What End Users and StakeholdersWant From Automated InsulinDelivery SystemsDiabetes Care 2017;40:1453–1461 | https://doi.org/10.2337/dc17-0400

OBJECTIVE

The purpose of this study was to rigorously explore psychosocial factors associatedwith automated insulin delivery systems among people living with type 1 diabetes.

RESEARCH DESIGN AND METHODS

Across four sites in the U.S. and U.K., 284 participants completed structured inter-views or focus groups on expectations, desired features, potential benefits, andperceived burdens of automated insulin delivery systems. Recorded audio files weretranscribed and analyzed using NVivo.

RESULTS

Three themes were identified as critical for uptake of automated insulin delivery:considerations of trust and control, system features, and concerns and barriers toadoption. Children and adolescents with type 1 diabetes primarily identified needsspecific to their life stage and social contexts (e.g., school). Adults with type 1diabetes, parents of youth with type 1 diabetes, and partners of adults with type 1diabetes weremost concerned about the accuracy, adaptability, and algorithm qual-ity alongside expectations that systems stabilize glucose levels and reduce risk forlong-term complications.

CONCLUSIONS

Incorporating stakeholder perspectives onuse of automated insulin delivery systemswill improve the adoption of devices, quality of life, and likelihood of optimal health.Efforts to build trust in systems, optimize user-system interactions, and provide clearguidance about device capabilities and limitations may help potential users achieveoptimal glycemic outcomes.

Automated insulin delivery systems represent an innovation that will change thelandscape of diabetes self-management by combining three previously distinct ele-ments of diabetes care: single (insulin) or dual (insulin and glucagon) hormone deliveryvia pump(s), continuous glucose monitoring (CGM), and predictive algorithms that useglucose data to generate automated decisions about hormone timing and doses. In theU.S., 40–62% of adults with type 1 diabetes use pump therapy and;21%of individualsabove age 25 years use CGM, with rates of CGM use increasing as device performancehas improved and nonadjunctive indication has been realized (1–6). Thus, there arereadyusers and otherswith type 1 diabetes not using diabetes deviceswhomayopt forautomated insulin delivery if there are glycemic and quality of life benefits. However,people with type 1 diabetes who initiate technologically advanced alternatives orsupplements to traditional insulin injections frequently report discontinuous or

1Department of Pediatrics, Stanford UniversitySchool of Medicine, Stanford, CA2Department of Social and Behavioral Sciences,University of California, San Francisco, San Fran-cisco, CA3Bournemouth University, Bournemouth, U.K.4Department of Psychiatry and Behavioral Sci-ences, Ann & Robert H. Lurie Children’s Hospitalof Chicago, Northwestern University FeinbergSchool of Medicine, Chicago, IL5Joslin Diabetes Center, Harvard Medical School,Boston, MA

Corresponding author: Korey K. Hood, kkhood@stanford.edu.

Received 24 February 2017 and accepted 23 July2017.

© 2017 by the American Diabetes Association.Readers may use this article as long as the workis properly cited, the use is educational and notfor profit, and the work is not altered. More infor-mation isavailableathttp://www.diabetesjournals.org/content/license.

Diana Naranjo,1 Sakinah C. Suttiratana,1,2

Esti Iturralde,1 Katharine D. Barnard,3

Jill Weissberg-Benchell,4 Lori Laffel,5 and

Korey K. Hood1

Diabetes Care Volume 40, November 2017 1453

CLIN

CARE/ED

UCATIO

N/N

UTR

ITION/PSYC

HOSO

CIAL

interrupted use of these devices (1–5).Automated insulin delivery systemsrepresent a promising frontier in themanagement of type 1 diabetes; thisstudy provides preliminary data to helpresearchers and clinicians understandand address the lack of interest in tech-nological advancements aswell as barriersto initiating and maintaining consistentuse of such systems (7).Seeking to optimize benefits for auto-

mated insulin delivery system users andresources for regulatory bodies and tech-nology developers, our research sought totranslate the experiences and expectationsof selected patient groups into tailoredmeasures to help assess type 1 diabetesmanagement experiences in using thesesystems. As part of a multisite study toidentify and describe perspectives on psy-chosocial factors relevant to automated in-sulin delivery from various stakeholders,researchers used qualitative researchmethods to understand patient and familyexpectations, attitudes, emotions, and ex-periences with automated insulin delivery.The purpose of the study was to deliverexperience-based findings to inform medi-cal device regulatory reviews and the de-velopment of validated psychosocialmeasuresfor use during imminent clinical trials by var-ious device developers.

RESEARCH DESIGN AND METHODS

Qualitative methods were selected tocapture a broad range of factors describinghow diabetes management devices fit intothe lives of people with diabetes (PWD)and why diabetes management devicesmight be initiated, continued, or discontin-ued. Qualitative methods are most appro-priate for developing more completeunderstandings of phenomena throughdepth as opposed to breadth and for gen-erating newmeasures (8). The goal of thisresearch was to identify and exploretopics related to interest in, uptake of,sustained use of, discontinuance of, andconcerns about automated insulin deliv-ery systems. The depth offered by thistype of study seemed most appropriatefor designing new survey measures to ac-company the rise of automated insulindelivery systems. Findings will also beused to help communicate realistic ex-pectations about emerging systems topotential users. Interviews were con-ducted January 2015 to December 2015,and data were analyzed December 2015to April 2016.

Sampling and RecruitmentIn order to learn from a variety of experi-ences and capitalize on existing researchefforts, this study was implemented byfour research teams in the U.S. (JoslinDiabetes Center, Lurie Children’s Hospital,and Stanford Children’s Health) and theU.K. (Bournemouth University). Researchteams included psychologists, diabetesphysicians, nurses, diabetes educators,and representatives from diabetes advo-cacy and family support organizations.Five patient populations were sampled:children, adolescents, and adults diagnosedwith type 1 diabetes, as well as parentsand partners of PWD. Participants wererecruited using existing type 1 diabetes re-search networks and infrastructure (e.g.,clinics, clinical trials, and support groups)with access to PWD with and withoutautomated insulin delivery knowledge orexperience. Across all four sites, every ef-fort was made to recruit diverse partici-pants with respect to socioeconomicbackground, sex, duration of diabetes,and technology use. In addition, weattempted to enroll equal numbers ofparticipants based on the five patientpopulations sampled. Common recruit-ment strategies across sites included ad-vertisements on social media and postedflyers at pediatric and adult clinics whereparticipants could self-select by respond-ing to the advertisement. Two sites of-fered either semistructured interviewsor focus groups (Stanford Children’sHealth and Bournemouth University). AtStanford Children’s Health, patients wereassigned to the interview modality thatworked best in their schedules, and atBournemouth University, patients weregiven the choice between individual andfocus groups based on personal prefer-ence. Joslin Diabetes Center only offeredsemistructured interviews, whereas LurieChildren’s Hospital offered only focusgroups; at both sites, researchers usedelectronic medical records to identifyeligible patients, who were later ap-proached in clinic, with the exception ofthe adults recruited at Lurie Children’sHospital, where they only used social me-dia and posted flyers in the hospital. Par-ents and partners were recruited throughthe PWD. Research protocols were ap-proved by ethical review boards at eachsite, and participants received $100 at U.S.sites and £20 at the U.K. site for their par-ticipation. Prior to their participation,adults and parents provided consent for

study participation; youth additionallyprovided assent.

Data Collection and Analysis PlanInterview and focus group questions ex-plored the following themes: expecta-tions, hopes and anxieties, perceivedbenefits and barriers, impact on dailyfunction, self-management, health careinteractions, and costs associated withautomated insulin delivery and consider-ations of comfort, wearability, and user-technology interfaces. Sample questionsfrom research discussions included thefollowing: 1) “What would be some ofthe tasks that would be involved in usingan automated insulin delivery system?,”2) “What are some of the possible bene-fits from the system?,” 3) “What are yourexpectations about what the systemmight do?,” 4) “Are there any aspectsof automated insulin delivery systemsthat you think might hurt your diabetesmanagement or worry you?,” 5) “Arethere particular times of day or situationswhen you might find an automated insu-lin delivery system particularly useful?,”and 6) “What would stop you from want-ing to try or use one of these systems orwhat might get in the way?” Data on ex-periences and expectations of both singlehormone and dual hormone systemswere collected; we refer to both typesof systems as automated insulin deliverysystems for simplicity. Findings were in-tended to help develop empiricallygrounded, self-administered quantitativemeasures to assess characteristics associ-ated with effective initiation and use ofautomated insulin delivery systems acrossage groups and family stakeholders. Investi-gators used hypotheses about themes thatmight emerge to identify 24 codes prior todata analysis (Fig. 1, step 1, and Fig. 2).

Focus groups and interviews were ledby facilitators who received advancedtraining on conductingmixed-methods re-search. Selected demographic data andnotes on the interview/focus group ses-sions were recorded by facilitators. Dataanalysis and measure developmentprocedures were divided into four steps(summarized in Fig. 1). In step 1, all focusgroups and interviews were audiore-corded and audio data were transcribedby an independent, Health Insurance Por-tability and Accountability Act (HIPAA)-compliant service provider (Medikin,New York, NY). Transcripts were reviewed,de-identified, and uploaded to NVivo

1454 End Users and Automated Insulin Delivery Systems Diabetes Care Volume 40, November 2017

software version 11.2 for analysis (QSR In-ternational Pty Ltd.) (9).Next, transcripts were analyzed using

thematic analysis based around the 24codes identified by investigators a priori.Thematic analysis offers aflexiblemethodfor identifying, analyzing, and presentingpatterns across qualitative researchmethods(10). Beyond “giving voice” to study par-ticipants, our thematic analysis acknowl-edged the expertise and experiences ofour researchers (10).Over 9 weeks, nine coders represent-

ing all four sites reviewed and assignedcodes to data from 137 transcripts con-sisting of.3,500 pages of data. Codes arewords or phrases that succinctly capturewhat a segment of text is about. The cod-ing process consisted of multiple stepsdesigned to support the replicable identi-fication of themes and quality monitoringof data and analytic products. A qualita-tive data coordinator worked with siteteams to ensure that initial coding com-bined a priori codes as well as emer-gent codes identified by coders as theyread and analyzed transcripts. One tran-script was coded by all coders and dis-cussed using a consensus process tobuild similar understandings of conceptsand themes before coders were assignedgroups of transcripts for independentcoding. Transcripts were assigned tocoders by thequalitativedata coordinatorbased on transcript length, transcript va-riety, and coder availability. Coders metweekly via HIPAA-compliant BlueJeansVideo Conferencing to discuss criteriaused in selection and nonselection ofcodes as well as challenges encountered

when analyzing transcripts. At each stageof analysis, coders were able to high-light memorable quotations and outliersor recommend alternative language orinterpretations of data. Emergent codeswere grouped into clusters with related apriori codes or set aside for future analy-ses. Attitudes toward technology in soci-ety, rather than diabetes technology inparticular, is one example of an emergentcode; some data relevant to this codewere captured under the thematic clusternamed “technological and technical as-pects of automated insulin delivery sys-tems,” other aspects of that code werenot. Twenty-five percent of transcriptswere randomly selected for double cod-ing. In order to quantitatively and quali-tatively compare interrater agreement,pairs of coders were randomly selectedeach week from among those assigneddouble-coded transcripts. Rather thanseeking statistical guideposts, researchersrelied on levels of agreement and consen-sual identification of themes for in-depthanalysis. Measures developed from thisresearch would later be subjected to cog-nitive interviews and validity testing. Bythe 4th of 9 weeks of coding, no newthemes arose and thematic saturationwas achieved around the research ques-tions of interest.

Next, in step 2, 24 a priori codes aswell as emergent codes were condensedto 12 thematic clusters (see Fig. 2). Clus-ters comprised closely related codesgrouped under thematic headings de-rived from iterative research dialogues.Two additional analytic methods wereused. First, coded data were synthesized

into cluster-specific “idea units” to repre-sent the range of ideas discussed by eachof the participant groups (children, ado-lescents, and adults with type 1 diabetesand their parents andpartners). Second, adata matrix was designed to distill analy-ses into possible survey items for themeasures to be developed. In step 3,summaries of analyses were presentedto the entire research team during aface-to-face study meeting. Integration ofqualitative results and expertise with PWDenabled the team to harness the power ofparticipants’ narrative and meaning whileensuring that distillations for quantitativeitems did notmiss critical factors influenc-ing the use of diabetes managementdevices. Further, data volume contributedto both theoretical saturation and cross-participant verification of themes. Step 4consisted of measure developmentand piloting and is not discussed in thismanuscript.

RESULTS

Demographic Characteristics ofSampleTable 1 summarizes demographic charac-teristics of participants and data sources.Interview and focus group data were ob-tained from 284 participants across thefour sites. Participants were predomi-nantly white, non-Hispanic (92.0%) and75.0% reported current pump use forthe person diagnosed with type 1 diabe-tes. Race/ethnicity andpump informationwere not available for children and teensat all sites.

Twelve thematic clusters representedthe majority of data analyzed and areshown in Fig. 2. Current analyses focuson threeprevailing themes: trust and con-trol, features, and concerns and trade-offs. These themes were most discussedthroughout the data across stakeholdersubgroups; differences noted betweensubgroups were primarily related to em-phasis. The trust and control theme wasfocused on user-system interaction andcontained subtopics related to user con-trol versus automation and thoughtsabout taking breaks from technologyover long-term diabetes management.The features cluster included preidenti-fied concepts related to potential fea-tures of automated systems, as well asfeatures identified by the participants asdesirable. Concerns and trade-offs delib-erated changes in self-management bur-den, the adaptability/flexibility of these

Figure 1—The data analysis and subsequent measure development procedures implemented forthis study. Four steps and nine substeps have been identified in the sequential list.

care.diabetesjournals.org Naranjo and Associates 1455

systems, as well as the socially embeddednature of type 1 diabetes managementand device use. Quotes representingeach of these themes are provided inTable 2.Overall, negative themes such as de-

vice concerns and areas of uncertaintywere more endorsed than positivethemes such as device benefits. Ideasabout potential benefits of automated in-sulin delivery systems in specific situa-tions, relationships, or contexts werealso mentioned frequently, such as night-time and eating routines. Emphasis on

the negative or unknown in participantexpressions may reflect both the devel-opmental stage of the systems as well aslow levels of knowledge about such sys-tems among themajority of participantsinterviewed.

Trust and ControlTrust and control data discuss user-system interactions, including one’sability to trust systems to operate effec-tively, suspend and restart usage, over-ride system miscalculations, and sharesystem data with caregivers. All groups

mentioned difficulties they might havetrusting new systems and questionsabout whether the system or the userwould have ultimate decision-makingauthority.

In order to consider using automatedinsulin delivery systems, participantsstated that systems had to be trusted todo what they “promise” to do (eliminatehigh and low blood glucose variabilitywhile subsequently reducing risks of com-plications). Participants desired systemsthat would diminish the daily prominenceof type 1 diabetes self-care activities and

Figure 2—The condensing of 24 investigator-identified a priori codes into 12 thematic clusters deemed most salient by a multistep,qualitative analysis and consensus process. *Eating and exercise/physical activity were related to several clusters. Trade-off analysis (#24)coding was used for other analytical purposes and is not represented in the 12 clusters here. Clusters discussed in this manuscript are in boldprint.

1456 End Users and Automated Insulin Delivery Systems Diabetes Care Volume 40, November 2017

recognized that automated insulin deliv-ery systemsmight be able to relieve someof the burden of type 1 diabetes manage-ment. Further, approximately half of thetranscripts contained discussion abouthow automated insulin delivery sys-tems might reduce human error in self-management. Automated insulin deliverysystems were perceived as being able to

stabilize glycemic control while improv-ing quality of life. More specifically, par-ticipants indicated that systems shouldbe automatic, accurate, efficient, andadaptive and should be associated withimproved mood, sleep, relationships,and greater normalcy of eating andexercising behaviors. Finally, participantnarratives described automated insulin

delivery systems that could lead to reduc-tions in monitoring, stress, conflict,hospitalizations, and future complications.

Reflecting a common uncertainty abouthow much trust one could place in auto-mated insulin delivery systems and theiralgorithms, PWD of all ages and parentswanted to be able to override decision-making functions of the systems.Other par-ticipantswere cautious about trusting auto-mated insulin delivery systems in general,and early generation systems in particular,due to their novelty. Despite skepticismabout the functioning of various systemcomponents, most participants were hope-ful about the advancement of automatedinsulin delivery systems and the future ofdiabetesdevices. Participants expected thatdevelopers would make efforts to builduser trust and confidence but believedthat part of the incremental research anddevelopment process meant that new de-vices would still entail new learning and aperiod of adjustment for PWD and theirloved ones.

In terms of subgroup emphasis, stake-holders discussed their beliefs and con-cerns about system trust and controldifferently. Children were more con-cerned with how automated insulin de-livery systems might affect specific socialsituations, school, and time with friends;they hoped that system adaptabilitywould ease these concerns. Adults withtype 1 diabetes and parents of youthwithtype 1 diabetes weremore worried aboutdevice safety than others. Similarly, part-ners raised each of the aforementionedtrust and control themes as well as theirpartners’ past experiences with diabetes-related technology. Participants experi-enced with advanced type 1 diabetesdevices such as pumps and CGM weremore likely to presage the potential ben-efits of using automated delivery systemsthan people who followed lower techmanagement regimens. “High-tech” par-ticipants discussed trust and control,physical features, and past experiences(positive and negative) with technologymore than their “low-tech” counterparts.

In sum, participants described a rangeof expectations that would help themtrust the automated insulin delivery sys-tems and feel comfortable with the pro-cess of controlling such systems.

FeaturesPotential system features were of greatinterest to study participants. Data

Table 1—Characteristics of participants and qualitative data sources

Descriptive characteristics of participants and data sources Value (range)

Overall number of focus groups 48

Overall number of semistructured interviews 89

Adults, n 113Data from semistructured interview 31.0Age, years 39.5 (18–77)Female 70.8Race/ethnicityBlack/African American 1.8Hispanic/Latino 0.9Asian/Pacific Islander American 3.5White, non-Hispanic 92.0Other 0.9

Bachelor’s degree or higher education 73.5Current pump use 72.6Current CGM use 54.5Hemoglobin A1c 7.5% (5.0–11.8%)

Adolescent/young adult with type 1 diabetes, n 35Data from semistructured interview 45.7Age, years 14.7 (12–20.8)

Children with type 1 diabetes, n 16Data from semistructured interview 43.8Age, years 10.3 (9–11)

Parents, n 65Data from semistructured interview 25.1Relationship to childMothers 79.7Fathers 17.2Other 1.5

Responding for female child 61.7Child’s race/ethnicityBlack/African American 1.5Hispanic/Latino 5.3Asian/Pacific Islander American 0.0White, non-Hispanic 89.9Other racial group 3.3

Child’s current pump use 71.8Child’s current CGM use 53.8Child’s hemoglobin A1c 8.1% (6.4–13.0%)

Partners of people with type 1 diabetes, n 55Data from semistructured interview 20.0PWD race/ethnicityBlack/African American 0.0Hispanic/Latino 1.8Asian/Pacific Islander American 1.9White, non-Hispanic 94.5Other racial group 1.9

PWD current pump use 83.6PWD current CGM use 74.1PWD hemoglobin A1c 6.9% (5.0–9.2%)

Total participants, n 284

Values are presented as percentage or mean (range) unless otherwise indicated.

care.diabetesjournals.org Naranjo and Associates 1457

about features emerged in a qualita-tively different manner than otherthemes; participants listed desired phys-ical, mechanical, or technical featuresand described how the identifiedfeature(s) could make a difference intheir daily lives. Seven categories of fea-tures were identified: features borrowedfrom prior technologies, physical fea-tures, comfort and wearability, the algo-rithm and computing architecture, userinterface features, device educationand support programs, and other fea-tures. First and foremost, participantswanted a device they could “set and for-get.” Participants thought automatedinsulin delivery systems should approxi-mate their understanding of a biologicpancreas.Various configurations and compo-

nents of automated insulin delivery sys-tems were envisioned, such as oneadolescent’s belief that systems shouldsimply have “glucagon, insulin, CGM, notubes, [and be] waterproof.” In addi-tion to preferring wireless or tubelessand waterproof devices, the majorityof participants worried about the size,weight, and appearance of possibledevices. Other participants describedthe neglected importance of aestheticsin the development of type 1 diabetesdevices; the ever-present nature ofthese devices makes them part of the“daily uniform” of PWD. Children andteenswanted devices that came in colors,could visually function as extensions oftheir personalities, and were built ontheir generation’s immersion in tech-nological devices and applications. Ado-lescents spent a great deal of theirinterviews/focus groups discussing theirconcerns about the physical features,wearability, and comfort of potentialsystems. Finally, some parents sought de-vices to help lessen the sense of differ-ence that their child experiences; manyparents believed that less visible andless intrusive devices would improvetheir children’s self-esteem and socializ-ing. This example demonstrates howcommon interests in less visible systemsmay stem from different beliefs (childrenand teens also wanted less visible sys-tems, but so that they could control towhom their diagnosis was revealed).Youth were not explicitly concernedwith enhancing self-esteem but sharedparents’ interest in systems that mayhelp improve socializing.

Table 2—Sample participant data for three emergent themes

Exemplar quotes representing three major themes

Trust and control “I think that’s a critical part of a closed-loop system. . .theaccuracy and dependability. And there are a lot of piecesthat can fail. Just infusion sites are sodone day they workand the next they don’t. So, there are a lot of human factorsin there. I think that a lot of those could be alleviated, butthat’swhat I worry about is if they don’t address at least theones that are the most critical, then I think the systems justwon’t be accurate. You won’t have faith in them. You won’tbe able to trust them. And then, you probably spent a lot ofmoney on something that’s just not working.” (Adult withtype 1 diabetes)

“If I was able to take over, manually, the action of the pump,then I wouldn’t feel worried. The worry is letting it workautomatically and it not doing it correctly and I go too lowortoo high, particularly at nighttime if I’masleep and it makesthe wrong decision in terms of a dose. That would be aworry.” (Adult with type 1 diabetes)

“Having seen somedata nowandbeing exposed to it a little bitmore like actually seeing those nondiabetic blood sugarson the screen and showing patient after patient afterpatient that are on these systems that have zero variationand theyall stay betweenmaybe 80 and120 for someof themore advanced studies, I am a believer now. I am a closed-loop convert. I can’t wait to not have to make thosedecisions.” (Adult with type 1 diabetes)

“Inevitably with any new type of medical treatment ortechnology, there is going to be a bit of a teething periodwhere you are just kind of getting your head around how itworks.” (Adult with type 1 diabetes)

“My ultimate aim would be to forget that I have diabetes inthe first place. [Other participants in the room agree.] Thathas to be everybody’s dream that we become whatever iscalled normal...that is my ultimate expectation within mylifetime. I don’t think they’ll find a cure but I would like toforget that I’ve got it.” (Adult with type 1 diabetes whowanted to trust a new system)

Features “Would it be great to have a system that you just plug it in andjust live your life and you don’t have to do your fingerpricksand you don’t have to worry about telling it what you’regoing to do because it activates and it makes decisions onthe spot?” (Adult with type 1 diabetes)

“Well, this is not scientific, but sort of what [automated insulindelivery]means tome iswith state-of-the-art technology totry to imitate what the pancreas does for the human bodyby using several devices and/or drugs in combination sothat a person with type 1 diabetes can live a normal life.”(Parent of youth with type 1 diabetes)

“I kind of think it would be quite big, especially if there werewires and stuff connecting the pump to the cannula, andyou have a sensor and a pump and the other gadget to carryaround. I kind of think itwould bequite heavy.” (Adolescentwith type 1 diabetes)

“Is it discreet? Is it bulky? Is it going to draw attention to itselfwith alarms and things like that?” (Parent of youth withtype 1 diabetes)

[Device companies] “forget that it is old-school looking. . .thisugly device [becomes] part of our wardrobe, it is part of ourdailywear. Forme, especially, as a femalewho dresses up inbusiness wear every day, I want something that is attractiveonmybody and I think sometimes that gets forgotten in thediscussion.” (Adult with type 1 diabetes)

Continued on p. 1459

1458 End Users and Automated Insulin Delivery Systems Diabetes Care Volume 40, November 2017

ConcernsParticipants were most concernedabout two areas of uncertainty and theability of automated insulin delivery sys-tems to diminish social challenges relatedto type 1 diabetes and its management.First, participants were uncertain aboutthe types and intensity of changes inman-agement burden that would result fromusing automated insulin delivery systems.Would the management burden bereduced? Or would burden simply beshifted so that users were now responsi-ble for different, technology-orientedtasks such as device charging or mainte-nance? Lack of knowledge about auto-mated insulin delivery systems andwhat they will require from users raisedanxiety and wariness among partici-pants. For the majority of participants,the thought of possibly increasing and

not decreasing their self-managementburden seemed untenable. Even amongthose with “out of control” blood glucoselevels, status quo equilibria have beenachieved through investments of timeand effort. Quality of life or health improve-ments are critical for PWD to embrace anew system and way of managing theirdiabetes.

Similar uncertainty existed about theadaptability and customizability of auto-mated insulin delivery systems. Partici-pants wanted systems that could adaptto the specifics of their lives. Descriptionsof personalization preferences for sys-tems included the ability to be temporar-ily removed, devices that “learn” fromone’s habits and routines over time, andcustomizable alarms and sounds. Partici-pant interest in customization parallelssocietal advancement of personalized

medicine and routinization of on-demandaccess to technological devices. Socialtrends in personal health managementthrough data quantification and monitor-ing were also reflected in participantinterest in using glucose data to improveinsights about their own self-management.Despite these developments, teens andchildren described potential prohibitionsagainst devices in some settings. Youthmentioned rules against insulin pumpsin some sports and against phones insome schools. For school-aged youth, sys-tems must be recognizable medical de-vices clearly distinguishable from othertechnological devices, and yet many chil-dren expressed a desire for the technolo-gies to blend in and not alert others totheir diabetes. When type 1 diabetes de-vices are suitable for individuals’ distinc-tive lifestyles, participants believed thatPWD would be more inclined to use thedevices.

Study participants across subgroupswanted devices that would improvesocial and situational aspects of diabe-tes management. Participants exten-sively discussed how self-managementcontributed to stress, conflict, worry,and dependency in their relationshipsand interfered with many situations andactivities of daily life. Their hope wasthat a system could automate, replace,or postpone some of the managementtasks and help them regain somesense of normalcy; yet, they were con-cerned that the promise of auto-mated insulin delivery systems mightbe exaggerated.

CONCLUSIONS

Our research reveals multiple psychoso-cial factors that influence initial andpersistent use of automated insulindelivery systems. Our findings expandprevious research findings about psycho-social factors in medical device use(11–14) and offer new insights. First, psy-chosocial factors are influential in PWD’sconsiderations of changes to their self-management regimens. Participantsseek technical innovations that will im-prove blood glucose stability and reducethe likelihood of long-term complicationswithout increasing their management-related burden. Releasing clear expecta-tions and potential benefits of systems,achieving high levels of satisfaction in pub-lished trials, and including technical sup-port and fail-safes that place system

Table 2—Continued

Exemplar quotes representing three major themes

Concerns and trade-offs “If it means less interaction and better control, yeah. If itmeans more interaction and better control, no. If it meansthe same interaction and the same control, no. Whychange? Why learn a new system?” (Adult with type 1diabetes)

“If I had to carb count, I don’t know if that’s really alleviating it,because it’s themental stuff that I don’twant todeal. I don’treally care about the physical stuff as much. I’ll wear it, I’llchange it, and whatever. It’s the mental stuff I am tired ofdoing.” (Adult with type 1 diabetes)

“I don’twant tobewearing tendifferent thingsonme. It’s kindof like already burdensome as it is.” (Adult with type 1diabetes)

“Well, it should be customizable. You should be able to choosewhat your alarm of choice is and how often you want it. Itneeds to be responsive todit needs to be customizable forthe user so they can set it up the way they want. I mean acell phone is. This should be also.” (Adult with type 1diabetes)

“Well, I think that varying blood sugars and also the fact thatwe carry this burden and live so close to death, I think thathas a lot of effect on the kind of person that you are andhow you relate to other people and the highs and the lowsand your moods and that sort of stuff. So I think the betterblood sugars that you can have then the less weight thatyou have to carry on your shoulders every day. So I thinkthat perhaps that could have a positive effect onrelationships with other people because you just are nothaving that burden all the time.” (Adult with type 1diabetes)

“I became obsessed with the idea of being normal again andhaving something thatwouldmake you normal to an extentwhere you feel normal and I mean what is normalobviously; but being in a situation where you don’t have toworry about everything, being in a situation where youdon’t have to carb count, being in a situationwhere you cango out with your mates and not really care or just have it inthe back of your mind or something.” (Adult with type 1diabetes)

care.diabetesjournals.org Naranjo and Associates 1459

control in users’ handsmayhelp engendertrust.Second, device features figure promi-

nently in users’ daily experiences withtype 1 diabetes devices. Possible systemfeatures were related to participants’trust, expectations, and concerns. Fea-tures like being wireless and waterproofwere almost unanimously requested andcould improve uptake. Supporting find-ings by Alsaleh et al. (12), participantswere concerned about the size, visibility,and attention demands of automated in-sulin delivery systems. Being able to cus-tomize device features and attributesbased on one’s lifestyle was also de-scribed as important. Device featureswere more than specifications and mate-rials; they were facilitators with thepower to lessen stigma and improveuser experiences and trust while support-ing PWD’s interest in control, indepen-dence, and data-driven responsivenessof devices.Third, participants expressed concerns

about whether automated insulin deliv-ery systems would add to or complicateregimen burden and how systems wouldfit into specific aspects of participants’lives. This finding extends the conclusionof Lowetal. (13) that PWDweremost con-cerned about the size, portability, andtechnical problems associated with theirinsulin pumps. Participants were inter-ested in how such systems would adaptto help manage diabetes in the presenceof moderate or intense physical activity.Concerns were also linked to low levels ofinformation about automated insulin de-livery systems among participants. In fact,many participant concerns may be ad-dressed as more information about auto-mated insulin delivery systems is revealedand developers communicate the capa-bilities of their devices. Despite concerns,participants were cautiously optimisticabout the potential impact of automatedinsulin delivery systems and their interestin trying such systems as they becomeavailable. Across the researchers’ experi-ences with diabetes technology trials, pa-tients almost unanimously affirm theirinterest in continuing the use of test de-vices even when patients experience anumber of difficulties throughout trialimplementation.Limitations of this study include a study

design where participants were not al-ways classified by characteristics thatmight have allowed for more in-depth

comparative analysis as well as volumi-nous qualitative data that requiredresource-intensive analytic proceduresin order to distill important patterns andthemes. Although site data collection wascoordinated and questions were stan-dardized, differences in site implementa-tion and procedures limit our ability tocompare across groups based on demo-graphic and behavioral characteristics,e.g., sex or pump users. Some sites re-ported aggregated demographic data,and different demographic variableswere captured across sites. Our samplesappear to differ from the broader popu-lation of people affected by type 1 dia-betes in that CGM usage is higher in oursamples than in other studies, possiblyindicating differential access, motiva-tion, and/or technological sophisticationthan the general population of those af-fected by type 1 diabetes (1). Further-more, in an effort to present the mostsalient themes, not all themes discussedin participant data are mentioned, justthose most commonly brought up. More-over, the interview and focus groupguides focused on experiences, expecta-tions, and potential effects on quality oflife, so the absence of some themes maybe artifactual. For example, concernsabout insurance coverage and additionalsupplies were raised, but discussionsabout direct costs of automated insulindelivery systems were infrequent. Over-all, study data depict cross-geography andcross-stakeholder verification of themesrelated to stakeholder understanding ofhow automated insulin delivery systemsmight change their lives. We hope thatthe insights provided by the study out-weigh the study’s limitations.

By unveiling the importance of psy-chosocial factors in the early stages ofautomated insulin delivery systemdevelop-ment, this research affirms a critical role forqualitative approaches to clinical questionsin diabetes research. This study contributesto limited literature about the psychosocialfactors that support or detract from medi-cal device utilization. As a qualitative study,this research illuminates these factors in thewords and experiences of five criticalgroups, including three developmen-tally distinct groups of PWD and their lovedones. Reinforcing research on other type 1diabetes devices, our study also found so-cial and emotional elements of diabetesmanagement and care to be critically im-portant in considerations of device “fit”

within one’s life (7,12–15). Althoughstudy participants may differ from theoverall population of people living withtype 1 diabetes in terms of both theirglycemic control anduse of diabetes tech-nological devices, we believe the varietyof experiences across geographies, payertypes, and disease management perspec-tives adequately outlines domains of ex-perience that should be examined withfuture quantitative studies. Our grouphas recently used this study’s results toguide the development of comprehensivesurvey measures to predict uptake andsustained use of automated insulin deliv-ery systems. These survey measureswill be deployed in future studies wherestatistical analyses, including those de-pendent upon individual-level charac-teristics, are planned to examine thedistribution of specific factors and ex-periences associated with automatedinsulin delivery system interest anduptake, across diverse groups affectedby type 1 diabetes. In preparation forbroader use of automated insulin de-livery systems and related experiencesurveys, researchers should includeitems that assess device features andpatient expectations and concerns, aswell as potential changes in regimens,management behaviors, vigilance, gly-cemic control, and physical and emo-tional burdens.

Acknowledgments. The authors acknowledgethe research contributions of the site researchteams, including Annie Astle (BournemouthUniversity), Sarah Collard (Bournemouth Uni-versity), Persis Commissariat (Joslin DiabetesCenter), Kimberly Garza (Lurie Children’s Hospi-tal), Pramod Regmi (Bournemouth University),Danielle Slaughter (Lurie Children’s Hospital),and Amanda Whitehouse (Joslin Diabetes Cen-ter). Marisa Hilliard (Baylor College of Medicine),Molly Tanenbaum (Stanford School of Medicine),and Sarah Hanes (Stanford School of Medicine)also provided invaluable assistance.Funding. This study was supported by fundingfrom The Leona M. and Harry B. HelmsleyCharitable Trust.Duality of Interest. No potential conflicts of in-terest relevant to this article were reported.Author Contributions. D.N. designed the quali-tative methods, helped coordinate data collection,led implementation of the data analysis, manageddata, and contributed to manuscript writing andposthocmeasuredevelopment.S.C.S.designedthequalitative methods, analyzed data, and wrote themanuscript. E.I. assisted in qualitative analysis de-sign,wrote thefirst list of codesusedbycoders, andcollected, managed, and analyzed data. K.D.B.,J.W.-B., L.L., and K.K.H. served as site principal

1460 End Users and Automated Insulin Delivery Systems Diabetes Care Volume 40, November 2017

investigators and were involved in the review andimplementation of ethical approvals, research de-sign, data collection, guidance of analytic discus-sions, and post hoc measure development. Allauthors reviewed and edited the manuscript. K.K.H.istheguarantor of thiswork and, as such, had fullaccess to all the data in the study and takesresponsibility for the integrity of the data andthe accuracy of the data analysis.

References1. Miller KM, Foster NC, Beck RW, et al.; T1DExchange Clinic Network. Current state of type 1diabetes treatment in theU.S.: updateddata fromthe T1D Exchange clinic registry. Diabetes Care2015;38:971–9782. Pickup J. Insulin pumps. Int J Clin Pract Suppl2011;65(Suppl. 170):16–193. Renard E. Insulin pumpuse in Europe.DiabetesTechnol Ther 2010;12(Suppl. 1):S29–S324. Beck RW, Tamborlane WV, Bergenstal RM,Miller KM, DuBose SN, Hall CA; T1D Exchange

Clinic Network. The T1D Exchange clinic registry.J Clin Endocrinol Metab 2012;97:4383–43895. White HD,GoenkaN, FurlongNJ, et al. TheU.K.service level audit of insulin pump therapy inadults. Diabet Med 2014;31:412–4186. Wong JC, Foster NC, Maahs DM, et al.; T1DExchange Clinic Network. Real-time continuousglucose monitoring among participants in theT1D Exchange clinic registry. Diabetes Care2014;37:2702–27097. Naranjo D, Tanenbaum ML, Iturralde E, HoodKK. Diabetes technology: uptake, outcomes, bar-riers, and the intersectionwith distress. J DiabetesSci Technol 2016;10:852–8588. Creswell JW. Qualitative Inquiry andResearchDesign: Choosing Among Five Approaches.Thousand Oaks, CA, SAGE Publications, Inc.,20129. NVivo Qualitative Data Analysis Software.Melbourne, Australia, QSR International Pty Ltd.,201510. Braun V, Clarke V. Using thematic analysis inpsychology. Qual Res Psychol 2006;3:77–101

11. Barnard KD, Hood KK, Weissberg-Benchell J,AldredC,OliverN, Laffel L. Psychosocial assessmentof artificial pancreas (AP): commentary and reviewofexisting measures and their applicability in AP re-search. Diabetes Technol Ther 2015;17:295–30012. Alsaleh FM, Smith FJ, Taylor KM. Experiencesof children/young people and their parents, usinginsulin pump therapy for the management oftype 1 diabetes: qualitative review. J Clin PharmTher 2012;37:140–14713. LowKG,Massa L, LehmanD,Olshan JS. Insulinpump use in young adolescents with type 1diabetes: a descriptive study. Pediatr Diabetes2005;6:22–3114. Ritholz MD, Smaldone A, Lee J, Castillo A,Wolpert H, Weinger K. Perceptions of psychoso-cial factors and the insulin pump. Diabetes Care2007;30:549–55415. Barnard KD,Wysocki T, Allen JM, et al. Closingthe loop overnight at home setting: psychosocialimpact for adolescents with type 1 diabetes andtheir parents. BMJ Open Diabetes Res Care 2014;2:e000025

care.diabetesjournals.org Naranjo and Associates 1461