Individuals with type 1 diabetes (T1D) are confronted with complex tasks through which to manage their blood glucose (BG) levels. T1D is an autoimmune disease where the beta cells in the pancreas no longer produce insulin, resulting in dangerously high BG levels or hyperglycemia. The person diagnosed with T1D is subsequently required to self-administer insulin. This involves regular self-monitoring of BG levels and calculation of appropriate insulin doses. There is a delicate balance between the reductions of the risks of long-term complications (often associated with hyperglycemia) and those of hypoglycemic events. This puts emphasis on adherence and patient behaviors. It has been suggested that large numbers of people with T1D are nonadherent [1]. Additionally, Patton [2] highlights multiple social, emotional, and cognitive barriers. The prevalence of new and emergent technologies to support self-management of T1D through personal data logging processes and support for decision making may have the potential to address these issues.

There may be a dilemma for health care providers due to the economic implications of adopting such technologies for individuals compared with potential public health benefits. This raises the issue of identification of adults with T1D who may benefit the most. There are associated questions around how to investigate and evaluate the benefits of such technology with respect to specific populations in such a way as to inform future design decisions. Thus, consideration of psychological and behavioral constructs alongside evaluation of the usability of devices, also known as human factors, is an integral component of any investigation that involves clinical consideration for emergent technology aimed at self-management of T1D.

The objective of this review was to describe the relationship between human factors and technology adherence for data logging processes in adults with T1D and to explore the factors that influence this association.

The aim of this systematic review was to describe the relationship between human factors and adherence with technology for data logging processes in adults with T1D and to explore the factors that influence this association.

An integrative literature review research design was chosen because it provides a more holistic conceptualization on a complex topic [25] such as human behavior and facilitates inclusion of diverse methodologies and theories, given the interdisciplinary approach toward research in the field.

A protocol was developed (Multimedia Appendix 1) to clarify the aims, sampling strategy, exclusion and inclusion criteria, methods, outcomes, language, and search strategy.

A systematic search of the literature was performed in accordance with the preferred reporting items for systematic reviews and meta-analyses (PRISMA) [26] in January to March of 2017 (Multimedia Appendix 2)

The following databases were searched: Computing Research Repository (2006 to January 2017); PsycINFO, EMBASE, and MEDLINE (2006 to January 2017); Web of Science (2006 to January 2017); Zetoc (2006 to January 2017); Excerpta Medica and Scopus (2006 to January 2017); and ProQuest (2006 to January 2017). Only research that was undertaken during the last 10 years was included as technology for the self-management of T1D has been developing rapidly during this time. Search terms included: Diabet* AND Techno* AND Behavi*; Self-manage* OR self-manage* OR manage* OR self-care OR self-care; technolog* OR telehealth OR telemedicine OR reminder system* OR text messag* OR application OR app*; adhere* OR compliance OR barrier OR problem* OR obstacle: MH Diabetes Mellitus, Type 1*.

Searches were limited to adults (over 18 years) and filtered to studies of adults published in English. Reference lists were also searched in addition to subject-specific websites and key journals (Multimedia Appendix 1). The search strategy was carried out in collaboration with a university health care librarian. Unpublished studies (dissertations and theses) were excluded, in addition to editorials, opinions, and discussion papers. Studies were reviewed for the following criteria: (1) primary research; (2) empirical data on adherence to data logging processes with the use of advanced technology for adults with T1D; (3) an investigation of the relationship with psychological, social, and human factors; and 4) the psychological outcome measures were explicit (quantitative studies) or alternatively included a clearly described picture of the phenomenon that included the user perspective (qualitative studies).

The search strategy produced 1 article in the Computing Research Repository; 348 articles in PsycINFO, EMBASE, and MEDLINE; 40 articles in the Web of Science; 84 articles in Zetoc; 38 articles in Excerpta Medica and Scopus; and 36 articles in ProQuest. Once duplicates were removed, additional articles were excluded due to limitations associated with unclear abstracts or for not meeting the inclusion criteria (ie, children, type 2 diabetes, and gestational diabetes). In total, 72 citations were retained and each abstract was read for relevance.

Also, 3 citations were found from searching reference lists and key journals. One study, which included children and their carers, was retained because outcomes were compared with adults who also participated within the study [27]. To reduce bias and ensure that only the most relevant articles were selected, the second and third authors reviewed the titles and abstracts regarding the protocol criteria and a consensus was reached about the articles to be included in the review. In total, 22 articles met the inclusion criteria, and these included 14 quantitative studies, 5 qualitative studies, and 3 mixed-method studies (Multimedia Appendix 2).

Whittemore and Knafl’s approach [25] of using as many instruments as necessary to evaluate the quality of the data was taken because this is an integrative review, and the data are drawn from more than one disciplinary area that use a range of research traditions that align with quantitative, qualitative, or mixed-method research designs. The instruments for appraisal were selected from the University of South Australia International Centre for Allied Health Evidence [28] databases of critical appraisal tools. The following criteria were taken into consideration for types of study design: demographic information of the participants and statement of research question, appropriateness of the research question for the selected study design, and approach to recruitment reported (Table 1). The criteria for quantitative study designs included power analysis reported response rate, reliability and validity of study instruments and method of data analysis (Table 2). The following criteria were considered for qualitative studies: theoretical perspectives, audit trail, member checks, peer review of qualitative data, and method of data analysis (Table 3). The first author undertook the quality appraisal of each study, which was peer-reviewed independently by second and third authors.

Following the critical appraisal process, 4 studies were further excluded for poor methodological design.

The review data were categorized and synthesized into the themes that underpinned the human constructs that were examined and the outcomes that were reported. Mile and Huberman’s [46] approach to coding of data, which involves data reduction and comparison, was utilized.

There was inconclusive evidence about the relationship between adherence to data logging process and measurable outcomes. For example, Kirwan et al [35] examined a freely available iOS app—Glucose Buddy—combined with text messaging feedback from a diabetes educator aimed at the improvement of glycemic control. The intervention group showed a significant decrease in HbA_1c (mean −1.10; SD 0.74; (P ≤.001) over the 9-month period of the study; however, linear regression showed no significant relationship between the level of engagement with the app and these outcomes. This result may be interpreted with caution, given the small sample size (n=27). Furthermore, there was a potential socioeconomic bias in that participants were required to have iOS ownership.

Groat et al [29] analyzed individual participant internet protocol address data to characterize the relationship between adherence to insulin bolus dosing, logging of carbohydrate intake, and BG monitoring and glycemic control for a 1-month period. The only significant outcome was that an increase in daily insulin bolus doses had an impact on increasing the number of days that the BG was at target (r=.93). The reported results were based upon an extremely small sample (n=8) and described as regression analysis, which contradicts the researchers’ claims for undertaking a qualitative study.

Some findings suggest that adults with T1D may feel more satisfied with their treatment on transition to advanced technology. For example, Gonzalez et al [34] evaluated the overall effect of adding a telemedicine system for adults with T1D who were treated with an insulin pump and real-time CGM. This was a longitudinal study that measured the physical and psychological outcomes of the intervention. Mean plasma HbA_1c was significantly lower at 6 months compared with prebaseline (6.97 vs 7.5; P=.01); there was a significant reduction in glucose variability at 6 months compared with baseline (53.1 vs 68.7; P=.04) and prebaseline (53.1 vs 67.3; P=.04), and time spent interacting with the sensor correlated positively with time in normoglycemia (r=.72; P=.03) and negatively with occurrences of mild hypoglycemia (r=.64; P=.02). From a psychological perspective, there was an improvement in quality of life scores at 6 months in comparison with baseline (92.4 vs 86.9; P=.01), and participants with poorer glycemic control had significant improvements with prior dissatisfaction with treatment (34.3 vs 31.6; P=.01).

The authors acknowledged that the findings were based on a small sample size (n=15), and therefore, it is not possible to generalize the outcomes. The authors also questioned whether the point of being observed affected the outcome measures. However, the study did show that there may be benefits for well-controlled individuals using CGM in conjunction with telemetry in terms of HbA_1c and quality of life as reported in the previous paragraph.

There is some consistency regarding the perceived physical outcomes and satisfaction of the above study with the findings of Barnard et al [45], who measured the relationship between satisfaction when transitioning to the then-current insulin pumps (Animas Vibe CGM-enabled system IV) and personal glycemic control. The most significant contributing factor to treatment satisfaction was when BG levels were consistently <7.00 mmol/L (P=.009). The limits of this study are that the findings are based on self-report, and it is not clear why only 22 items of the 50 on the Insulin treatment satisfaction questionnaire were included on the survey instrument.

Frequent users of existing diabetes technology may find an easier transition to more advanced options. For example, Tansey et al [27] examined the perceived barriers and benefits to CGM use and how this related to frequency of use. Engaged CGM users were more satisfied, with higher frequency users less bothered by the “hassles” of the device. Frequent users were classified as engaged with CGM for more than 6 days per week and infrequent users less than 4 days per week. Adults and parents of users had higher total and subscale scores on the CGM satisfaction survey (P=.0009). All respondents reported that visualization of glucose trends and the opportunity to detect hypoglycemia were the best aspects of use of CGM (text item responses in the questionnaire).

Martinez-Sarriegui et al [33] analyzed patient behavior when using the intervention of telemedicine system combined with CGM to identify how the CGM data captured participant interactions with the mobile system. In 2 phases of the experiment (with and without the telemedicine system), participants were provided with tools for visualization, management of monitoring data, and wireless downloading of data from an insulin pump via a personal smart assistant running on a personal digital device. The number of times interacting with the system was higher during the intervention phase (29.0 vs 18.8; P=.04), and the total time spent interacting with the system was also higher during the intervention phase (04:27:11 vs 01:47:07; P=.009).

Kamble et al [32] compared weekly estimates of time, changes in time, and patient time costs associated with diabetes-related care between SAPT and MDI. They used data on patient-reported time collected over a 52-week period. Participants were required to log the total time spent per week on diabetes management for a range of diabetes-related variables. The total time spent on the SAPT arm of the study was higher than time spent on MDI during and after pump initiation within the overall 52-week study. The reported weekly time estimates were as follows: SAPT 4.4 hours and MDI 3.4 hours (95% CI 0.4-1.7). However, all adults with T1D in the study reported that they spend considerable time on diabetes care.

Each of the above 3 studies suggests that engagement with technology is time consuming. Given that the inclusion criteria for the Tansey et al [27] study were prior high frequency of self-monitoring, it is not clear if the technology was a mediating factor for engagement. The Martinez-Sarriegui et al [33] study was limited by a small sample size (n=10) and did not include any details about how the study instrument was developed or how the participants valued the feedback from the telemedicine system. Furthermore, there was a possibility for margin of error with the Kamble et al [32] study as it was not clear how participants measured time costs.

Some researchers have attempted to understand the way in which the human complexities of diabetes have the potential to be mediated with the use of advanced technology.

Some researchers have adopted an ethnographic approach to explore how technology is used in the context of the everyday lives of adults with T1D.

O’Kane et al [42] took a sociocultural perspective and reported on 3 qualitative studies that examined how devices for adults with T1D are adopted, carried, and used in a variety of everyday contexts. This is based on the premise that adults with T1D are encouraged to self-regulate by HCPs, but the nature of everyday life is contingent upon the dynamics of the unfolding situation. The following themes emerged from the data analysis: misuse, inappropriate use, and unintended use of the technology. The authors’ main point is that any individual can report a wide variation in normal use of their technology across a continuum of public use, work-life use, and in the company of friends and family. This was based on the perceived emotions and attitudes of the other party within a given context. Uncertainty in discrete situations can lead to hiding a device, whereas showing off the device in other situations can lead to normalization and control of a situation. This corresponds with the findings of Ritholz et al [41], which were reported in the previous section regarding the place of significant others in uptake and use of technology.

Research that examines the meaningfulness and perspectives of the user has an important role to play in the future and ongoing development of advanced technology. Shepherd et al [40] explored both desires and concerns regarding the use of CGM for self-management. The findings suggested that adults with T1D who already used insulin pumps and CGM had a diverse range of attitudes and concerns along a continuum regarding personalized glucose advisory systems. Participants would have liked advice from the system on suggestions for correction boluses, basal rates, insulin-carbohydrate ratios, and alerts to the risks of hypoglycemia. However, it would be necessary for the individual to understand how the advice was generated, trusting that all personal variables would be considered to develop the confidence to relinquish control to an automated system. A shortcoming of the study is that it was not entirely clear how the themes were arrived at.

Lawton et al [44] (2014) found evidence of similar themes during a longitudinal study of the use of insulin bolus calculators following the intervention of a dose adjustment for normal eating course. Adults with T1D were motivated by the device because it saved time and effort in calculations; however, those who were confident in their mathematical ability undertook their own individual calculations and were paradoxically less likely to use the device over time. Reliance on the calculator alone had a detrimental impact on glycemic control. Some participants left the ratios unchanged until their next clinician/study review, and for some, this was attributed to not knowing how to change the settings. Underconfidence in carrying out personal calculations or not knowing how to change settings led to loss of trust in the technology.

Storni [43] contends that diabetes is more than a disease and should be regarded as a complex lifestyle. People with T1D develop lay expertise that is unique to their situation. This creates implications for technology design, and it is crucial to involve the user in the process. This perspective is based upon findings that emerged from an ethnographic study on diabetes support groups and by following individuals with T1D within the context of their everyday lives [43]. The purpose was to examine what participants really did in dealing with their condition as opposed to what they were told to do by clinicians. These findings influenced the design of a tagging system for events from everyday life to link them to carbohydrate intake and BG readings to create meaning between the events and a log for the individual on a mobile device. A shortcoming of the study is that the report provided a lack of demographic information about participants, which is important in qualitative research to determine transferability to other contexts. Nevertheless, there is an emergent field of research that addresses the diverse needs of people with T1D in the design of technologies.

Advanced technology for the management of T1D needs to have clear benefits that are meaningful to adults with T1D. The aim of the review was to describe the relationship between human factors and adherence with technology for data logging processes in adults with T1D and to explore the factors that influence this association.

There was inconclusive evidence about the relationship between adherence to data logging and measurable outcomes in relation to the review question. However, clinical values may have less importance than perceived outcomes for individuals. The review did suggest increased satisfaction with treatment on transition to advanced technology; however, this was biased toward frequent users of existing technologies and with an acceptance of the time required to spend on diabetes care.

The review also showed some benefits of advanced technology for older people by mediating complexities and fear of hypoglycemia. There appears to be a wide variation in the normal use of technology for adults with T1D across a continuum of sociocultural contexts. There is also a variability regarding user involvement in the design of future technologies and the role of “significant others” and this requires further research. People need to be able to trust technology as the capacity for intelligent decision-making advances.

In the literature that was reviewed, participants appeared to be a highly selective group biased toward white populations. Another limitation was the relatively small sample sizes of some of the quantitative studies included within the review, only 1 study [35] reporting on a power calculation, thus making it hard to generalize the findings.

A significant issue was that where demographic characteristics were reported (Table 1), 95% of the participants were described as white. The data suggest that those from higher socioeconomic groups are more likely to have access to and engage with technology in their self-management behaviors [38]. Of the studies, 2 [30,45] purposefully selected participants with prior adherent behaviors; however, 1 study recruited participants who were less engaged with technology and adherence [32].

The predominance of white participants, combined with the fact that 6 of the reviewed studies were samples drawn from parent clinical trials, suggests that the data are based on a highly selective group. This may not be representative of the general adult population with T1D. The mean baseline HbA_1c of 7.9 implies that participants had relatively good control before entering one of the respective studies, which may suggest a largely adherent sample.

Although qualitative research is not considered to be necessarily generalizable by some audiences [47], it is incumbent on the researcher to provide full demographic descriptions so that the generic reader from an interdisciplinary audience can decide about the transferability of findings to his or her own practice, research, or development context. Furthermore, trustworthiness of the findings can be clarified based on participants’ checking of data and peer review of data analysis. This was a shortcoming of some of the literature that was reviewed.

Engaged participants spend considerable time on diabetes care, so it is important that they receive support to make informed choices. On the basis of this review, it was found that these are the people most likely to benefit from the affordances of advanced technology; however, this creates a tension between these populations and hard-to-reach groups who may be at increased risk of diabetes complications. Furthermore, Lawton et al [44] suggest that in general HCPs lack knowledge about the scope and purpose of advanced technology for diabetes. This is important, given the potential information overload and the frustrations that adults with T1D are presented with when using technology demonstrated within this review and other literature [9,19,41].

What is meaningful for the adult with T1D might not be important for the clinician and may therefore require mediation. Storni [43] found that patient-generated tags for mobile devices developed by participants were not of interest to clinicians who were more focused on numerical values.

James et al [47] have explored the perceptions and experiences of diabetes educators when supporting the use of advanced technology and suggest that there are challenges for all parties. This includes device costs, access to Wi-Fi, and appropriate mobile devices. CSII puts demands on diabetes services, and there are also challenges associated with keeping up to date with technology, such as the skills to analyze data from patient mobile devices. This research study suggested that there is a need for mentorship of HCPs and a review of service configurations as technology advances.

Engaging people in the design of technology for T1D is essential for meeting the requirements of the user. Within this review, O’Kane et al [42] suggested that the design of devices needs to be both discrete and more public for context-dependent behavior. Lawton et al [44] suggested that voice recognition for entering data would make data logging practices easier for some people. Engaged participants appeared to be able to deal with the hassle and time required for diabetes-related tasks. However, a challenge for designers is to build in time-efficient capabilities.

There is a requirement for studies within the context of day-to-day data logging that are representative of the general adult population with T1D. There is more scope for research that explores how technology could be used to engage hard-to-reach groups. Many of the studies in this review were short-term; however, the study undertaken by Lawton et al [44] on the use of insulin bolus calculators was in-depth and over time (1 year), thus providing a rich and diverse view of adherence and nonadherence along a trajectory, which provided important nuances about human factors. There is also a need to study the role of significant others within data logging processes [41]. There appeared to be a dearth of mixed-methods studies, which if conducted through a rigorous methodological process have the potential to capture the complexity of human factors by maximizing the advantages of more than one research design. There is also a need for future studies that explore the sociocultural and demographic factors associated with technology uptake.

A limitation of this review is that the data were drawn from databases, which excludes emergent unpublished research in a fast-moving field. However, this was mitigated by extracting the data from sources retrieved from 9 key databases covering the fields of health, medicine, and computer science, and the search was performed in collaboration with a university health care librarian.

The application of critical appraisal frameworks used in this review made it possible to evaluate the reliability, validity, and trustworthiness of each of the studies under consideration. This review presents a contribution to the field in comparison with descriptive mapping reviews and highlights areas where research design could be improved. By abstracting data from each of the studies, it was possible to compare the findings and focus on the human factor constructs of adult populations with T1D, including older people.

The purpose of this systematic review was to explore the relationship between human factors and the adherence to technology for data logging in adults with T1D. The research design was an integrative review, given the interdisciplinary nature of research in the field and the diverse methodological approaches taken to inquiries.

The aim of the review was to analyze the relationship between human factors and adherence to technology for data logging in adults with T1D. Overall, the sample was drawn from homogeneous populations that may not be the complete representation of adults with T1D. Inconclusive evidence was found about the relationship between adherence to data logging with advanced technology and measurable outcomes. There was some suggestion that adults with T1D may feel more satisfied with their treatments on transition to advanced technology. Qualitative research suggested that the way in which technology is used by any individual varies along a continuum and is contingent upon the sociocultural context in which technology is used. As technology continues to advance, there is a need for more research into how trusting the individual is of personal treatment advice, which is generated through advanced technology.