Abstract
Objective:
Evaluations of technology to help adults manage type 2 diabetes (T2D) have yielded mixed results. We analyzed the effectiveness of a free app linked to a glucose meter to study reductions in glucose levels over time among a self-selected sample of adults with T2D.
Research Design and Methods:
Adults with T2D >12 months, >21 years, ability to read English (insulin using—IU and non-insulin using—NIU) who independently elected to pair their CONTOUR NEXT ONE meter with the CONTOUR DIABETES App were invited to participate. Glucose data from baseline to 16 weeks were uploaded to the cloud (N = 461). Assessment of diabetes distress, medication taking, quality of life, and hypoglycemia concerns occurred at baseline, six, and 16 weeks.
Results:
Findings indicated a significant decrease in weekly glucose levels over time: baseline mean = 169 (62.0) (9.4 mmol/L; 3.44); 16-week mean = 146.5 (36.0) (8.1 mmol/L; 2.0) (P < .001), with no IU and NIU differences. Largest reductions occurred during the first six weeks, with no later rebound effects. Significant, though modest, improvements in global quality of life (P = .03), hypoglycemia concerns (P = .01), and diabetes distress (P < .001) occurred over 16 weeks.
Conclusions:
Making an App for monitoring glucose easily available for download with a glucose meter can be helpful for self-selected adults with T2D. Effective utilization assumes that users are sufficiently motivated and engaged, are comfortable and trusting of the technology, and have sufficient knowledge of how to make use of the glucose data.
Introduction
There are now literally hundreds of smartphone technologies and related Applications (hereafter called “Apps”) to help adults with diabetes monitor and enhance disease management over time, including close assessment of glucose levels, diet, and physical activity.1-3 Careful evaluation of both the short- and long-term use of these Apps and their effect on glycemic outcomes has been limited, and the results have been mixed.4,5 Furthermore, most studies have utilized highly selected samples in controlled settings, thus limiting the generalization of findings to the broader, heterogeneous population of adults with diabetes. 5
The Real-World Evidence of Diabetes App Use in Daily Life study (REALL) was designed to address these issues. This study followed a self-selected, national, community-based sample of insulin-using (IU) and non-insulin-using (NIU) adults with type 2 diabetes (T2D) (N = 461) over 16 weeks who, on their own, had independently elected to download and pair the publicly available CONTOUR DIABETES App with their CONTOUR NEXT ONE glucose meter to assist with glucose monitoring and self-management. In a previous report we described the frequency of App use and glucose testing over time, along with predictors of utilization and testing. 6 Results indicated a steady and dramatic reduction in both the frequency of App use over time for both IUs and NIUs, from about five times per week during the first week to less that once per week by week 16. About 40% ceased App use altogether by 16 weeks. Frequency of glucose testing went from about 17 times during the first week to about eight times a week by week 16, with IUs testing more frequently than NIUs. Continuous App users were older and they reported higher baseline diabetes distress, better baseline medication taking, and more positive attitudes toward technology at baseline than those who stopped App use during the study period.
Whereas our initial report focused on the frequency of both App use and glucose testing, in this report we address the efficacy of App use with respect to changes in glucose values over time. We posed three research questions. First, what was the effect of App use and frequency of testing on changes in glucose levels over 16 weeks? Second, did glucose change vary by participant demographic, diabetes-related characteristics, and frequency of App use and glucose testing? Last, what was the effect of App use on quality of life, diabetes distress, hypoglycemia concerns, and medication taking, and did these effects vary as a function of participant demographic and diabetes-related characteristics? With the understanding that a single technology is rarely uniformly effective for all users, information from this study will help identify which subgroups of T2D adults can make best use of this straightforward aid to glucose management.
Methods
All adults in the United States who paired their CONTOUR NEXT ONE meter with the free and readily available CONTOUR DIABETES App from March to July 2020 were invited to participate. Inclusion criteria were a self-reported diagnosis of T2D for at least 12 months; use of oral medications, non-insulin injectables, or insulin; ability to read English; and age 21 or older. Individuals were excluded if they reported current use of a continuous glucose monitoring device, had experience with the App prior to receiving an invitation to participate, or who did not log at least one glucose reading during the first four weeks after paring their App, and/or did not complete the baseline survey.
Participants completed an online informed consent and an online survey within four weeks of pairing their App (baseline). Similar online surveys were completed at six and 16 weeks after baseline. Glucose data recorded on the meter could be uploaded by the user using the App during the 16-week study duration. Each new upload contained all glucose data recorded since the last upload, and no test results could be excluded from an upload. Participants could then view their glucose data over time through pattern recognition displays and other formats. The App also allowed tracking of a variety of user-entered disease management information that could be combined and integrated with their glucose data, for example, medication taking, exercise, carbohydrate consumption. Participants received an electronic gift card or check for $25 at completion of the baseline survey, $30 for completion of the six-week survey, and $45 at the completion of the 16-week survey. The study received Human Subjects approval from Advarra IRB.
Measures
Demographic characteristics included age, gender, education (less than high school, high school, some college, four-year college graduate, graduate degree), ethnicity (non-Hispanic white/non-white), years with diabetes, current insulin use (yes/no), employment status (yes/no), and most recent self-reported HbA1c.
Three measures of App use, and two measures of glucose testing were calculated: duration of App use (number of days from initial to final App use during the 16-week study); frequency of App use (total number of times the App was used during a participant’s use period); intensity of App use (average number of App use per week—ie, total frequency divided by duration, in weeks); frequency of glucose testing (total number of glucose tests during a participant’s use period); and intensity of glucose testing (average number of glucose tests per week—frequency divided by duration, in weeks). Note that a single App upload could include glucose levels from multiple tests.
Several measures were included in the baseline survey and repeated in the six- and 16-week surveys. Included were the five-item Regimen Distress (RD) (alpha = .90) and the five-item Emotional Burden (EB) (alpha = .88) subscales of the Diabetes Distress Scale. 7 Items were scored on a six-point scale from “not a problem” to a very serious problem.” Medication taking was assessed by a modification of the seven-item Hill-Bone Medication Adherence Scale (alpha = .78). 8 Items were scored on a four-point scale from “none of the time” to “all of the time” and the total score was the sum of the items, with higher scores reflecting poorer medication taking. The five-item World Health Organization-5 scale (WHO-5) is a non-diabetes-specific measure of quality of life (alpha = .86). 9 Items were scored on a five-point scale from “at no time” to “all of the time,” and scores were converted to a 100-point scale with higher scores indicating a higher quality of life. The 14-item Hypoglycemia Attitudes and Behavior Scale (HABS) assessed attitudes and concerns about hypoglycemia. 10 It yielded a scale total plus three subscales: Anxiety (five items, alpha = .74), Avoidance (four items, alpha = .71) and Confidence (five items, alpha = .80). Items were scored on a five-point scale from “strongly disagree” to “strongly agree.”
Statistical Analyses
Data were cleaned and distributions were reviewed graphically; no significant deviations from normality requiring transformation were noted. Hierarchical Linear Modeling (IBM SPSS Statistics v. 25) was used to conduct the primary analysis. Multilevel modeling using full maximum likelihood (FML) examined change over time in glucose levels. Time (in weeks) was the level 1 predictor and the main effect of interest was weekly glucose average as the level 1 outcome. Cross-level interaction effects (time/level 1 by baseline predictor/level 2) were then added to examine the effect of several baseline variables (age, insulin status, gender, self-reported baseline HbA1c, WHO-5, medication taking, HABS, EB, and RD) and patterns of App use and glucose monitoring (duration, frequency, intensity) as potential predictors of reductions in glucose levels over time. Finally, the same multilevel modeling approach described for glucose levels was used to examine change over time in participant-reported outcomes, and potential moderation of this change by select baseline variables (age, insulin status, gender, self-reported HbA1c). Missing data were minimal and were handled using FML.
Results
During the invitation period, 461 adults with T2D (n = 108 IU, n = 353 NIU) linked their meter with the App and met inclusion criteria (Table 1). Average age was 51.6 (11.6) years, 48% were female, average time with T2D was 6.9 (7.9) years and average self-reported baseline HbA1c was 8.1% (2.0) (mmol/mol = 65 [8.9]). IUs had T2D significantly longer, were more likely to be employed for wages, and reported a significantly higher HbA1c than NIUs (all P < .001). Of the 461 participants who completed a baseline survey, 453 (98%) completed the six-week survey and 461 (100%) completed the 16-week survey.
Participant Characteristics, N = 461.
Changes in Glucose Levels Over Time
Findings indicated a significant decrease in glucose levels over time: the mean weekly glucose level at week 1 was 169 (62.0) (9.4 mmol/L; 3.44) and the mean weekly level at week 16 was 146.5 (36.0) (8.13 mmol/l; 0.99) (P < .001) (Figure 1) . The largest reduction in glucose levels occurred during the initial weeks, between weeks 1 and 6, with no significant reductions thereafter. Notably, no rebounding increases in glucose levels occurred after the initial reductions. Additional analyses indicated no significant differences in the pattern of glucose reductions between IUs and NIUs, nor were there significant differences based on age, gender, frequency, intensity and duration of App use, or frequency and intensity of testing. In addition, reductions in glucose levels were the same for those who stopped using the App before the 16-week end of the study, compared with those who continued to use the App through the entire 16-week period. Thus, as shown in Figure 1, about six weeks of App use led to a significant reduction in glucose levels for a broad range of users.
Weekly mean blood glucose levels (mg/dL) over 16 weeks (P < .001), N = 461.
Predictors of Reductions in Glucose Levels Over Time
Multiple regression analyses evaluated reductions in glucose levels as a function of participant characteristics recorded on the baseline survey: HbA1c, WHO-5, medication taking, HABS and EB, and RD. Only one significant finding occurred: those who reported higher EB at baseline recorded greater reductions in glucose levels than those who reported lower EB scores at baseline (P = .01). Baseline levels of HbA1c, quality of life, medication taking, RD, and concerns about hypoglycemia were unrelated to changes in glucose levels with App use over time.
Changes in WHO-5, Diabetes Distress, Concerns About Hypoglycemia, and Medication Taking Over Time
Additional analyses evaluated changes in the survey scale scores across baseline, six and 16 weeks (Table 2), including moderation tests for differences in findings based on participant age, gender and insulin use. Overall, across the 16 weeks, significant, although modest, improvements occurred in quality of life (WHO-5) (P = .03), HABS total (P = .01), HABS Avoidance (P = .05), HABS Confidence (P = < .001), HABS Anxiety (P = .007), and EB (P < .001). No significant changes occurred for RD (P = .07) or medication taking (P = .44). Across the 16-week study, therefore, we found modest, but significantly improved quality of life, reductions in hypoglycemia concerns and anxieties, and reductions in diabetes-related EB.
Changes in Survey Scale Data, N = 461.
Data are mean (SD) unless otherwise indicated.
Abbreviation: WHO, World Health Organization.
Interestingly, where significant changes in the psychosocial measures were documented, most change occurred only between six and 16 weeks. For example, significant changes occurred only between six and 16 weeks for EB, and HABS total, Avoidance, and Anxiety (all P < .05), without any significant change occurring between baseline and six weeks for these variables. Significant change between baseline and six weeks only occurred for RD and HABS Confidence (both P < .05).
Three significant moderation effects also occurred: improvements in WHO-5 scores were greater for younger, compared with older participants (P = .04); reductions in HABS Avoidance were greater for IUs than NIUs (P = .009); and improvements in medication taking were greater for IUs than NIUs (P = .04).
Discussion
In this sample of self-selected App users, we find a significant reduction in average weekly glucose levels between week 1 of App use and week 6, with no further reductions or rebound effects thereafter through week 16. There are no differences in glucose changes between IUs and NIUs over time, although IUs tested more frequently than NIUs. 6 Participants with higher initial EB demonstrate greater glucose reductions than those with lower initial EB. Interestingly, reductions in glucose levels are unrelated to frequency or intensity of App use or glucose testing during the study period, and, as reported earlier, frequency of use and testing decrease steadily and substantively from baseline to week 16. 6 It may be the case that once participants realize that they are becoming successful in reducing glucose levels, they may not require the same frequency of App use and testing to maintain their success over time.
Two major factors may have contributed to these notably positive results, especially in light of the high attrition rates and mixed findings of previous studies.3,11,12 First, about 46% of participants report that they elected to use the App because it “came with the meter,” and another 30% sought out the App “on my own.” 6 Thus, from the outset, more than three quarters of participants appear to be independent, self-motivated, and self-engaged in the task of reducing their glucose levels on their own without necessarily undertaking this task at the instigation of or in collaboration with a health care provider (only 10% reported that their provider had recommended the App). Furthermore, the higher initial EB among those who are most successful in reducing their glucose levels may stimulate greater motivation and engagement to reduce glucose levels. 13 Second, participants most likely know how to use the individual and patterned glucose data that the App provides (very few participants utilize the carb counting and exercise functions). 6 The App does not include specific suggestions for lowering glucose levels once glucose measures are obtained, which has been shown previously to be a crucial component of effective glucose feedback.14,15 Thus, use of this free-standing App and participation in this study attracted a well-defined subgroup of T2 adults: they are tech savvy and tech trusting, 6 are highly motivated to address glucose problems on their own, and they have the resources or knowledge about how to make use of the App displays to reduce glucose levels.16,17 These findings suggest that even a relatively straightforward, uncomplicated technology can be effectively utilized by a particular sub group of the adult T2D population. This finding contrasts with the generally negative or mixed effects of similar and often far more complicated technologies when applied to the T2D population in general. Given the heterogeneity of adults with T2D, technologies may be best and most efficiently used when they are tailored to meet the specific needs, motivations, and resources of different groups of adults with T2D, rather than considering a one-size-fits-all approach to all adults with T2D. 5
Significant, although modest, improvements in quality of life and reductions in both concerns about hypoglycemia and EB also occur across the 16-week study period. Furthermore, given these total sample findings, in general, younger participants and IUs report greater improvements than older adults and NIUs. IUs tend to be more dependent on glucose data than NIUs, and younger adults with T2D may be more tech savvy and tech trusting than older T2Ds.
It is interesting to note that, by and large, although the largest reductions in glucose levels occur between baseline and six weeks, the psychosocial improvements recorded by the survey scales occur primarily between weeks six and 16. These findings underscore how changes in psychosocial and emotional experiences often follow changes in diabetes-related management. Individuals with diabetes often feel less concerned and burdened by their chronic disease once they demonstrate that they can affect major management changes that have concrete, observable, disease-related consequences. 18 This finding suggests that changes in psychosocial outcomes following management changes may require longer time periods to observe than is usually considered. Without such consideration, the risk of false-negative results increases.
This study has several strengths: it employed a large, unselected group of participants, it assessed change over multiple time periods, and, through the survey assessment, it provided a glimpse into the real-world lived experience of adults with T2D. Several limitations should be kept in mind. First, recruiting a self-selected sample of individuals who elected to use the App in a real-world study prevented the inclusion of a control group that did not receive the App. Thus, between-group differences in glucose and psychosocial changes over time could not be determined. Furthermore, since this was a real-world study, we had no access to the frequency of glucose testing or to average weekly glucose levels prior to downloading the App. Thus, change due to the introduction of the App could not be assessed directly. Second, no information is available to document additional factors that might have influenced the significant changes in the glucose levels we observe, such as links to health care providers and availability of other diabetes-related resources to assist with glucose management. Third, the study assessed change only over 16 weeks. The time lag between changes in glucose levels and changes in psychosocial variables suggests that longer-term follow-up may provide important, additional information.
Conclusions
We find that making a basic App for monitoring glucose widely and easily available for download with a glucose meter can be helpful for some adults with T2D to reduce their glucose levels. Doing so, however, assumes that users are sufficiently motivated and engaged, that they are comfortable and trusting of the technology, and that they have sufficient knowledge of how to make use of the glucose data provided by the App. These findings also suggest the need to tailor diabetes technologies to the different needs, motivations, and preferences of different groups of adults with T2D and that, with sufficient motivation and engagement, many adults with T2D can profit from relatively straightforward technological aids to assist with glucose management.
Footnotes
Acknowledgements
Appreciation is expressed to Caterina Florissi, Chester Lu, James Richardson, Keaton Stoner, and Richard Wood for their support throughout this project.
Abbreviations
IU, insulin using; NIU, non-insulin using; App, applications; EB, emotional burden; RD, regimen distress; REALL, Real-World Evidence of Diabetes App Use in Daily Life; SD, standard deviation; T2D, type 2 diabetes; T1D, type 1 diabetes.
Authorship Contribution
All authors contributed to the implementation of the study. LF and AF designed the analyses and wrote the initial manuscript, JK and AS reviewed and edited the manuscript.
Declaration of Conflicting Interests
The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Jennifer Knaebel and Andreas Stuhr are full-time employees of Ascensia Diabetes Care.
LF is a consultant to Eli Lilly and has received research funding from Eli Lilly. JK and AS are full-time employees of Ascensia Diabetes Care. No conflict of interest was reported by any author.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was funded by an unrestricted research grant from Ascensia Diabetes Care.
