently described.

Improved Glycemic Outcomes Across All Age Groups and Genders

To assess glycemic outcomes, participants were asked to report their or their child’s 3 most recent HbA_1c results before as well as the first 3 HbA_1c results after commencing the open-source AID. HbA_1c levels decreased significantly following open-source AID implementation (P<.001) from an average of 7.14% (SD 1.13%; mean 54.5 mmol/mol, SD 12.4) to 6.24% (SD 0.64%; mean 44.7 mmol/mol, SD 7.0), with an effect size of −0.9% (Figure 2).

‎View this figure

The average self-reported TIR across adults and children with diabetes significantly increased by +17.4%, from 62.96% (SD 16.18%) to 80.34% (SD 9.41%; P<.001; Figure 3). Similar outcomes were observed separately for adults and children with diabetes (Multimedia Appendix 3) and were independent of age and sex (Figure 4). Overall, 92.3% (286/310) of the respondents reported a decreased average HbA_1c level (Figure 5).

‎View this figure

‎View this figure

‎View this figure

Discussion

Principal Findings

This study is the first to systematically analyze the motivations found within the #WeAreNotWaiting movement of people with diabetes, who have built and maintained their open-source AID systems and created their own ecosystem of international self-support networks. To the best of the authors’ knowledge, this is also the largest study reporting the self-reported clinical outcomes of open-source AID users across several continents. We found large effect sizes for self-reported improvements in HbA_1c (−0.9% on average) and TIR (+17.4% on average), indicating considerable biomedical benefits associated with open-source AID, which were independent of sex and age.

Why #WeAreNotWaiting: Main Motivators to Choose an Open-source AID

The main motivators for adults were improvements in overall glycemic and long-term outcomes and quality of life, whereas the strongest motivation for caregivers was improvement of their own sleep, followed by improved glycemia of the child and possibility of remotely controlling glycemia and insulin delivery via the internet. The results indicate that motivations are configured differently among caregivers and that other motivations also scored a high level of consensus among the respondents. These findings suggest that motivation to transition toward open-source AID is multifaceted and complex, with reasoning and decision making bound up with the psychological and social intricacies of individuals’ lives.

Improvement in Sleep Quality

Caregivers experience reduced sleep quality because of fear of hypoglycemia, which often requires them to regularly check their child’s glucose levels overnight [22]. In our study, caregivers reported experiencing fewer demands and less apprehensiveness regarding their child’s glucose levels at nighttime. As shown in the free-text responses, open-source AID also appears to offer caregivers with the reassurance necessary to provide their child more autonomy and engage in activities that might otherwise present a risk, such as having a sleepover with friends. Previous studies in adults using open-source AID have shown self-reported improvement in sleep quality [23,24]. These initial findings indicate a substantial benefit for users and caregivers for sleep and most likely for their psychological and physical well-being. Poor quality of sleep negatively affects the psychological well-being, cognitive functioning, and a diverse range of hormones that affect the regulation of appetite and our homeostatic systems as well as the immune system [25-34]. Recent research also points to sleep as impacting the actual maintenance of the brain and our DNA regenerative systems [35]. Thus, AID may play an important role in improving the psychological and physical health of people with diabetes and their family members. However, it has been noted elsewhere that the potential discomfort or inconvenience of wearing devices and overnight alarms may also hamper the benefits for some users [36].

The Importance of Customizability: One Size Does Not Fit All

The majority of participants reported that currently approved and available commercial therapy options may not be sufficiently flexible or customizable to fulfill their or their children’s individual needs. Among caregivers, features only available in open-source AID, in particular, the possibility of remote management was the main additional motivation. A wider range of features and adjustable settings to improve user experience may be beneficial for people with diabetes of all ages, which mirrors a recent study in very young children using a commercial AID [36]. Interestingly, for many adult respondents, curiosity was cited as an important motivation. In contrast, curiosity was a much lower motivating factor for caregivers who chose to build a system for more practical or psychosocial reasons.

Do-it-Yourself Is Not Do-it-Alone: The Impact of Peer Support

The ability to receive and provide support within the do-it-yourself (DIY) community and observe the success of others was an important motivating factor associated with opting to use open-source AID for some. Obtaining and exchanging information and advice from open forums limits the spread of misinformation because other users constitute a community of inquirers ready to challenge and correct spurious or misleading information [37]. Although open-source AID is individualized and patient focused, it is also a grassroots community-driven movement. The number of responses to our survey reflects the enthusiasm and importance of open-source AID. In challenging traditional top-down hierarchies in medicine, open-source AID presents a patient-focused initiative that serves to empower people with diabetes through personalized technology. Because of the availability of current technology and individualized innovations, open-source AID has previously been described as having the potential to democratize health care, revolutionizing treatment and the way people with diabetes as well as other stakeholders such as care teams, researchers, and device manufacturers view chronic conditions such as diabetes [38]. The importance of peer support in the context of open-source AID use has recently been highlighted elsewhere, and a sense of community underpinning the development and diffusion of open-source AID has been emphasized by individual users [39]. Further research should consider community phenomena as an integral part of the DIY experience.

Strengths and Limitations

This study is the first to investigate motivations of users and caregivers to build and use open-source AID. In addition, this is the largest study that reports self-reported clinical outcomes of open-source AID users globally and adds to the existing evidence base around glycemic outcomes in smaller cohorts [11-15]. At the time of data collection, it surveyed the majority of open-source AID users worldwide, with 897 respondents of a population estimated in 2018 (N=1500). The sample is impressive not only in size but also because people with diabetes from various continents and regions of the world are represented. Of other strengths, this study has been conducted by an interdisciplinary consortium, with members of the diabetes community directly involved. However, this firsthand experience should be acknowledged as a potential bias. In addition, a key limitation of the study is the fact that self-reported outcomes have not been corroborated by clinical records. Some may consider this has potential for inaccuracy, that is, by lacking precision as witnessed by the overaccumulation of rounded TIR values or to be biased by the specificity of the population that participated. However, other studies have found that real-world data are robust and reliable [40]. We acknowledge that open-source AID users constitute a specific group of people with diabetes who may be highly motivated, engaged, and willing to improve the quality of their diabetes care and life, limiting the scope of our findings to this group. However, a recent study of newly available commercial AID systems indicates that users are similarly motivated to achieve the best possible outcomes but are dissatisfied with postprandial glycemic outcomes under commercial AID systems [41]. The earliest adopters of available commercial AID technology may be more similar to the group of people with diabetes choosing open-source AID in terms of motivation and engagement than expected. The lack of widespread availability of AID technology in general, including commercial systems with regulatory approval, at the time of the study likely influenced the perspectives of people with diabetes choosing open-source systems. In the future, wider availability—and, importantly, better funding or insurance coverage of commercial AID systems—may further influence this cohort. Similarly, it remains to be seen if the predicted second generation of commercial systems—with a hypothesis of increased sophistication or improvements on the first-generation devices and algorithms—will enable people with diabetes to achieve results similar to those they are currently achieving with their chosen open-source AID system.

It should also be noted that those who benefit from and continue to use open-source AID may be motivated to share their positive experiences. Although the survey was open to people with type 2 and gestational diabetes, it was completed almost exclusively by adults and caregivers of children with type 1 diabetes. This is likely a reasonable reflection of the DIY community, but efforts need to be made in the future to encourage participation of those with other types of diabetes. The high percentage of respondents from Europe may be influenced by the fact that the majority of the research team is EU based, which may be another bias. This may also be explained by the characteristics of the European health services provision and reimbursement of diabetes-related technology, which may provide a greater degree of accessibility of the underlying components needed (eg, pumps and continuous glucose monitors). Language barriers may have limited responses from other parts of the world as the survey was only available in 2 languages. Finally, the majority of participants had a university degree, suggesting that open-source AID uptake is more common among people of higher socioeconomic status. Increasing socioeconomic inequalities in access to the underlying technologies needed to build an open-source AID system may help to explain some of these variations. Thus, further investigation into how the wider diffusion of open-source AID is conditioned by factors such as social class, gender, age, and geographic location is required.

Conclusions

This study provides new insights into the factors that motivate people to adopt DIY solutions in relation to their diabetes and beyond. Our findings contribute to a better understanding of the unmet needs of people with diabetes and some of the current challenges in the uptake of AID technology. This study, alongside other efforts in the DIY community space, can help key stakeholders, including academia, the medical device industry, regulators, health care providers, and care teams, to better address some of the fundamental gaps and needs that still exist for people with diabetes worldwide, even with the advent of first-generation commercial AID systems. The DIY movement has resulted in impactful solutions addressing the unmet needs of people with diabetes and represents an exemplary case of how informed and connected patients are shaping the direction of technological innovation in diabetes care and potentially for other areas of health care in the future.