Conflicts of Control: Continuous Blood Glucose Monitoring and Coordinated Caregiving for Teenagers with Type 1 Diabetes

Continuous blood glucose monitoring (CGM) is a bio-sensing technology designed to help individuals with diabetes understand and manage their blood sugar levels. Better control and management have the promise of extending and improving the quality of life with diabetes. For teenagers with Type 1 Diabetes (T1D), CGM has the potential to allow for a more collaborative T1D management and control experience by sharing real-time blood glucose data with caregivers. To understand how well CGM designs are living up to this promise and to identify potential challenges, we used the Asynchronous Remote Communities (ARC) with 16 teenagers with T1D and their informal caregivers over a 6 month period to investigate families' lived experiences with CGM and its data sharing function. We found three challenges in caregiver-teenager collaboration; lack of visibility into behaviors behind data, difficulties balancing teenage daily life with T1D control, and technical limitations of current CGM systems. We also found three issues that hinder informal caregiving coordination; unclear task assignment, difficulties in maintaining standards of care, and a lack of informal caregiver rotation. We provide a number of design implications based on these findings to better facilitate home care for teenagers' T1D.


INTRODUCTION
Marina is a typical 16-year-old teen who happens to be living with Type 1 Diabetes (T1D).This weekend she was finally allowed to join a sleepover at her friend Gina's home.Marina wasn't sure about snacks that Gina's mom prepared but decided to give them a try.After all, Marina just wanted to be like any other teenage friends-eat some junk food and have a fun movie night.She did a careful carb count and gave herself an early bolus dose of insulin to cover the anticipated additional sugars.The cookies turned out to be higher on the Glycemic Index (GI) than she expected, and her blood sugar started to spike.This spike was quickly detected by her Continuous Glucose Monitoring (CGM) system.Almost immediately, Marina got an angry call from her mother asking why she wasn't taking care of herself and if she forgot to pre-bolus for snacks.Marina was on the episodes of hyperglycemia (high blood sugar levels) and hypoglycemia (low blood sugar levels).Besides the short-term effects, blood sugar fluctuations over the long-term can lead to complications that damage major organs such as the heart, kidneys, eyes, and blood vessels.Therefore, people with T1D need to monitor and treat out-of-range blood sugar levels (highs and lows) by giving extra amounts of insulin or consuming sugars as carbohydrates 24/7.
Measuring blood sugar levels is critical for T1D control.As biosensing technologies become more accurate and at lower cost, technological tools such as the Continuous Glucose Monitoring (CGM) system promise to improve patients' T1D management and lower the risk of severe highs and lows [4,30,74,85].CGM is an FDA (U.S. Food and Drug Administration) approved medical device that works through a subcutaneous sensor that samples body fluids just under the skin to measure real-time blood glucose information throughout the day and night.The sensor is worn by a person with T1D and connected to a transmitter, which sends the blood glucose information wirelessly to a monitor (either a receiver or a smart device compatible with the CGM app) every five minutes.The monitor needs to be carried by the person with T1D.The monitor displays current blood sugar value and trend (Fig. 1).The monitor generates alarms (Fig. 2) under several conditions: 1) the blood glucose value is lower than (low alarm) or higher than (high alarm) a certain control range (commonly set as 70mg/dL -180mg/dL); 2) the blood glucose value is rising (rise rate alarm) or dropping (fall rate alarm) fast; 3) the blood glucose value might go beyond 55mg/dl in less than 20 minutes (urgent low soon alarm); 4) the blood glucose value is below 55mg/dl (urgent low alarm); 5) there is a technical issue (signal lost or no reading alarm).CGM can also share blood glucose information with caregivers in real time.By downloading and installing an app (Fig. 3) for followers (caregivers), caregivers can remotely receive the same blood glucose information and alarms as the person with T1D receives.
CGM is also an essential part of the "Closed-loop" insulin delivery system, which is known as an artificial pancreas device 2 (APD) [43,87].This system automatically administers insulin to patients with type 1 diabetes.It includes a computer or smartphone with a fully automated algorithm, a CGM, and an insulin infusion pump.The APD collects information from the continuous glucose monitor and utilizes the algorithm to determine how much insulin should be delivered by the pump [18].If the CGM readings are low, the insulin flow is paused, and if they are high, the flow is increased [80].Four APDs are currently available commercially and are widely used, including CamAPS FX, Medtronic 670G + Guardian sensors, Medtronic 780G + Guardian sensors, and Control IQ [26].In addition, there are DIY closed-loop systems like Loop, OpenAPS, and AndroidAPS that connect commercially available insulin pumps and CGMs with an open-source algorithm [80].These systems are not regulated or commercialized, but are used by patients and caregivers seeking low-cost availability and increased customizability [19].Although referred to as "artificial pancreas, " all current commercial and DIY closed-loop systems are hybrid systems [18,80,87], meaning they are not entirely automated and require users to manage some aspects manually, alongside the automated parts.To ensure proper system functionality and maximize effectiveness, users must comprehend the control algorithm (adjustment), verify system status (CGM readings are accurate, pump infusion set/cannula is functioning, and the algorithm is operating appropriately), keep track of carbohydrate intake, administer prandial boluses, understand how quickly carbohydrates will reach the bloodstream, and manage glycemic levels during and after exercise [18,26,87].
According to the American Diabetes Association's Standards of Medical Care [2]: "CGM is essential for creating the ambulatory glucose profile (AGP) and providing data on time in range, percentage of time spent above and below range, and variability."However, access to CGM is  reportedly difficult for patients in the US and globally due to the technology's high cost [41].The estimated cost of using a CGM is more than $15 per day [84], making it difficult for people with Fig. 3. Dexcom G6 app (left) and Dexcom Follow app (right) for caregivers.The Dexcom Follow app is designed to allow caregivers to remotely monitor the glucose levels of a person with diabetes who is using a Dexcom Continuous Glucose Monitoring (CGM) system.The app displays real-time glucose readings and trend arrows, alerts, and a trend graph.lower incomes, inadequate insurance policies, or those living in regions where CGM is not publicly covered to access the device [61].This high cost has resulted in a global disparity in equitable access to CGM [61].Moreover, even universal healthcare systems across the world do not universally offer coverage or reimbursement for CGM [29].In the US, a study [3] has shown that uninsured patients and patients with Medicaid are far less likely to access or use CGMs compared to their counterparts with commercial health insurance.
Furthermore, the emotional well-being of both patients and caregivers can be affected by continuous exposure to CGM measurements.Research [65] indicates that being constantly aware of diabetes results and sometimes the perceived failures to control diabetes can be stressful.In addition, some caregivers believe that CGM places additional pressure on children, making them feel isolated from their peers [37,65].Moreover, CGM alarms can generate significant stress and anxiety, as they may go off at inappropriate moments, interrupting conversations and serving as a reminder that blood sugar levels are outside the optimal range [62,65,73].This can make it challenging to get a good night's sleep.
CGM is the "standard of care for patients with Type 1 diabetes (T1D) [24]." However, the adoption rate of CGM is still low and many users have difficulty sustaining usage over time [57].In addition to the cost and emotional issues we discussed above, design issues are also believed to negatively affect the perceived usefulness of CGM [73].Little is known about how to fit the device into the daily lives of people who rely on it [69].To improve the usability of CGM, several third-party apps such as mySugr 3 , BlueStar4 , and Sugarmate5 allow users to combine CGM data with additional services.These apps focus on providing better insights into the data and personalized coaching.
This study is focused on the data sharing and collaborative functions of CGM.We seek to understand how families with teens navigate the care work supported by CGM.

Type 1 Diabetes in Teenagers
T1D is an increasingly prevalent disease in US children and teenagers [52].In 2020, roughly 200,000 people under 20 years old are living with T1D [23,25].The number is expected to hit 600,000 by 2050 [15].Though managing diabetes is difficult for patients of all ages, adolescence is the most challenging time for diabetes control.Among any age group, teenagers with T1D have the lowest rate of adherence and the worst control outcomes [90].Only 21% of them can reach the optimal control goal set by the American Diabetes Association (ADA) [16].This makes them more susceptible to earlier disease complications and premature mortality [5,16].Borus and Laffel [5] surveyed common barriers to teenagers' adherence.They concluded social and peer pressure, premature shift in control responsibility, risk taking behaviors, limited knowledge of diabetes and its risks, diabetes fatigue, and greater insulin resistance associated with physiological changes contributed to teens lower rates of optimal control.Managing teenagers' T1D is indispensable to family caregivers' involvement in day-to-day care tasks, including but not limited to giving insulin, monitoring blood sugar levels, treating out-of-range blood sugar levels according to blood sugar monitoring results, adjusting insulin delivery based on blood sugar level and activities, scheduling and preparing for meals and snacks, exercising regularly, arranging clinic appointments, managing supplies, and teaching others (e.g., teachers, relatives, friends) about diabetes [77,83].Caregiver's involvement in these tasks can be direct (caregiver and teenager perform a task together) and indirect (caregiver question/remind teenager about a task) [91].However, caregivers' involvement improves diabetes control only when caregivers and teenagers create a partnership "with shared responsibility and interdependency" that avoids diabetes-specific family conflicts [5,68].A greater perceived controlling, critical or restrictive parenting behavior in the partnership leads to lower adherence, worse control outcomes, greater family conflicts, and poorer quality of life [9,91].CGM is believed to increase family involvement in teenagers' diabetes control by sharing real-time blood glucose information with informal caregivers.While there are a few studies of patients' and caregivers' experience with CGM, there is a lack of in-depth inquiry about how greater parental involvement through CGM affects the parent-teen partnership in diabetes management [42,55,68].Our work begins to address this gap by studying caregivers' and teenagers' lived experiences of sharing diabetes control through CGM.We further propose design implications that mitigate conflicts in their collaboration, to improve teenagers' diabetes outcomes.

Supporting Informal Caregivers
Informal caregivers, also called family caregivers, can be defined as family members "who are heavily involved with day-to-day patient care activities" [12,79].Providing care is a demanding (usually full time) job, and informal caregivers are prone to a variety of psychological, physical, emotional and economic burdens associated with caregiving [12,79].Specifically, informal caregivers for people with T1D carry burdens such as responsibilities for providing care, misunderstandings between kids and parents, conflicts in the family, adhering to diet restrictions, difficulties in learning diabetes management skills, decreasing social activities, and financial difficulties [39].These burdens can lead to a "diabetes caregiver burnout" 6 , in which the caregivers feel exhausted, irritable, socially isolated, depressed, and anxious.Research in HCI has focused increasingly on designing to support informal caregivers.Studies have examined informal caregivers' needs in providing family care for a variety of chronic conditions-such as diabetes [64], cancer [71] and autism [38]-and designed social and technological systems to support family care work.Numerous studies have also considered computer-mediated social and emotional support for family caregivers and care recipients [12,22,45,46,81].
In addition, care is a complex endeavor that cannot be accomplished by one caregiver alone, thus informal care is often described as a cooperative work that is built upon coordinated efforts of informal caregivers and care recipients [6,59,78].Prior literature has explored ways to support collaboration and coordination in informal care [10,17,32,56].For example, Miller et al. [56] designed to support caregivers and patients during a hospital stay.Hong et al. [32] created an online social network for young adults with autism to request help and caregivers to collaborate remotely to distribute the labor of care.Buyuktur et al. [10] explored ways to construct the independence of care recipients.Findings from theses studies showed that the independence and autonomy of both caregivers and care recipients must be co-constructed through collaboration and coordination efforts of the care network.
In recent years, increased access to bio-sensing technologies made tracking an important part of informal care.Prior research discussed how individuals track their personal health and wellbeing [20,44,70].Studies also proposed health monitoring applications to assist individuals' understanding and reflecting upon their own collected data [48,49].Using the theoretical framework of sensemaking, Mamykina et al. [50] presented a method of understanding how individuals make sense of their chronic disease through self-monitoring data.Unlike the normative decision-making approach that centers around one-time decisions like therapy choices, or the problem-solving perspective that focuses on overcoming barriers to self-management goals, sensemaking involves organizing and finding patterns in lived experiences, identifying connections and dependencies, and making numerous daily choices related to vital self-management tasks.Katz et al. [35] investigated the usability and user experience of diabetes self-management apps that allow users to track and monitor their glucose levels, medication, diet, and exercise.Their findings indicate that while these apps offer a general overview that can aid in evaluating performance and serve as motivation for some users, they may not adequately address the cognitive and emotional needs for decision making.To address this limitation, the authors propose designing interfaces that offer explicit and specific actionable information while also keeping users mentally engaged with their data.The authors also put forward the concept of Fluid Contextual Reasoning to explain the response of individuals with diabetes to complex situations [36].This concept has practical applications as it can provide an alternative approach to the conventional habitual and sense-making thinking, and can address the cognitive demands of users in specific contexts that are not adequately accounted for in existing models.Additionally, Owen et al. [63] found that recording contextual details such as food intake and exercise (whether planned or completed) can aid the interpretation of glucose monitor readings and provide better assistance during non-routine occurrences.Similarly, through studies [60] on how user experience can impact the choices made by adults with T1D in regards to the adoption and usage of self-management technologies, O'Kane has uncovered contextual factors that impact the decision-making process when it comes to utilizing these technologies in real-world settings.These factors include the physical environment, social circumstances, cultural context, and individual differences.O'Kane's research highlights the importance of conducting in-situ studies to gain a better understanding of how the design of mobile medical devices can influence their adoption and use in everyday life.
Current health informatics design solutions typically treated tracking as an individual activity.Few studies examined the use of personal tracking in relation to a family's collaboration around managing complex chronic conditions [66].Toscos et al. [82] investigated how the use of a Blood Glucose Monitoring technology that automatically sends children's test results to caregivers impacted diabetes management and the family dynamics.Kaziunas et al. [37] studied a group of parents' practice-hacking their T1D kids' monitor to track blood glucose in real time.They found though "caring through data" gave parents peace of mind and teenagers freedom, it resulted in "extreme monitoring and a sense of loss of control" as parents obsessed over data and kids felt the numbers were prioritized over their social and emotional needs.In addition, Cha et al. [11] investigated the collaborative efforts of children with T1D and their parents in monitoring and managing blood glucose levels.The study also examined how parents can support their children in transitioning to independent self-care.The authors emphasized the importance of evaluating the children's knowledge and motivation levels and suggested developing appropriate strategies to ensure a balanced participation of parents and children in their collaborative efforts.
Recent advances have made tracking technologies and continuous blood glucose monitoring widely available to caregivers.Parents no longer have to hack a device to support their teens, and new sharing functions further allow the data to be disseminated among a set of connected caregivers.This creates new collaboration patterns of diabetes control not only between parents and kids, but also between additional connected caregivers.We build on the past work by identifying challenges and issues in supporting caregiver-teenager collaboration and informal caregiving coordination through real-time blood glucose sensing.

METHODS
We studied families' lived experiences of using Continuous Glucose Monitoring (CGM) through the creation of two Asynchronous Remote Communities (ARC).We created two ARCs on Slack and ran 20 ARC activities with a total of 32 participants (16 caregiver-teen dyads) over a 6 month period.The study and activities were approved by our university's Institutional Review Board (IRB).In the remainder of this section, we describe our recruitment and participant demographics followed by a discussion of activities we conducted in the ARC groups and how we analyzed the resulting data.

Participants
Participants of this study were recruited through six major Facebook support groups for Type 1 Diabetes (T1D) patients, families, or parents.We obtained the group moderators' permission prior to posting our recruitment message to follow the university's human subject regulations and Facebook's group policies.Families had to meet the following criteria for inclusion: 1) have one teenager (ages [13][14][15][16][17][18][19] in the household with T1D; 2) have been using a CGM with the data sharing function for over one year; and 3) have at least one caregiver of the teenager willing to participate in the study with the teen.
Forty seven interested individuals who expressed interest in the study were directed to complete a screening survey that asked the teen's gender, current age, age at diagnosis, and contact information for both the teen and the caregiver who intended to participate.We followed-up with an email providing guidelines for participating in an ARC and a link to an online consent form for individuals who are 18 years and older and an online assent form for teenagers that are too young to give legal consent.Twenty two families completed the screening survey and agreed to participate.Eighteen families enrolled in the study groups, while five families decided to withdraw due to time constraints and/or inability to access Slack.A total of 36 participants (18 1 shows demographic characteristics of the participants.Although we aimed to recruit a diversity of families, all participants were located in the United States and self-identified as white and all caregivers identified as women.One family (T08, C08) withdrew from the study after Activity 6, as a change in their health insurance policies discontinued their CGM usage.Another family (T14, C14) quit the study right after the study started for unknown reasons.In the 16 families that completed the study, all teenage participants wore Dexcom7 CGM sensors (G5 or G6) since it was the only brand that enabled automatic glucose-data sharing and sound alarms for blood glucose issues [27].All caregiver participants used the Dexcom Follow App8 to receive blood glucose data remotely.

Activities
We adopted the Asynchronous Remote Community (ARC) method [47] in order to conduct groupbased studies online with distributed participants.We created two online communities (workspaces) on Slack.We chose the Slack platform because it allowed participants to use pseudonyms for anonymity.We put teenage participants in one workspace and their caregivers in another to protect their privacy and minimize the impact of caregivers' presence on teenagers' responses.In each workspace, we created a weekly activity channel to run ARC activities.
We planned to run two activities each week for 10 weeks.After the first two activities, we decided to conduct only one activity per week to accommodate participants' busy work, school, and diabetes management schedules.We also paused the study for teenagers' finals and winter break.For each activity, we posted the description on the weekly activity channels and emailed individual participants a notice with completion deadline.Participants then replied to the pined activity post as instructed in the description.We anticipated an activity to take no more than 20 minutes for an individual participant to finish.However, some participants volunteered to spend more time each week to reply to others' posts.This often started deeper follow-up discussions around the study topic.
In this paper, we present our findings based on the first eight activities.We have observed that data saturation was achieved by this point, as participants began to echo their perspectives on family practices, opportunities, and challenges related to the use of CGM.Table 2 shows the topic of each activity.In Activity 1, we asked participants to introduce themselves and talk about their diabetes management routine.We also asked about how they use CGM and their expectations on continuous blood glucose tracking technologies.In Activity 2, we created a task-based scenario in which a teenager with T1D (for the teenage group) or caregiver of a teenager with T1D (for the caregiver group) was considering adopting a CGM.We asked participants to provide peer advice for using CGM in a family, focusing on challenges/conflicts that might arise.After Activity 2, we read through participants' posts and identified three major aspects that led to family challenges/conflicts, tracking remotely, setting and responding to alarms, and making treatment decisions with CGM readings.Activities 3-5 each involved discussing one major aspect, asking participants to elaborate on their practices, what works well and what does not, and potential strategies for improvement.In Activity 6, we surveyed individual participants' benefits and burdens in CGM usage, using scales developed by Suh et al. [75,76].Instead of quantifying participants' experience, we adopted the scales to probe unique user benefits and burdens CGM brought by comparing it with standard computing systems for which the scales were created.Therefore, we asked participants to comment on benefits and burdens that were not covered by the scales upon completion of the survey.In Activities 7 and 8, we involved participants in designing technology solutions to improve family collaboration around teenagers' T1D control through CGM data.We asked participants to sketch their design ideas, take a picture of the sketch, write to explain their ideas, and post the picture together with the writing on the weekly activity channel.The sketching activities help us explore design opportunities and start discussions among participants as they comment on each other's design ideas.In Appendix A, we provide a sample activity description we posted on the weekly activity channel and a response we got from one of the participants.
Data for activities were exported from the online study groups to a password protected machine.The Data contained 717 posts, 68 survey responses, and 34 participant design artifacts.We removed all personally identifiable information and assigned a unique identifier to each participant (Teenager: T01-T18; Parent: P01-P18).We adopted the reflexive variation [8] of thematic analysis approach [7] to conduct iterative coding analysis on the data.In the first round of analysis, the first author conducted an interpretative, reflexive, open-coding of the data to gain an understanding of the family practices, opportunities, and challenges associated with CGM usage.The initial themes developed by the first author were shared and discussed with the study team in group meetings.Based on the initial thematic codes, the study team decided to concentrate on exploring how CGM can facilitate caregiver-teenager collaboration and informal caregiving coordination.The team then iterated on related codes and generated commonly occurring codes through further discussion in group meetings.

FINDINGS
The data analysis revealed the peace of mind caregivers obtained and the freedom teenagers got from remote Continuous Glucose Monitoring (CGM).More importantly, it also revealed issues in the collaborations supported by remote CGM.These issues introduced conflicts not only between caregivers and teenagers but also among caregivers who care for the same teenager.In what follows, we discuss the issues in the context of supporting caregiver-teenager collaboration and informal Participants were asked to reflect on how they treat blood sugar issues using a CGM, causes for over-correction, and potential ways to improve their decision making.

Text 6
Survey.Participants filled out online surveys about user benefit and burden scales then commented about unique benefits and burdens in CGM usage that were not covered by the surveys.

Survey & Text 7 & 8
Sketching Activities Participants were asked to comment and upload pictures about their technology designs that improve family collaboration around teenager's T1D control through CGM data.
Draw & Text caregiving coordination.CGM is designed to provide a visual representation of blood glucose change over time.We recognize that for readers who are not familiar with this T1D technology, a textual description of a potential problem might be too abstract.Further, it might be hard to gain experience with this technology without a T1D friend or relative.Therefore, we have attempted to help the reader visually situate our findings in reference to the current technology by incorporating screenshots of the first author's own Dexcom G6 app in operation (Fig. 4-8) to illustrate a number of common T1D situations that our participants describe.

Caregiver-Teenager Collaboration
Prior to the availability of modern CGM systems, a few tracking tools were created that allow parents to access children's blood glucose readings remotely.These tools informed the design, implementation and FDA approval of remote CGM systems.Prior studies [37,82] in HCI have investigated experience of families that adopted these experimental tools for children's T1D management.
Results of these studies revealed the eagerness of parents to gain peace of mind over children's health and safety through remote blood glucose data sharing.However, they also pointed out the potential of creating social and emotional conflicts by "caring through data [37]."In this study, through an analysis of caregivers' and teenagers' lived collaborative experience with a modern CGM system, we further identify social and technological issues that caused conflicts between caregivers and teenagers.We found that a lack of visibility into the behaviors behind blood glucose data, difficulties in balancing a teenage life with T1D control, and technical limitations of current CGM systems are three major issues that challenge the caregiver-teen collaboration supported by remote glucose data sharing.

Lack of Visibility into Behaviors behind Data.
When connected to the CGM system, caregivers can view teenagers' blood glucose levels in real-time and be alerted when the levels are (heading) out of a control range.This allows caregivers who track remotely to be aware that an issue occurs in their teenager's blood glucose level.However, there is not a convenient way for caregivers to gain knowledge about the cause of the issues and more importantly, whether and how the issue is addressed by the teenagers.This is reportedly anxiety-inducing for caregivers who track remotely because if left untreated, a blood glucose issue would threaten their kids' imminent safety and future health.As a result, caregivers reportedly having to "jump on", call/text the teenagers as soon as they receive a CGM alarm to figure out whether the teen has acknowledged the alarm and made appropriate treatments, . . .my husband and I appreciate the cgm so much and it gives us such a breath that we get thrown off whenever there are issues.I feel kind of bad for my son when he has a high or low because both my husband and I text or call or yell for him.I know he finds it annoying.So it's great that we know about the high/low but at the same time I think sometimes we jump on him faster than necessary.-P11 To gain more visibility into teenagers' reactions to the CGM alarms, some caregivers reported developing strategies that require teenagers to report their behavior back to caregivers immediately.For example, P09 set up a rule on how her daughter should text parents a code to acknowledge alarms and report treatments, If her [CGM] alarms whenever Im NOT with her (at school, at a friends, etc), we ask that she responds in a code she set up.She landed on texting c for a correction for a high alarm and s for snack in response to a low alarm.If its not urgent low, I give her until the 2nd glucose notification (5 min) to text me (c or s) and if she doesn't then I text/call.If urgent low I give her about 30 sec-1 min.This annoys her at times, but she doesn't quite understand the parent perspective of especially urgent low notifications.-P09 The teenager set up codes that represented different treatments for when an alarm went off and she was not with her parents.The girl was asked by her parents to treat and text the code to her parents within 5 minutes for non-urgent alarms or within 1 minute for urgent alarms.Otherwise, P09 will call and ask about whether/how the issues were resolved.Additionally, some caregivers reportedly send helpers to check the teenagers if the teenagers did not report back promptly as the caregivers asked, I would give her some time to correct one way or the other, but if I didn't see that happening, I would call the nurse, who would buzz the classroom and call her to the office to treat the high/low.This my daughter didn't like, of course.-P18 We found while teenagers acknowledged good intentions of their caregivers, they did not appreciate the strategies caregivers adopted to "bug [T16]", "text [T06, T11, T04]", "micromanage [T15]", "call [T04]", and "ask [T02]" about teenagers actions all the time, When your parents are constantly able to see your blood sugars they will be asking you if you've treated high and low blood sugars which starts arguments because you get annoyed with them asking you all the time.-T02 Teenagers reportedly argue that they were not trusted in the collaboration as efforts they made to address blood glucose issues were not acknowledged by their caregivers, as T13 complains "I know that I need to fix it but she feels she needs to remind me every time."In addition, the continual feeling of being nagged prohibited some teenagers from taking ownership of their diabetes, and instead they considered it their parents' responsibility to make sure it gets taken care of, because "[I would like] taking care of myself without them nagging me but they're my parents so I can't really stop that.[T11]".This, in turn, makes parents argue about their teenagers' lack of accountability.
Moreover, having to answer phone calls, reply to status inquiries and text caregivers reportedly distract teenagers from dealing with the blood sugar issues.As we've discussed above, teenagers have to figure out an blood glucose issue and deliver treatment within a few seconds/minutes after an alarm goes off.However, as P11 realized, caregivers "freak out and start texting and calling him in what is already a stressful situation" is sometimes counterproductive and frustrates the teenagers.
Our finding shows that while remote CGM informs caregivers of their teenagers' blood sugar issues, this does not bring total peace of mind to the remote caregiver-teen collaboration.Without visibility into teenagers' reactions to the alarms and awareness of treatments delivered, caregivers had to contact the teenagers to assure their health and safety.However, caregivers' frequent inquiries about CGM alarms leads to caregiver-teen conflicts because teenagers reportedly feel not trusted in the collaboration and get distracted in already stressful situations where they deal with unpleasant blood sugar problems.

Difficulties in Balancing a
Teenage Life with T1D Control.The fact that teenagers agreed on wearing a CGM and sharing data manifests their willingness to collaborate with their caregivers for better T1D outcomes.When they track remotely, we found parents always expect their teenagers to address blood glucose issues instantly.However, sometimes teenagers cannot attend to an issue as soon as the alarm goes off.Teenagers' response could be delayed by some other day to day challenges in a teenage life, as T01 says: "I do have a life and I'm not always able to immediately answer a blood sugar alert, just like I can't always respond to texts right after they're sent."We found three major challenges that delay teenagers' response to CGM alarms, feeling normal in public, managing body image, and maintaining friendships.
The first challenge is feeling normal in public.Receiving CGM alarms and making treatments in public reportedly bring up a feeling of abnormality.This feeling makes teenagers like T16 "hate blousing in public/in front of friends and sometimes I don't."Therefore, teenagers were found to pause CGM alarms and wait until they could find a private space like a restroom to check and address blood glucose issues, as P17 observed, One reason is that alarms going off in public tend to embarrass her.As a result she will sometimes silence her phone and/or receiver. . .I have also noticed her acknowledging the alarm notification on her phone but not actually checking the number.Kind of like hitting the snooze button with your eyes shut.-P17 T17 reported feeling embarrassed when an alarm goes off in public because it sounds loud and often attracts unwanted attention.As a result, she would silence it as soon as possible and go back to check the numbers later.P17 does not like this approach because she cannot know whether an alarm was missed or intentionally paused when she's not with the teen.
The second challenge is managing body image.Concerns about weight gain reportedly make some teenagers delay treatments for non-urgent lows.When blood sugar levels dropped below 70 mg/dL, teenagers were instructed to use carbs to bring them up.The common conception that carbs drive the numbers is only partially true [13].Different organs in the human body retain or release glucose at different concentrations and will respond to the amount of glucose in the bloodstream.This can result in occasions where the numbers can go up a little bit without taking additional carbs.Teenagers like T06 observed this and therefore she would like to "ride" non-urgent, close to target (55 -69 mg/dL) lows by not giving fast-acting carbohydrates immediately but instead waiting to see if her blood glucose level would fluctuate back to the control range itself.This way she can avoid weight gain from taking additional carbohydrates.It could also eliminate the need to treat potential blood glucose spikes caused by overtreating lows, My daughter sometimes wants to ride a low (68-70) before treating in case it goes up on her own.I worry about that approach.Seeing the #s every 5 min makes me worry when I'm not seeing the # go up.There are times it adjusts itself and she's higher after 15-20 but that worries me.Price you pay for seeing constant #s I guess.I do understand her perspective of not liking to eat or drink when she's not hungry or thirsty and since the diagnosis she's gained weight which for a teen wasn't really on her agenda.-P06 While P06 acknowledged that there were times when her daughter's "riding a low" approach (Fig. 4) worked out, she did not like this approach.Because not seeing the numbers going up in a short amount of time made P06 worry about her teen's safety.Fig. 4. Visual example of riding a low.The user let their blood sugar levels naturally rise after experiencing a hypoglycemic (low blood sugar) event without immediately consuming carbohydrates to raise their blood sugar levels.This image is from a CGM user who was not part of the study and is just to help situate the participants' statements.
Last but not least, maintaining friendships is challenging for teenagers with T1D.These teenagers are asked to stay vigilant over blood sugar which often distracts them from engaging in social activities.To care for their friends' feelings, some teenagers reportedly let lose diabetes control by not treating high blood glucose levels when they hang out with friends.Because living with high blood sugar levels for a short period cannot cause immediate harm to their health.Besides, it has little impact on the long-term blood glucose control goal which is commonly set at 70% time in range and less than 25% time in hyperglycemia.Teenagers not treating high blood glucose levels during social activities is in parallel with caregivers who reported not adjusting highs overnight because it "is not going to adversely impact their health in the long run" but has other benefits such as higher sleep quality, as P07 explained, Not every high BG needs to be adjusted overnight.Some will come down on their own and even if they don't, waking up once in a while about their target BG is not going to adversely impact their health in the long run.Reacting to every single high glucose alarm, however, will leave you (and your child, if they're a light sleeper like mine) chronically sleep-deprived -which does cause issues in cognitive function and health both in the short-term and over time.Never underestimate the power of a chunk of uninterrupted sleep in making good decisions (about T1D care and life in general).-P07 However, when tracking remotely, caregivers reportedly take teenagers ignoring glucose alarms as an indicator of "gotten distracted" by friends.Therefore, they would "keep a closer eye on the numbers" and send reminders, I feel like I need to keep a closer eye on the numbers when my daughter spends the night away from home or even when she's out with her friends.She is super responsible but she has gotten distracted a few times when out with friends [because] she just wants to feel like a typical teen.-P06 Above we discussed challenges teenagers encounter in their day to day life living with T1D.These challenges could impair teenagers' ability to stay vigilant over CGM data and delay their response to address blood glucose issues.However, teenagers not addressing glucose issues immediately is at odds with caregivers' goal for tighter control.Some teenagers reported "getting in trouble" with their parents, I would get in trouble with my parents if my blood sugar did not change within a few minutes of the high or low alert going off.I got called irresponsible and told they were going to take my phone back to the store if I wasn't going to use the app, etc. -T01 This creates caregiver-teen conflicts because teenagers feel that they are being abstracted as a number and that their parents largely see them as such.They "can't get away with anything [T16]" As a result, some teenagers reported over-treating an issue to speed up changes in the readings to avoid parents' judgment, interference, and punishment, Another time I was coming home from a friends birthday party and my blood sugar was in the 300s so I tried to get it down as fast as possible to avoid my mom getting upset over it so I bolused around 12 units when it should've been like a 5 unit correction and of course I went low by the time I got home.Next time I need to remind myself it is ok for my blood sugar to be off sometimes but over correcting just sends it into a bad loop.-T18 In this case, T18 reported over-correcting her high blood sugar levels by giving more insulin than needed to "get it down as fast as possible to avoid my mom getting upset over it."Over-correcting (Fig. 5) is extremely dangerous and the teenager ended up with low blood sugar levels when she arrived home.

Technical Limitations of Current CGM Systems.
In addition to the teenage life challenges we discussed above, we found some technical limitations of current CGM systems such as false alarms, lag time, and stubborn blood glucose levels also lead to caregiver-teen conflict around teenagers not acting on and fixing the numbers speedily.
The first limitation is false alarms.While being the "standard of care" for people with diabetes, current CGM does not run error-free.There are times the system gives inaccurate readings especially when the wearer is asleep, exercising, or dehydrated 9 .Making treatment according to a false alarm is harmful so the wearers are always recommended to compare CGM readings with how they feel and a finger prick value before making any treatments 10 .When teenagers believe that an alarm was a false one, they reportedly ignore it intentionally.For example, T06 said he had ignored some urgent alarms because they were caused by compression, Fig. 5.A visual example of over-correcting.The user gave more insulin than needed to quickly bring down their high blood glucose levels.This results in a rapid drop in blood sugar levels, also known as hypoglycemia.This image is from a CGM user who was not part of the study and is just to help situate the participants' statements.
False readings (mostly false lows) -honestly, most of the time I forget I have it on.Unfortunately with the [CGM], if I lay on that spot for a while, then I am very likely to get a false low.. this causes a lot of additional stress because then my parents are freaking out about my urgent low and it keeps beeping over and over again. . .-T06 When CGM users put too much pressure on the sensor area, the wire cannot get enough fluid for accurate readings.This leads to compression lows (Fig. 6).Compression lows do not need to be treated and they are commonly resolved a few minutes after the pressure is removed.Fig. 6.A visual example of a compression low.Compression of the CGM sensor at night caused the sensor to give a lower blood sugar reading than the actual blood sugar level.When the user changed their position, the compression was relieved, and the sensor could then detect the actual blood sugar level, resulting in a gradual recovery of the blood sugar levels on the CGM graph.This created the V-shape pattern, where the blood sugar levels dropped rapidly and then recovered almost as quickly.This image is from a CGM user who was not part of the study and is just to help situate the participants' statements.
The second limitation is the lag time.CGM uses interstitial fluid to test blood glucose levels, which can have a 5-20 minute lag relative to the actual blood sugar concentration.As a result, While teenagers reported making a treatment promptly, it takes some time for the treatment to be reflected on CGM readings and visible to caregivers who track remotely, as T06 complained, 99% of the time, my [CGM] readings are super accurate.However, there also can be a 15-20 min delay with readings.this gets really annoying at times because i'll get a notification that i'm low, and then i'll treat right away, and then [CGM] keeps going off because it hasn't updated yet.. so then my parents are again constantly texting and calling all worried even though i've already told them i treated.a lot of the time, this makes me feel as if i have no freedom and like my parents think i can't deal with my diabetes myself.-T06 Last but not least, having "stubborn (Fig. 7)" highs and lows reportedly delays reflection of treatments on CGM readings.P12 explained what a stubborn low is, My daughter has had some stubborn lows that would just take forever to come up.Readings would range from 34,50,57.She would eat 15-20 carbs and we would wait and test and she would still be right around the same number so I would give her additional carbs.Wait and check and [she] was still low so [I] gave a few additional carbs.-T06 Fig. 7.A visual example of a stubborn high.Consistently high blood sugar levels that were difficult to bring down during the menstrual cycle where hormonal changes caused insulin resistance.This image is from a CGM user who was not part of the study and is just to help situate the participants' statements.
In the case of false alarm, lag time, or stubborn glucose level, teenagers either do not make any treatments and wait for the issue to resolve itself or make a treatment but wait for the treatment to be reflected/redone.However, similar to teenagers not being able to address an issue immediately due to life challenges, they reported "getting in trouble" with caregivers for not fixing the numbers as soon as possible, as T06 says, "so then my parents are again constantly texting and calling all worried even though I've already told them I treated.a lot of the time, this makes me feel as if I have no freedom and like my parents think I can't deal with my diabetes myself."This causes caregiver-teen conflict around trust and freedom as T06 mentioned.It also makes teenagers overcorrect to move the readings, as P01 stated, "[T01] has a few times [of overcorrecting] and it is REALLY scary.Especially around period time, she will correct and correct, and if it doesn't move, she gets really frustrated."

Informal Caregiving Coordination
Prior studies [37,82] discussed children's blood glucose data sharing between a caregiver-child dyad.However, we found all families in our study shared teenager's CGM data with multiple caregivers, that could include relatives, school teachers, sports coaches, camp facilitators, healthcare workers, teenage friends, and friends' parents.This reportedly extend the caregiver-teenager collaboration on T1D control by enabling primary caregivers to collaborate with others involved in the teenagers' caregiving, Being able to use the [built-in data sharing feature] with multiple users is a plus.It gives us an extra set of eyes and ears for those times that we are not nearby or just miss a low for whatever reason at that moment.-P17 As P17 says, the built-in data sharing function of modern CGM systems allowed multiple informal caregivers to track and care for the teenager simultaneously.However, we found three issues that inhibit effective informal caregiving coordination supported by remote CGM, unclear task assignment, difficulties in maintaining standards of care, and a lack of caregiver rotation.

Unclear Task Assignment.
When connected to CGM, all caregivers received the same blood glucose data and alerts at all times.However, there is not a dedicated place for caregivers to communicate task assignments -who was supposed to respond to what CGM alarms at which time.Participants reportedly used ad hoc communication to assign these caregiving tasks, which is not effective at making task assignments clear to all caregivers who receive CGM data and alarms simultaneously.This reportedly caused "chaotic" situations in which several caregivers responded to the same alarm altogether as well as stressful times when no caregiver would respond to an alarm at all.For example, T01 complained about receiving calls from multiple caregivers for the same alarm, Too much information about your blood sugar is suddenly available to everybody who has access to the app.When I first got the [CGM], my parents gave the app code to the school nurse, my relatives, my friends parents etc...It was a nightmare.I spent more time out of class than in it because the nurse would call for me every time a low or high alert went off.When more than five adults at a time can see your blood sugars, it's chaotic.I would get into arguments with my parents because one of them would text me about an alert, then that parent wouldn't tell the other parent that they'd already asked, and I'd get a second panicky phone call and usually end up snapping at whoever called after the first time.Then came the texts-my uncle from across the country would text my about lows, stuff like that.Clueless friends' parents would wake me up every hour when I tried to sleep over anywhere because they weren't sure if the number was okay.-T01 As T01 stated, caregivers who followed her CGM readings did not communicate with each other about who was supposed to take action and update it to others when an alarm went off.Instead, they all called or texted her individually.Having to answer several redundant inquiries for the same alarm severely disrupted T01's daily activities such as school and sleep.She often "end[s] up snapping at whoever called after the first time." As stressful as caregivers all responding at the same time was, another concern was that no one would respond to an alarm at all.Participants reported this often happened when a group of caregivers were following together.Without assigning a lead responder explicitly, the diffusion of responsibility made individual caregivers assume that other caregivers either are responsible for taking actions or have already made a treatment, so they all ignored the alarm or turned it off unintentionally.This puts the teenagers' life at risk, especially at night when an untreated low could lead to the Dead-in-Bed Syndrome [86], I frequently argue with my husband if he doesnt hear the alarm and then [T01] gets mad at us if we dont hear the alarm and we are upset with her if shes sleeping with the phone right next to her alarming (not really upset, but you know what I mean).-P01 P01 described that her family members got mad at each other when someone finally woke up from a beeping alarm but found others in the house sleeping through it, because they are supposed to be responsible for it too.To avoid such occasions, P09's family elected "night guards" that were responsible for all alarms at night.However, they found caregivers who were not on duty could not opt out from receiving alarms.There was no way for the night guards to turn off an already treated alarm for other followers remotely.Therefore, P09 and her husband had to go back and forth to their daughter's room to tap her phone so she would not be woken up by alarms that had already been taken care of by her parents.
Last night was one of those nights where my husband and I were up 5+ times fighting BG numbers.Her [CGM] would alarm, so we'd go tap it and respond.Then 5 min later ours would go off.Then 25 min later hers would go off and we'd have to get up, go back into her room and tap her phone because her BG was coming down but still high.This happened all night long.-P09 Unclear task assignment not only puts teenagers' health and safety at risk, but also significantly impairs caregivers' daily activities as they have to pay attention to CGM numbers all the time and respond to all alarms regardless of who else is tracking and would help the teenagers at the same time.This is a waste of already overwhelmed caregivers' time and efforts.And it causes "chaotic" situations in the teenagers' lives as they have to respond to redundant inquiries from different caregivers for the same alarm.

Difficulties in Maintaining
Standards of Care.Caregivers for teenagers with T1D need to follow certain treatment processes as the standards of care for treating high/low blood sugar levels to achieve best control outcomes.The standards include how frequently the CGM readings should be checked, whether a CGM reading was confirmed with a finger prick prior to treatments and how many carbs/units of insulin were used for treatment each time.Participants in our study found that keeping the standards of care consistent among caregivers became challenging as the data sharing function of CGM allowed more people to participate in teenagers' diabetes care.For example, in P15's family, P15 and her husband were not on the same page for trusting CGM readings to make treatment decisions, standards of care.Is everyone on the same page for things like confirming a low / high with finger stick?Checking for ketones?Our example is that if my husband got up at night he was much more comfortable relying on the cgm number and treating a low or high.I preferred a finger stick as confirmation -P15 P15 prefered to double-check with a finger prick to make sure the readings were accurate.Her husband did not usually follow this extra step.Similarly, P15 accused her husband of being more guilty of over treating their son's lows at night, My son is very difficult to wake up and therefore we would often give him an entire [Glucose Control Drink], which is about 35 carbs vs suggested 15.One parent was more guilty of this than the other.-P15 When a low happened, families were commonly suggested to follow the "15-15 Rule [1]" by doctors -give 15 grams of carbs and wait for 15 minutes.If the low persists, repeat.P15's husband seemed to be less patient about waking his son multiple times and giving small portions of carbs each time.Instead, he tends to give "an entire [Glucose Control Drink]," 35 carbs altogether.The overtreatment could lead to a rebound high shortly after.
Participants discussed difficulties in maintaining standards of care.The most frequently reported difficulties in achieving consistency between caregivers are caregivers having different knowledge and experience levels of care, frequent "hand-offs" between caregivers, and personal factors such as sleep deprivation.
The varying levels of knowledge and experience reportedly make it difficult for some caregivers to be knowledgeable about and follow the standards of care.What is considered as a habitual practice for primary caregivers might take considerable time for other caregivers to learn.For example, P17, the primary caregiver of her T1D daughter, described why she does not trust other caregivers for monitoring, "Most people just don't get it.They don't realize how quickly numbers can change." -P17 P17 worried that other caregivers would not check CGM readings as frequently as herself because they do not know how quickly the teenager's blood sugar levels change.
Another difficulty reportedly comes from making frequent "hand-offs" between two caregivers.When one caregiver substituted for the other one, substantial information needed to be communicated to help the successor make informed treatment decisions.However, many families did not have a formal way to facilitate the information exchange.Sometimes the successor cannot follow the standards of care and "give an appropriate correction" because an important piece of information was missed, I've rage bolused a couple times recently because my daughter grazes at night and estimates carbs...and in my humble opinion doesn't count enough carbs...so at night she's high and I don't know exactly what she ate, so it's hard to give an appropriate correction.-P18 P18 was responsible for guarding her daughter at bedtime.She reported having "rage bolused (Fig. 8)" -using an aggressive amount of insulin than what is suggested by the standard of care to correct frustrating high blood sugar levels because she was not told what her daughter ate for dinner and how the carbs were counted.Fig. 8.A visual example of rage bolus.The user incorrectly estimated the number of carbohydrates in a meal and then gave a large dose of insulin to compensate for the mistake, which resulted in a rapid drop in blood sugar levels.This image is from a CGM user who was not part of the study and is just to help situate the participants' statements.
Personal factors such as sleep deprivation also reportedly impacted individual caregivers' ability to follow the standards of care.For example P04 says she makes erroneous treatments generally at night due to lack of sleep, Generally it happens in the middle of the night or late at night and we are simply impatient or we forget to finger poke to double check the [CGM].I don't make my best decisions at 3 a.m.-P04 Families in our study reported trying different strategies to keep the standards of care consistent.For example, P04's family "have a set action plan written out on the fridge for dealing with lows." It reminded caregivers not to over treat a low by letting the teen "eat up the whole kitchen [P16, P04, P09]." Similarly, P03 reported that they "keep a testing Glucose monitor in his room " so caregivers remembered to double-check with finger sticks at night.While these strategies might have improved the consistency of care, the notes and artifacts were cumbersome to use and hard to be passed between different caregivers.Besides, there was no way to make sure that the "message" was captured by everyone.

A Lack of Informal Caregiver
Rotation.Above we discussed unclear task assignment and difficulties in maintaining standards of care as two issues in supporting caregiver coordination.We found that these two issues also result in a lack of informal caregiver rotation.The primary caregivers are reluctant to share care responsibilities with other caregivers, because they are afraid of missing CGM alarms.Additionally, they feel that the risks of alternate caregivers providing uneven treatment because of unclear task assignment or not following the standards of care.For example, P16 insisted that she always needed to stay on top of the CGM alarms as well as the teenager's care because she did not trust other caregivers (her husband and teen) for those, I find myself waking up to check my phone.She doesn't wake up for her alarms, and my husband is also a deep sleeper, so I feel that if anything is going to happen at night, I need to be on top of it.-P16 This lack of rotation of caregivers hinders informal caregiving coordination as well.We found it exacerbates the primary caregivers' fatigue and disengages other caregivers.Having to monitor and respond to CGM alarms 24/7 exacerbates the primary caregivers' fatigue, which reportedly impared their response to alarms and decision-making ability.Primary caregivers like P18 found it harder for her to wake up by CGM alarms at night due to care fatigue, I have to set an alarm on my phone to get up and go check on her, which is hard because I'm so fatigued that sometimes I don't wake up.-P18 Additionally, some primary caregivers reported finding themselves making defective treatments at night because "Needing/wanting better sleep sometimes drives poor decision-making [P09]."For example, P09 overtreated her teen's low, Stubborn lows in the middle of the night. . .[CGM] went off about 6 times, reading urgent low (50s) within a matter of 4 hours, giving 15-20 carbs each time.After the 6th alarm, I was so utterly exhausted that finally gave her a chocolate milk box (28 carbs) AND a 20 carb fruit snack.She skyrocketed by early morning, but I was so done with treating lows.Middle of the night stubborn lows are the death of me.-P09 The lack of rotation could also disengage other caregivers in that they rely entirely on the primary caregivers for the teens' care.This reportedly makes other caregivers impractical and inexperienced.However, primary caregivers cannot always be available for caregiving.As Chen et al. [12] discussed, primary caregivers also needed to balance caregiving with their own physical, emotional, and social needs.Sometimes the primary caregivers have to give priority to other life obligations and pass the caregiving tasks to others.Accidents reportedly happened when caregiving tasks were transferred to other caregivers.For example, P15 had a painful experience with trusting other caregivers for night duty while she was out of town, Im primarily in night duty as the [same].A couple years ago, while out of town, my son had a low at 3:30 am.Nobody in my house responded.I was about to call 911 and send a squad over when his low stabilized.-P15 No caregivers in her house responded to her son's low alarms at midnight.This was scary and she was about to call 911 for help remotely.Luckily, the low stabilized itself and did not progress to an urgent low in the end.
Our analysis shows that CGM and its data-sharing function not only support remote collaboration between a caregiver and a teenager but also among caregivers for T1D management.This aspect of collaborative care has not been documented before.As well, we found several issues that need to be addressed to make the caregiver-teen collaboration and informal caregiving coordination effective and efficient to eliminate conflicts and protect teenagers' health and safety.We discuss some of these implications more thoroughly in the next section.

DISCUSSION
The findings from this study help us to better understand a special type of real-time, synchronous sensing that has not previously been available.This new kind of real-time synchronous sensing is a special case that can help to inform growing capabilities in health tools designed for collaborative use between parents, teenagers, and other caregivers.Our findings suggest that the current design of such tools can generate conflicts and lead to potential health risks, even despite the benefits of the devices.In the following subsections, we discuss our empirical findings working to link them to prior literature in HCI/CSCW.As well, we provide implications for the design and study of real-time, synchronous health tracking tools that attempt to improve caregiver-teenager collaboration and support for informal caregiving coordination.

Design to Leverage Teenagers' Self-care Assets
Through our ARC activities, participants shared their lived experiences and revealed the challenges that adolescents with Type 1 diabetes face in balancing the demands of their condition and their teenage life.For example the need to take additional carbohydrates to treat low blood sugar levels conflicting with weight control and body image management.Adolescents also struggle to maintain friendships due to the constant need to monitor their blood sugar levels, which can be distracting during social activities.Additionally, they face difficulties in feeling normal and living a typical teenage life due to medical alerts and treatments in front of others.
CGM has the potential to address the challenges faced by adolescents with T1D by facilitating family caregivers' involvement in teenager's blood glucose monitoring and control.However, our study revealed that the current design of CGM devices, while providing parental support, has led to conflicts in the caregiver-teenager collaboration.Adolescents feel micromanaged and distrustful of their own ability to manage their condition, leading them to shrink their self-care responsibilities or take life-threatening risks.
Current CGM design views teenagers with chronic conditions as passive recipients of care, in need of constant monitoring and management.However, our research suggests that adolescents play an active role in managing their condition and have developed experiential knowledge and strategies that are often overlooked by their caregivers.For instance, some teenagers "ride" nonurgent, close-to-target lows to control weight gain and avoid potential blood glucose spikes caused by over-treating.Others "pause" CGM alarms and find a private space for treatments to avoid feeling embarrassed in front of peers.Additionally, some teenagers allow their blood sugar levels to run high temporarily and then treat them after social events, as this does not pose an immediate risk to their health and benefits their social life.Therefore, we suggest future CSCW/HCI researchers and designers adopt an asset-based approach [40,51] to empower teenagers' self-management role and ensure balanced participation of caregivers and teenagers in their collaboration through CGM.
The asset-based approach is a framework that aims to identify and leverage the existing capabilities, strengths, and strategies of individuals or communities in technology design.This approach differs from traditional needs-based approaches (e.g., [32,38,71]) that focus on identifying and addressing deficiencies or gaps in knowledge, skills, or resources.Instead, the asset-based approach recognizes the positive attributes or assets that individuals or communities possess and builds upon them to create tailored, empowering solutions that are more effective and sustainable in the long term.By adopting this approach, designers can better support individuals or communities in achieving their goals and fostering their self-efficacy [28,34,40,51,72,88].
In the context of T1D management and CGM design, an asset-based approach would involve designing to identify and leverage the experiential knowledge and strategies that teenagers have developed to manage their blood glucose levels and navigate their social lives.One way to achieve this is to incorporate features that allow teenagers to record, track, and share their own strategies for managing blood glucose levels with caregivers.For instance, the "riding a low" strategy mentioned in the previous finding could be documented within the CGM device, enabling teenagers to reflect on their decision-making while helping caregivers to better understand the teenagers' tracking needs and circumstances.Additionally, CGM devices could leverage teenagers' strategy for feeling normal in public and feature a discreet mode that enables teenagers to pause non-urgent alarms without drawing attention to themselves.It is important to inform caregivers that the teenager is using this mode and has acknowledged and paused the alarm to give the caregivers peace of mind that important alarms are not being missed, and avoid sending unnecessary reminders that might cause embarrassment.Furthermore, the CGM could offer greater customization and personalization options to help teenagers balance social lives and diabetes control.For example, teenagers could set their own time-in-range and average blood glucose goals and earn "cheat time" by meeting those goals.During cheat time, their blood sugar level could run slightly higher without triggering alerts or reminders from caregivers.This could provide a sense of autonomy and control for teenagers while still ensuring their safety and well-being.
In addition, the approach would need to be tailored to the specific needs and circumstances of each teenager, taking into account age-dependent differences in skills and experiences.For example, a 14-year-old might have developed strategies for managing T1D in sleepovers, such as taking breaks to check blood sugar levels or carrying snacks to treat low blood sugar.A 19-year-old, on the other hand, might have developed different assets, such as the ability to manage their condition while living independently or balancing the demands of college with T1D management.An asset-based approach need to identify and leverage these assets to develop solutions that are tailored to the individual's specific needs and circumstances.
Overall, an asset-based approach to designing CGMs and their interfaces would involve a shift away from the traditional need-focused approach that views teenagers with T1D as passive recipients of care, and toward a more collaborative and empowering approach that recognizes teenagers strengths and capabilities in the collaboration.

Design for Caregiver Coordination
Previous CSCW/HCI literature of T1D has primarily focused on the collaboration between individual caregivers and children or teenagers with T1D [37,82].However, our study suggests that the CGM data of teenagers is often shared with multiple caregivers, and there is a lack of technological support for coordinating these caregivers who simultaneously monitor the data.Caregivers, at best, use ad-hoc communication when collaborating with each other, which is ineffective in assigning tracking tasks, maintaining care standards, and arranging caregiver rotations.When caregivers fail to collaborate effectively, teenagers' lives become chaotic, their health and safety are put at risk, and conflicts arise among informal caregivers.
Researchers in CSCW/HCI have conducted extensive studies on informal caregiving and developed technologies to facilitate coordination among home caregivers for various groups, including the elderly [6,14,31], hospitalized patients [56], depressed family members [89], and children/young adults [32,33].However, our study has identified three complexities related to CGM that pose significant challenges to coordinating informal caregivers for teenagers with T1D.
The first complexity we identified is that informal caregivers for teenagers with T1D, when connected to the CGM system, must collaborate synchronously.This means that they receive the same glucose readings and alarms simultaneously, making it challenging for caregivers to communicate and keep track of who is currently responsible for the teenager's care.This is different from caregiving for the elderly, hospitalized patients, and depressed family members, where caregivers collaborate asynchronously and usually only one caregiver needs to be present at a time [6,14,31,56,89].We found that synchronous collaboration often results in multiple caregivers responding to the same alarm, creating a chaotic situation for the teenager and potentially leading to conflicts between the caregivers and the teenager.Moreover, there were instances where no caregiver responded to critical alarms due to the distribution of responsibilities.This ineffective collaboration could potentially cost the teenager's life.
Secondly, caregivers for teenagers with T1D require much shorter response times for communication.Previous designs for coordinating caregiving [33,89] have mainly focused on tracking and reflecting on symptoms, illness experiences, and caregiving activities, with a cycle of tracking and reflection that could last for days or even weeks.However, our study revealed that caregivers often need to make treatment decisions within a few seconds or minutes after a CGM alarm goes off.This demands timely access to information, faster communication, and better coordination among caregivers who may be working remotely.
Thirdly, previous studies on caregiver coordination have mainly focused on a single care context, such as care for elderly people or children at home, or care for patients in the hospital.However, our study, similar to Raj et al. [67], found that teenagers with T1D often move across different contexts, such as school, gym, work, home, summer camp, playdates, and sleepovers.Each context requires unique tracking and care needs.For instance, while parents can assist with treatment decisions when at home, they need to grant CGM access and collaborate remotely with other caregivers, such as school nurses, camp coaches, friends' parents, and other relatives when the teenagers are at school, summer camp, or a sleepover, or when the parents are traveling.Therefore, caregivers must frequently and sometimes passively transition caregiving roles and adjust their coordination modes to accommodate teenagers' changing contextual settings.
These complexities emphasize the importance of developing CGM technologies that can facilitate both synchronous and asynchronous communication among informal caregivers for effective coordination.To support asynchronous communication, it is crucial to incorporate features that prevent duplicated efforts and ensure timely responses to critical alarms.This can be achieved by implementing a rotation schedule where each caregiver has a specific window of time to monitor the teenager's CGM readings and respond to alarms.Additionally, the design should enable clear communication and awareness of which caregiver is currently responsible for the teenager's care.This can be achieved through a shared dashboard or interface that displays the caregiver's assignment and an alert system that notifies them when a task has been assigned or completed.Furthermore, the design of future CGM technology should enable faster communication during passive transitions in caregiving roles to ensure that there are no delays in managing the teenager's CGM readings.This can be achieved by incorporating real-time communication channels that allow caregivers to quickly exchange information and coordinate handoffs.Alarm prioritization and escalation protocols could also be helpful to ensure that critical alarms are not missed and addressed promptly.
In addition to the above, it is essential for future CGM technology design to incorporate features that facilitate Asynchronous communication among informal caregivers.These features should enable effective tracking and reflection on symptoms, illness experiences, and caregiving activities with the ability to share information across multiple contexts and caregivers.Moreover, the design should provide adequate support to caregivers with different levels of knowledge to ensure that they can comfortably use the technology and respond appropriately to different CGM alarms and notifications following the standards of care.This could be achieved by incorporating step-by-step checklists, tutorials, and recommendations for treatment options in the design.
In summary, previous CSCW/HCI literature on T1D has mainly focused on individual caregiverteenager collaboration, but our study highlights the lack of technological support for coordinating multiple caregivers who simultaneously monitor the CGM data of teenagers with T1D.We identified three complexities related to CGM that pose significant challenges to coordinating informal caregivers for teenagers with T1D: synchronous collaboration, short response times for communication, and frequent transitions in caregiving roles across different contexts.Future CGM technology designs should incorporate features that facilitate both synchronous and asynchronous communication among informal caregivers, prevent duplicated efforts, ensure timely responses to critical alarms, enable clear communication and awareness of caregiver responsibilities, and provide adequate support to caregivers with different levels of knowledge.

Design to Support Caregiver-Teenager Collaboration
The findings of our study indicate that the current continuous glucose monitoring (CGM) systems that share glucose readings, trends, and alarms between caregivers and teenagers are insufficient to support their collaboration effectively.This is because there are significant contextual details, such as food intake, medication, pump site connection, stress level, CGM site, and exercise, that must be communicated alongside the CGM data to enable caregivers and teenagers to understand the readings and alarms and make informed treatment decisions.Unfortunately, caregivers and teenagers currently rely on text messages and phone calls to communicate this important contextual information, which can be inefficient and result in treatment delays.Studies [35,36,63] have previously identified several contextual details that are critical for interpreting blood glucose levels, and we recommend that future CGM system designs incorporate the sharing of this information alongside the CGM data to enhance the support for caregiver-teenager collaboration.
Furthermore, our study revealed that the extent of contextual information that must be exchanged between caregivers and teenagers is contingent on the teenagers' level of knowledge and capability for self-care.For example, a 19-year-old might easily recognize that her menstrual cycle is the main cause of her elevated blood glucose levels.In this case, sharing just two pieces of contextual information -"MC" and "current basal rate" -with her caregiver would be enough for them to make informed decisions on necessary basal changes to reduce her blood sugar levels collaboratively.However, a 14-year-old may not possess the knowledge or experience to independently identify the cause of high blood glucose levels.Thus, a more substantial sharing of contextual information between the teenager and her caregiver would be necessary to thoroughly explore and rule out all potential causes.
Therefore, we suggest that future CGM system design should prioritize tailored contextual information sharing that accounts for the age and experience levels of teenagers.This can be achieved by evaluating individual teenagers' knowledge and experience with self-management [11] and providing appropriate levels of information sharing accordingly.For younger and less experienced users, CGM systems should provide more comprehensive sharing of contextual information, including meal and snack details, medication doses and timing, physical activity levels, and other relevant details.To ensure ease of use, user-friendly interfaces and intuitive data input methods, such as voice commands or image recognition, could be incorporated.For older and more experienced users, CGM systems can provide more streamlined sharing of contextual information, with a focus on essential details that are critical for making informed treatment decisions.

Design to Incorporate Social Cues and Signals
The use of Continuous Glucose Monitoring (CGM) allows for remote collaboration between teenagers and their caregivers by sharing real-time blood glucose information with caregivers.Despite the benefits such as increased freedom and peace of mind, our study found that the remote collaboration is less effective and more prone to conflicts among contributors compared to traditional, in-person care.The lack of visibility of social cues and signals in CGM data is a major cause of this difference.During face-to-face collaboration, caregivers can easily notice social cues and signals such as a teenager checking the CGM readings, doing finger pricks, and administering carbs or insulin.Similarly, teenagers can rely on caregivers for assistance because they can see that they are attending to their CGM readings and alarms.However, when CGM data is used for remote collaboration, these social cues and signals of caregivers and teenagers are not visible to each other, leading to frequent communication through phone calls or text messages to ensure mutual awareness of readings, alarms, and treatments.Our findings suggest that this can cause conflict between caregivers and teenagers due to excessive communication that can be perceived as annoying, which can ultimately lead to feelings of mistrust and strain on the collaborative relationship.
To mitigate the conflicts that arise from the lack of social cues and signals in caregiver-teenager collaboration, we recommend that future CGM designs include social translucence features [21,53,54] that make contributors' actions more translucent to each other and build awareness of the social cues and signals.For instance, a social translucence feature could include a log of actions taken by both the caregiver and teenager, allowing them to see when the other person checked their readings, acknowledged alarm or administered treatment.Another potential feature could be the ability for the caregiver and teenager to send short messages to each other through the CGM app, providing real-time, quick check-ins and reducing the need for frequent phone calls or text messages.
By incorporating social translucence features into CGM system, Future designs can enhance remote collaboration and create a more seamless and effective system.This will not only improve the quality of care for teenagers with diabetes but also ease the burden on caregivers.

Limitations
We acknowledge several limitations in our study.First, all families participating in the study had access to costly CGM devices and have a set of caregivers contributing to the teenagers' diabetes tracking and management.We suspect that this makes our sample skewed toward families with higher income levels and technology literacy, although those specific demographic variables were not part of our data collection.Future studies should focus on families with fewer social, financial, and technical resources to investigate their tracking needs.Second, we were not able to include medical providers in our ARCs.Medical providers, physicians, nurse practitioners, pharmacists and others, are another important group of potential CGM users who collaborate with the families in data interpretation and treatment decision-making.Future work should consider this user group to cover a broader picture of collaborations around teenagers' diabetes control.Third, although we aimed to recruit for a diversity of family arrangements, we were only able to enroll women caregivers in two-parent, heteronormative households.Further work should explore CGM usage in additional family arrangements.Finally, there might be a potential for participants to influence each other after sharing their experiences within their perspective ARCs.We believe these influences should have minimal impact on participants' sharing their lived experiences, as many already participated in online T1D groups.

CONCLUSION
In this paper, we present an analysis of 16 families' lived experience of using a bio-sensing technology to manage teenagers' complex chronic conditions collaboratively.The study shows an exciting potential of expanding personal tracking to support families' collaborative caregiving.Our work contributes an empirical understanding of how families with teens navigate the care work in the context of CGM and the issues they encountered in their day to day collaboration.We believe the findings from this research can help inform the design of new family-based health informatics tools and contribute to our understanding of the caregiving work created by the specific design choices for these tools.

Fig. 1 .
Fig. 1.Dexcom G6 system.The monitor displays the current glucose reading, a trend arrow indicating the current direction and rate of change of the glucose level, and a trend graph showing the trend of glucose levels over the past several hours.

Fig. 2 .
Fig. 2. Examples of Dexcom G6 system alarms: From left to right, the examples illustrate an urgent low soon alarm, an urgent low alarm, a low glucose alarm (top), and a high glucose alarm (bottom).

Table 2 . 1 2 3 4 5
List of activities including the activity number and topicActivity TopicData type Introduction and icebreaker.Participants introduced themselves, their diabetes management routine and CGM usage, and expectations for CGM.Text Task-based scenario.Participants were asked to provide peer advice for using CGM in a family.Text Discussion on setting and responding to alarms.Participants were asked to talk about their current alarm settings, when they feel the alarms are helpful/unhelpful, and how/for what they adjust the alarm settings.Text Discussion on remote tracking.Participants were asked to identify scenarios in which parental tracking are necessary/unnecessary and provide examples.Text Discussion on making treatment decisions with CGM data.