The Use of Mobile Applications Among Adults with Type 1 and Type 2 Diabetes: Results from the Second MILES

Abstract

Background:

While the number of diabetes-specific mobile applications (apps) continues to grow, there is a lack of knowledge about their actual use.

Methods:

The second MILES (Management and Impact for Long-term Empowerment and Success)—Australia study was a national cross-sectional survey of the psychological, behavioral, and social aspects of diabetes for adults with type 1 diabetes (T1D) and type 2 diabetes (T2D). Associations between diabetes-specific app usage and demographic, clinical, and psychosocial variables were examined.

Results:

Of the 1589 respondents responding to the diabetes-specific app questions, 795 had T1D (mean ± standard deviation age 43 ± 14 years; 61% women; diabetes duration 19 ± 14 years) and 794 had T2D (age 60 ± 9 years; 40% women; diabetes duration 11 ± 7 years). Among adults with T1D, 24% (n = 188) reported using apps, with carbohydrate counting (74%; n = 139) as the most common cited purpose. App usage was significantly associated with shorter diabetes duration, more frequent glucose monitoring, and lower self-reported HbA1c. Among adults with T2D, 8% (n = 64) reported using apps, with glucose monitoring (62%; n = 39) as the most common purpose. For all respondents, the most commonly reported reason for not using apps was a belief that they could not help with diabetes self-management.

Conclusions:

A minority of adults with T1D and T2D use apps to support their self-management. App use among adults with T1D is associated with a more recent T1D diagnosis, more frequent glucose monitoring, and lower self-reported HbA1c. Future efforts should focus on this association and determine the mechanisms by which app use is related to better clinical outcomes.

Introduction

The prevalence of diabetes in Australia has more than doubled since 1981, with an estimated 7.4% of the population, or 1.7 million people, having some form of diabetes.¹

It is reported to be the fastest growing chronic condition in Australia and is a significant cause of disease burden.^2,3 People with a chronic condition are estimated to spend on average 6 h per year with their health professional, leaving about 8754 h in which they self-manage their condition.⁴ This self-management typically includes monitoring glycemic levels, taking medications as recommended, initiating and maintaining lifestyle changes, and coping with the physical and psychosocial consequences of the condition. The combination of these self-management activities can be a tremendous burden on the individual and it is no surprise that many people are not able to reach or maintain the recommended glycemic targets.^5,6

Interest in the potential for mobile health applications (apps) to support self-management of health has grown because they are accessible, portable, low-cost, convenient to the user, and have wide reach.⁷ For people with diabetes, apps can support a broad range of self-management activities through features such as medication reminders, carbohydrate and insulin dose calculators, food intake and physical activity trackers, glucose monitoring diaries and incentivization, and peer support. A recent industry report indicates that there are nearly 1800 diabetes-specific apps available for use.⁸ Moreover, a 2015 industry survey of over 5000 app developers showed that 70% believed diabetes to have the best market potential.⁹ The estimated global number of active users of diabetes-specific apps had increased from 2.2% in 2014 to only 3.3% in 2016, suggesting that the potential of mobile technology (commonly referred to as mHealth) in diabetes is far from being realised.⁸ The rapid proliferation of health apps and mobile phone ownership in Australia seems to suggest that the opportunity exists for adults with diabetes to utilize apps in their diabetes self-management.¹⁰

Several meta-analyses have indicated the efficacy of apps in improving self-management among adults with diabetes. In a recent meta-review of randomized controlled trials (RCTs), Hou et al. found that the use of diabetes apps resulted in a clinically significant reduction in HbA1c among adults with type 2 diabetes (T2D; average mean difference: 0.49%/0.6 mmol/mol).¹¹ Other meta-analyses of mHealth interventions for diabetes have found similar effects.^12
–14 Although the results are mixed, RCTs of app-based interventions have also demonstrated reductions in HbA1c among adults with type 1 diabetes (T1D).¹⁵ However, the majority of research in this area has focused on evaluating the efficacy of diabetes apps in the short term, in tightly controlled trial settings, and very few, if any, of those apps are freely available. Thus, little is known about the use of apps in the real world and their perceived effectiveness or usefulness outside tightly controlled research-focused conditions. The work outside efficacy trials has been largely limited to the evaluation of apps available in the Apple and Google app stores,^14,16
–18 and efforts to understand what factors influence acceptance of diabetes apps.¹⁹

To our knowledge, there are no large cross-sectional surveys of app use for self-management among adults with diabetes. A recent survey of Australian adolescents with T1D indicated that only 21% used an app to support their diabetes management.²⁰ Majority of these respondents reported carbohydrate counting as the most common purpose for using apps. A New Zealand survey of adolescents with T1D found a similar rate of use.²¹ Whether this frequency and purpose of app use will generalize to an adult population, including those with T2D, is unknown.

Gaining a better understanding of the popularity of diabetes apps, who is using them, and the kinds of apps that are being used will identify opportunities for future app development/enhancement and gaps where the needs of people with diabetes are unmet. Thus, our primary aim was to investigate the frequency of diabetes-specific app use among a sample of adults in Australia with T1D or T2D, including the name of the app, and reasons for use and nonuse. A secondary aim was to investigate associations between diabetes-specific app use and markers of physical health, self-management, and emotional well-being.

Methods

The second Diabetes MILES (Management and Impact for Long-term Empowerment and Success)—Australia study (MILES-2) was a national, online cross-sectional survey of the psychological, behavioral, and social aspects of diabetes. Eligible respondents could access the survey from 23 March to 11 May 2015. A hard copy version of the survey was available for those who requested it. The survey had a final sample of 2342 respondents (T1D = 1078; 46%; T2D = 1264; 54%) who completed a large number of measures, the full details of which have been reported elsewhere.²² The current report focuses on the subset of respondents who completed the questions on app usage.

Participants

Respondents were eligible to take part in the MILES-2 study if they were aged 18–75 years; self-reporting a diagnosis of T1D or T2D, living in Australia, and could complete the survey in English without any assistance. The primary recruitment strategy relied on accessing the National Diabetes Services Scheme (NDSS) database to contact potential participants (similar to previous MILES surveys^23,24). In addition, a variety of general recruitment strategies were employed (e.g., adverts placed in magazines, e-newsletter, and social media) to broaden the reach of the survey and attract participants.

Of the 1078 respondents with T1D, 132 did not answer the app use questions and were excluded. Respondents who reported that they did not own a suitable device to access apps or had no access to the internet were also excluded (n = 130). Finally, to allow for adjustment to living with diabetes, any respondent who had been diagnosed within the previous 12 months was also excluded (n = 21). This left a final sample of 795 eligible respondents with T1D.

Similarly, of the 1264 respondents with T2D, the following were excluded: n = 166, who did not answer the app use questions; n = 285, who reported they did not own a suitable device to access apps or had no access to the internet; and n = 19, whose T2D had been diagnosed with the previous 12 months. This left a final sample of 794 eligible respondents with T2D, of whom 342 (43%) used insulin to manage their diabetes.

Measures

The MILES-2 study included a wide range of measures that have been described in detail elsewhere.²² All data were obtained through respondent self-report. Four study-specific questions asked about whether or not the respondents used apps in their diabetes self-management; if yes, for what purposes? if no, why not? Respondents who indicated that they used apps were also asked to name them using an open text box. Wording of questions and response options are shown in Table 1.

Table 1.

Application Usage Questions and Responses from Survey Respondents with Type 1 and Type 2 Diabetes

Question	T1D (N = 795)	T2D insulin (N = 342)	T2D noninsulin (N = 454)
Do you use apps to help you with your diabetes? Yes	24% (188/795)	8% (26/342)	8% (38/454)
If yes, then …
What do you use your app(s) for? (select all that apply)
Carb counting	74% (139/188)	15% (4/26)	27% (10/37)^a
Recording blood glucose levels	36% (68/188)	69% (18/26)	57% (21/37)
Tracking my physical activity	21% (40/188)	42% (11/26)	43% (16/37)
Tracking my weight	19% (35/188)	35% (9/26)	54% (20/37)
Insulin dose calculation	19% (35/188)	4% (1/26)	0% (0/37)
Other	4% (8/188)	12% (3/26)	19% (7/37)
Monitoring my moods/well-being	4% (7/188)	8% (2/26)	3% (1/37)
Social support	2% (4/188)	4% (1/26)	8% (3/37)
Hypos/hypo symptoms	6% (11/188)	19% (5/26)	5% (2/37)
If no, then …
Why don't you use apps to help you with your diabetes? (select all that apply)
I don't think apps could help me manage my diabetes	42% (252/603)	47% (146/309)	48% (199/412)
I haven't found any good ones yet	35% (209/603)	20% (62/309)	18% (74/412)
I don't know how to use apps	13% (79/603)	29% (88/309)	21% (86/412)
Unaware of diabetes apps	5% (30/603)	7% (20/309)	9% (37/412)
Other	22% (133/603)	15% (47/309)	14% (58/412)

For each multiple choice question, the top three responses within each subgroup are shown in bold.

One participant did not provide any response for this section.T1D, type 1 diabetes; T2D, type 2 diabetes.

Self-reported demographic data included age, sex, employment status, education, and socioeconomic status (SES). The latter was determined from the respondents' postcode and based on the Index of Relative Socioeconomic Advantage and Disadvantage (IRSAD), one of the Socioeconomic Indexes for Areas (SEIFA).²⁵ Deciles of the IRSAD were used, where lower deciles indicate more disadvantage and less advantage. For the purposes of the current study, these deciles were collapsed into three SES groups; low (1–3), middle (4–7), and high (8–10).

The following self-reported clinical measures were included: diabetes duration (years), daily frequency of self-monitoring of blood glucose (SMBG), number of severe hypoglycemic events in the past 6 months (defined as “where you needed help or were unable to treat yourself”), impaired awareness of hypoglycemia (using the Gold score single item, “Do you know when your hypos are commencing?” scored on a 7-point scale, with scores ≥4 indicating impaired awareness),²⁶ and HbA1c (in the past 3 months).

Diabetes-specific emotional distress was assessed using the 20-item Problem Areas in Diabetes (PAID) questionnaire. Items are scored on a 5-point Likert scale (0 = Not a Problem to 4 = Serious Problem) and are summed and standardized to form a total score from 0 to 100, with scores ≥40 indicating greater distress.²⁷

Statistical analyses

The responses of participants with T1D and T2D were examined separately. Descriptive statistics were used to report app usage among respondents with T1D and T2D. Demographic, clinical, and psychosocial characteristics were spilt by app usage and examined using nonparametric Mann–Whitney and chi-square tests as appropriate.

Logistic regression analysis was used to determine the relationships between several demographic and clinical variables and app usage. However, logistic regression was not appropriate for adults with T2D due to the small proportion reporting app use and the absence of any significant univariate effects beyond those found for age and diabetes duration. Therefore, logistic regression analysis is only reported here for the sample of adults with T1D.

Univariate relationships significant at the 20% level were considered in the model,^28,29 and multicollinearity among these variables was checked before running the regression analysis.³⁰ Variables that were not statistically significant (P < 0.05) or did not improve the model were removed and the model was rerun. Results are reported as odds ratios (ORs) with 95% confidence intervals (CIs). To avoid overfitting the regression model, the event per variable (EPV) was calculated. The EPV is taken as the ratio of outcome events over the number of predictor variables and is expected to be at least 10.³¹ The Hosmer and Lemeshow test was used to assess model fit.²⁸ All statistical analyses were conducted using R, version 3.3.2.³²

Results

Sample characteristics

Of the 1591 eligible respondents, 795 had T1D and 796 had T2D. Of those with T1D, 61% (n = 482) were women and the mean ± standard deviation (SD) age was 43 ± 14 years (minimum = 18, maximum = 75). Two percent (n = 12) of respondents identified as being of Aboriginal or Torres Strait Islander origin. Seventy-four percent (n = 588) were employed and 11% (n = 89) were retired, with the remainder in education, unemployed, or not reporting their employment status. Their average duration of T1D was 19 ± 14 years (minimum = 1, maximum = 68), with an average HbA1c of 7.3% ± 1.3% (56.9 ± 14.7 mmol/mol). Of the 63% (n = 500) who self-reported their HbA1c, 63% (n = 314) reported an HbA1c of ≥7.0% (≥53.0 mmol/mol) and 8% (n = 42) reported an HbA1c of ≥9.0% (≥75.0 mmol/mol). Thirty-nine percent (n = 313) used an insulin pump and the majority checked their blood glucose four to six times a day (51.8%; n = 411).

Of those with T2D, 40% (n = 321) were women and the mean ± SD age was 60 ± 9 years (minimum = 30, maximum = 75). Two percent (n = 12) identified as being of Aboriginal or Torres Strait Islander origin. Forty-one percent (n = 328) were employed and 42% (n = 334) were retired, with the remainder in education, unemployed, or not reporting their employment status. Their average duration of T2D was 11 ± 7 years (minimum = 1, maximum = 44), with an average HbA1c of 7.2% ± 1.7% (55.2 ± 18.3 mmol/mol). Of the 54% (n = 428) who self-reported their HbA1c, 49.0% (n = 211) reported an HbA1c of ≥7.0% (≥53.0 mmol/mol) and 11% (n = 47) reported an HbA1c of ≥9.0% (≥75.0 mmol/mol). Forty-three percent (n = 341) used insulin to manage their diabetes.

Respondents with T2D were significantly older than those with T1D (60 ± 9 years vs. 43 ± 15 years; P < 0.001) and, relatedly, were more likely to be retired (79% vs. 21%; P < 0.001). Respondents with T2D also reported a lower level of education with 67% having a primary level or less compared with 21% of those with T1D (P < 0.001), and a larger number of respondents with T2D reported residential postcodes from a lower SES area than those with T1D (61% vs. 39%; P < 0.001).

App usage among adults with T1D

Overall, 24% (n = 188) of respondents indicated that they used apps to help support their diabetes management (Table 2). A total of 69 different apps were reported (respondents could name multiple apps). The Australian food database app, CalorieKing, was the most frequently reported app, used by 43% (n = 80) of app users with T1D. Another food database app, MyFitnessPal, was second, reported by 9% (n = 17). The popularity of apps that provide food information complements data presented in Table 1, indicating that the most common purpose for using apps was carbohydrate counting (74%; n = 139). Carbohydrate counting was higher among respondents using a pump (80%; n = 80) than those injecting their insulin (67%; n = 59) (χ² = 6.30, df = 1, P = 0.012). Of the 76% (n = 607) not using apps, the most commonly reported reason was due to a belief that apps would not help with their diabetes management (42%; n = 262).

Table 2.

Characteristics of Survey Respondents with Type 1 Diabetes by Current Diabetes Application Usage Status (N = 795)

Variable	N missing	Total sample (N = 795)	No app usage (N = 607)	App usage (N = 188)	Test statistic^a	P
Sex, female	2	61% (482/793)	60% (361/605)	64% (121/188)	χ² = 1.14	0.287
Age, years	0	44 ± 15	44 ± 15	40 ± 14	z = −3.29	<0.001
Education	3				χ² = 4.13	0.127
Primary or less		11% (88/792)	12% (74/606)	8% (14/186)
Secondary or trade		39% (309/792)	39% (239/606)	38% (70/186)
Tertiary		50% (395/792)	48% (293/606)	55% (102/186)
Employment	0				χ² = 3.55	0.17
In paid work		74% (588/795)	73% (442/607)	78% (146/188)
Retired		11% (89/795)	12% (75/607)	7% (14/188)
Other		15% (118/795)	15% (90/607)	15% (28/188)
Socioeconomic status	7				χ² < 1	0.914
High		52% (411/788)	53% (317/603)	51% (94/185)
Medium		33% (262/788)	33% (199/603)	34% (63/185)
Low		15% (115/788)	14% (87/603)	15% (28/185)
Diabetes duration, years	0	18 ± 13	19.6 ± 14	16 ± 13	z = −2.73	<0.001
Primary diabetes treatment	0				χ² = 18.95	<0.001
Insulin pump		39% (313/795)	35% (213/607)	53% (100/188)
Insulin injections		61% (482/795)	65% (394/607)	47% (88/188)
Blood glucose checks, per day	1				χ² = 17.39	<0.001
Three or less		27% (215/794)	31% (186/607)	16% (29/187)
4–6		52% (411/794)	50% (303/607)	58% (108/187)
Seven or more		21% (168/794)	19% (118/607)	27% (50/187)
Self-reported HbA1c in past 3 months, % (mmol/mol)	295	7.3 ± 1.3 (56.1 ± 13.6)	7.4 ± 1.4 (57.7 ± 15.5)	7.1 ± 1.1 (54.5 ± 11.6)	z = −1.4	0.08
Impaired awareness of hypoglycemia: gold score ≥4	24	26% (199/771)	26% (151/585)	26% (48/186)	χ² < 1	>0.999
Severe hypoglycemia: ≥1 event in past 6 months	24	16% (122/771)	16% (94/585)	15% (28/186)	χ² < 1	0.83
Diabetes-specific distress: PAID total score	107	25 ± 21.1	25 ± 21	24.0 ± 20.7	z = 0.04	0.52
Diabetes-specific distress: severe (PAID total score ≥40)	107	24% (162/688)	24% (122/517)	23% (40/171)	χ² < 1	>0.999

PAID: total score range 0–100, with greater scores indicating greater distress.

Statistic and P-value refers to group comparison between app and non-app users.

app, application; PAID, Problem Areas in Diabetes.

App usage among adults with T2D

Overall, 8% (n = 64) of respondents with T2D indicated that they used apps to support their diabetes management (Table 3). A total of 68 different apps were reported. The exercise-tracking app, Fitbit, was the most frequently reported app, used by 9% (n = 6) of app users with T2D. However, unlike the respondents with T1D, there was no single app used by the majority. To highlight the differences in app use as a function of treatment, Table 1 shows responses split by those using insulin to manage their diabetes and those not using insulin. Although both treatment groups indicated that recording blood glucose levels was their main purpose for using apps, weight tracking was reported more frequently by those not using insulin (54% vs. 35%). For both treatment groups, the most commonly reported reason for not using apps was a belief that apps would not help with their diabetes management (48%; n = 345).

Table 3.

Characteristics of Survey Respondents with Type 2 Diabetes by Current Diabetes Application Usage Status (N = 794)

Variable	N missing	Total sample (N = 794)	No app usage (n = 730)	App usage (n = 64)	Test statistic^a	P
Sex, female	2	41% (321/792)	40% (292/728)	45% (29/64)	χ² = 0.49	0.485
Age, years	2	58 ± 10	61 ± 9	55 ± 10	z = −4.01	<0.001
Education	1				χ² = 4.29	0.117
Primary or less		23% (181/793)	24% (171/729)	16% (10/64)
Secondary or trade		42% (329/793)	42% (305/729)	38% (24/64)
Tertiary		36% (283/793)	35% (253/729)	47% (30/64)
Employment	0				χ² = 4.5	0.106
In paid work		41% (328/794)	40% (295/730)	52% (33/64)
Retired		42% (334/794)	43% (315/730)	30% (19/64)
Other		17% (132/794)	16% (120/730)	19% (12/64)
Socioeconomic status	10				χ² = 0.05	0.974
High		42% (329/784)	42% (303/720)	41% (26/64)
Medium		35% (277/784)	35% (254/720)	36% (23/64)
Low		23% (178/784)	23% (163/720)	23% (15/64)
Diabetes duration, years	2	10 ± 7	11 ± 7	9 ± 6	z = −2.07	0.02
Primary diabetes treatment	0				χ² = 0.07	0.795
Noninsulin		57% (453/794)	57% (415/730)	59% (38/64)
Insulin		43% (341/794)	43% (315/730)	41% (26/64)
Blood glucose checks, per day	1				χ² = 0.84	0.358
Three or less		89% (703/793)	89% (649/729)	84% (54/64)
Four or more		11% (90/793)	11% (80/729)	16% (10/64)
Self-reported HbA1c in past 3 months: % (mmol/mol)	368	7.2 ± 1.7 (54.5 ± 18.3)	7.2 ± 1.7 (55.3 ± 18.3)	7.1 ± 1.7 (53.6 ± 18.2)	z = −0.47	0.32
Diabetes-specific distress: PAID total score	122	19 ± 19	19 ± 20	19 ± 18	z = −0.18	0.43
Diabetes-specific distress: severe (PAID total score ≥40)	122	15% (103/672)	15% (96/624)	15% (7/48)	χ² = <1	>0.999

PAID: total score range 0–100, with greater scores indicating greater distress.

Statistic and P-value refers to group comparison between app and non-app users.

Univariate and multivariate correlates of app usage among adults with T1D

Univariate analysis indicated that those who used apps were younger and were diagnosed more recently than those who did not use apps (P < 0.001; Table 4). Comparing three age categories (18–25, 26–55, and ≥56 years), although not statistically significant, the youngest group had the largest percentage of app use (38%; 24/63), followed by the 26–55-year-olds (34%; 127/379) and ≥56 years (22%; 37/165) age groups (χ² = 4.52, df = 2, P = 0.105).

Table 4.

Logistic Regression with Application Usage as Dependent Variable for Respondents with Type 1 Diabetes (N = 419)

Variable	Odds ratio (95% confidence interval)	P
Diabetes duration, years	0.97 (0.96–0.99)	0.002
Insulin delivery mode
Insulin injections	Ref.
Insulin pump	1.54 (0.98–2.43)	0.064
Blood glucose checks, per day
Three or less	Ref.
4–6	3.03 (1.6–6.09)	0.001
Seven or more	2.15 (1.01–4.74)	0.05
Self-reported HbA1c in past 3 months	0.81 (0.66–0.97)	0.027

There was also a significant relationship between app usage and insulin delivery mode, with more app users among respondents using an insulin pump (P < 0.001). Duration of diabetes was also significantly associated with app use (P < 0.001). Finally, a higher frequency of SMBG was observed among app users (P < 0.001).

The seven variables that were significant at trend level (P ≤ 0.2) in the univariate analysis (age, education, employment, diabetes duration, self-reported HbA1c, insulin delivery mode, and SMBG frequency) were entered into the logistic regression model. Three of the original seven variables remained significant in the final model (Table 4). The variables that were associated significantly with app usage were shorter diabetes duration (OR 0.97, CI 0.96–0.99, P = 0.002), four to six or seven or more SMBG per day compared with three or less (OR 3.03, CI 1.6–6.09, P = 0.001; OR 2.15, CI 1.01–4.74, P = 0.05), and a lower self-reported HbA1c (OR 0.81, CI 0.66–0.97, P = 0.027). The model correctly classified 81% of the respondents, and the Hosmer–Lemeshow test was nonsignificant (χ² = 10.04, df = 8, P = 0.62), indicating that the model fit the data well. The EPV was >10 (118/4), indicating that the model could be estimated accurately.

Discussion

In this first national survey of app usage among adults with diabetes, we found that nearly a quarter of respondents with T1D, but only 8% of those with T2D, used apps to support their diabetes self-management. Almost three-quarters of app users with T1D used apps to support their carbohydrate counting, whereas people with T2D were more likely to use apps to record their glucose levels and monitor their weight and activity levels.

The lower use of apps among adults with T2D is interesting as previous meta-analysis has indicated that these individuals showed greater improvement in glycemic control than those with T1D in 22 experimental studies on mobile phone interventions.¹² It may be that the benefit is greater among people with T2D, but outside an experimental trial, this potential is lost due to lack of real-world uptake of and engagement with apps. Further work is required to substantiate this as it may be the case that these individuals have tried (untested and unproven) apps that are widely available, but did not find them helpful.

There are several potential explanations for why app use was higher (24%) in respondents with T1D when compared with people with T2D (8%). First, respondents with T2D in the sample were older, more likely to be retired, and from a lower SES area than those survey respondents with T1D. This is in line with other research indicating that health app use is linked to age and socioeconomic status.^33,34 Related to this is the observation that more respondents with T2D were excluded from this study for not having a mobile device or internet access than those with T1D. Second, people with T1D and T2D may differ in terms of their treatment needs and trajectories. Newly diagnosed people with T1D may have an immediate and urgent need to adjust their dose of insulin based on their carbohydrate intake, which may prompt them to look for an app that helps them do this. The multiple lifestyle changes recommended to newly diagnosed people with T2D, although beneficial in the long term, are not urgent and do not have immediate consequences. It is also possible that the currently available apps do not tackle the multiple and complex behavior changes that are an important part of the treatment regime for people with T2D. It is clear that more research is needed into how the needs of people with T1D and T2D differ in terms of the support that apps can potentially provide.

Of the respondents who did not use apps, the primary reason given (regardless of diabetes type or treatment) was that they did not feel that apps would help with their diabetes management. This may be further indication that although the number of diabetes apps available on the market is increasing exponentially, they are not meeting the needs of people with diabetes. Relatedly, the most popular apps used for diabetes self-management were not diabetes specific. More research is needed into understanding how the needs of people with diabetes could be supported by an app. Additionally, the majority of apps available are unregulated, not adequately evaluated,³⁵ and are often not endorsed or promoted by health professionals.³⁶ More research is needed to understand the barriers and enablers that drive engagement with diabetes apps. An important aspect of this is the potential to integrate consumer input into the app development process to ensure that products are developed that meet expressed needs.^37
–39

In regard to our second aim, the significant relationship between app use among respondents with T1D and lower self-reported HbA1c is consistent with previous research.^12,13,40 This leads to an important question: To what extent can this association with a reduction in HbA1c be explained by carbohydrate counting? That is, could the benefits of carbohydrate counting⁴¹ underlie the relationship between app use and lower HbA1c? Unfortunately, this survey did not ask whether respondents were actively engaged in carbohydrate counting, nor could we draw causal conclusions from our cross-sectional data even if this information was available. This would be an important aspect to examine in future work as including this in a regression model may or may not eliminate the significant relationship between app use and HbA1c.⁴² The significant relationship between SMBG and app use in respondents with T1D is consistent with a previous study of Australian adolescents with T1D.²⁰ This association between SMBG and app use could be explained by the finding that glucose recording was the second most cited purpose for using apps among adults with T1D. However, elucidating this will require further work. Similarly, it is unclear why we found a significant relationship with diabetes duration and app use. One possibility is that in the early stages following diagnosis, adults with T1D are seeking out and utilizing a variety of treatment and support options, including apps, to see what is effective for them. It could be that only a minority continue using apps in the long term, therefore resulting in this association with diabetes duration. It should be noted that this association between app use and diabetes duration has previously been observed among Australian adolescents with T1D.²⁰

The lack of significant associations with app use, beyond age and diabetes duration, among respondents with T2D was unexpected. It may indicate that the questions we asked were insufficient to understand what influences app use in that population. Furthermore, the survey did not address the motivation for using apps, for example, was an app suggested by their health professional or recommended by a peer? Understanding such motivations could inform efforts to encourage app use in future studies.

This study has several limitations, which have been explored in detail elsewhere.²² In particular, we acknowledge that these data are subject to issues of self-report and selection biases and the questions about app use were necessarily study specific and unvalidated.

A detailed description of the sample representativeness of the MILES-2 study can be found elsewhere.²² It should be noted that the percentage of respondents with a self-reported HbA1c of 7.0% or above is higher than recent Australian primary care-based estimations.⁴³ However, with respect to self-selection, only 10% of the respondents reported an HbA1c of ≥9%. Furthermore, we found more women in the T1D sample than reported in the NDSS register (61% vs. 48%, respectively).⁴⁴

Another limitation of the current research is that despite the large overall sample, the number of people with T2D reporting app use was insufficient to conduct further statistical analysis. More research is needed to understand the associations between app use and self-reported diabetes-related health outcomes. One of the strengths of this study is that the data were derived from a nationwide survey, sampling adults with T1D and T2D, with the same questions on their app use to allow group comparison. It provides the first insight into how frequently apps are used to support diabetes self-management and the similarities and differences between adults with T1D or T2D.

In conclusion, the Diabetes MILES-2 study has found that a minority of adults with T1D and T2D use apps to support their self-management. Importantly, app use among adults with T1D is associated with a lower self-reported HbA1c. The association between HbA1c and app use might either be causal, irrelevant, or a consequence of using apps for carbohydrate counting. It appears that a belief that apps could not help was the main reason for not using apps, which highlights a potential avenue for future efforts to encourage diabetes-specific app use among adults with T1D and T2D. Such research needs to adopt a participatory design process to ensure that new developments in this field meet the expressed needs of intended users.

Footnotes

Author Disclosure Statement

No competing financial interests exist.

References

Tanamas

, Magliano

, Lynch

, et al.: AusDiab 2012. AusDiab: The Australian Diabetes, Obesity and Lifestyle Study. Melbourne: Baker IDI Heart and Diabetes Institute, 2013.

AIHW: Australian Burden of Disease Study: Impact and Causes of Illness and Death in Australia 2011. Canberra: AIHW, 2016.

Huo

, Shaw

, Wong

, et al.: Burden of diabetes in Australia: life expectancy and disability-free life expectancy in adults with diabetes. Diabetologia, 2016; 59:1437–1445.

Al-Ubaydli

: Personal Health Records: A Guide for Clinicians. John Wiley & Sons, UK; 2011.

Laiteerapong

, Fairchild

, Chou

C-H

, et al.: Revisiting disparities in quality of care among US adults with diabetes in the era of individualized care, NHANES 2007–2010. Med Care, 2015; 53:25.

Mendes

ABV

, Fittipaldi

JAS

, Neves

RCS

, et al.: Prevalence and correlates of inadequate glycaemic control: results from a nationwide survey in 6,671 adults with diabetes in Brazil. Acta Diabetol, 2010; 47:137–145.

Dimitrov

: Medical internet of things and big data in healthcare. Healthc Inform Res, 2016; 22:156–163.

Research2Guidance: Diabetes App Market Report 2016–2021. Research2Guidance, Berlin; 2016.

Research2Guidance: mHealth App Developer Economics 2015: The Current Status and Trends of the mHealth App Market. Research2Guidance, Berlin; 2015.

10.

Australian Communication and Media Authority: Communications Report 2011–2012 Series. Report 3—Smartphones and Tablets Take-Up and Use in Australia. ACMA, Melbourne; 2013.

11.

Hou

, Carter

, Hewitt

, et al.: Do mobile phone applications improve glycemic control (HbA1c) in the self-management of diabetes? A systematic review, meta-analysis, and GRADE of 14 randomized trials. Diabetes Care, 2016; 39:2089–2095.

12.

Liang

, Wang

, Yang

, et al.: Effect of mobile phone intervention for diabetes on glycaemic control: a meta‐analysis. Diabet Med, 2011; 28:455–463.

13.

Saffari

, Ghanizadeh

, Koenig

: Health education via mobile text messaging for glycemic control in adults with type 2 diabetes: a systematic review and meta-analysis. Prim Care Diabetes, 2014; 8:275–285.

14.

, Yao

, Vespasiani

, et al.: Mobile app-based interventions to support diabetes self-management: a systematic review of randomized controlled trials to identify functions associated with glycemic efficacy. JMIR Mhealth Uhealth, 2017; 5:e35.

15.

Kirwan

, Vandelanotte

, Fenning

, Duncan

: Diabetes self-management smartphone application for adults with type 1 diabetes: randomized controlled trial. J Med Internet Res, 2013; 15:e235.

16.

BinDhim

, Hawkey

, Trevena

: A systematic review of quality assessment methods for smartphone health apps. Telemed J E Health, 2015; 21:97–104.

17.

Chomutare

, Fernandez-Luque

, Årsand

, Hartvigsen

: Features of mobile diabetes applications: review of the literature and analysis of current applications compared against evidence-based guidelines. J Med Internet Res, 2011; 13:e65.

18.

Demidowich

, Lu

, Tamler

, Bloomgarden

: An evaluation of diabetes self-management applications for Android smartphones. J Telemed Telecare, 2012; 18:235–238.

19.

Scheibe

, Reichelt

, Bellmann

, Kirch

: Acceptance factors of mobile apps for diabetes by patients aged 50 or older: a qualitative study. Med 2.0, 2015; 4:e1.

20.

Trawley

, Browne

, Hagger

, Hendrieckx

, Holmes-Truscott

, Pouwer

, Skinner

, Speight

: The use of mobile applications among adolescents with type 1 diabetes: results from Diabetes MILES Youth—Australia. Diabetes Technol Ther, 2016; 18:813–819.

21.

Dobson

, Whittaker

, Murphy

, et al.: The use of mobile health to deliver self-management support to young people with type 1 diabetes: a cross-sectional survey. JMIR Diabetes, 2017; 2:e4.

22.

Browne

, Holmes-Truscott

, Ventura

, et al.: Cohort profiles of the cross-sectional and prospective participant groups in the second Diabetes MILES—Australia (MILES-2) study. BMJ Open, 2017; 7:e012926.

23.

Hagger

, Trawley

, Hendrieckx

, et al.: Diabetes MILES Youth–Australia: methods and sample characteristics of a national survey of the psychological aspects of living with type 1 diabetes in Australian youth and their parents. BMC Psychol, 2016; 4:42.

24.

Speight

, Browne

, Holmes-Truscott

, et al.: Diabetes MILES-Australia (Management and Impact for Long-term Empowerment and Success): methods and sample characteristics of a national survey of the psychological aspects of living with type 1 or type 2 diabetes in Australian adults. BMC Public Health, 2012; 12:120.

25.

Pink

: Information paper: an introduction to socio-economic indexes for areas (SEIFA), 2006. Canberra: Australian Bureau of Statistics (ABS); 2008.

26.

Gold

, Macleod

, Frier

: Frequency of severe hypoglycemia in patients with type I diabetes with impaired awareness of hypoglycemia. Diabetes Care, 1994; 17:697–703.

27.

Polonsky

, Anderson

, Lohrer

, et al.: Assessment of diabetes-related distress. Diabetes Care, 1995; 18:754–760.

28.

Hosmer

Jr. , Lemeshow

: Applied Logistic Regression. John Wiley & Sons, Hoboken, NJ; 2004.

29.

Maldonado

, Greenland

: Simulation study of confounder-selection strategies. Am J Epidemiol, 1993; 138:923–936.

30.

Pregibon

: Logistic regression diagnostics. Ann Stat, 1981; 9:705–724.

31.

Peduzzi

, Concato

, Kemper

, et al.: A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol, 1996; 49:1373–1379.

32.

R Core Team: R: A Language and Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing, 2016.

33.

Ernsting

, Dombrowski

, Oedekoven

, et al.: Using smartphones and health apps to change and manage health behaviors: a population-based survey. J Med Internet Res, 2017; 19:e101.

34.

Krebs

, Duncan

: Health app use among US mobile phone owners: a national survey. JMIR Mhealth Uhealth, 2015; 3:e101.

35.

Boulos

MNK

, Brewer

, Karimkhani

, et al.: Mobile medical and health apps: state of the art, concerns, regulatory control and certification. Online J Public Health Inform, 2014; 5.

36.

Hingle

, Patrick

: There are thousands of apps for that: navigating mobile technology for nutrition education and behavior. J Nutr Educ Behav, 2016; 48:213–218.e211.

37.

Baptista

, Oldenburg

, O'Neil

: Response to “Development and Validation of the User Version of the Mobile Application Rating Scale (uMARS).”. JMIR Mhealth Uhealth, 2017; 5:e16.

38.

Fishbein

, Nisotel

, MacDonald

, et al.: Mobile application to promote adherence to oral chemotherapy and symptom management: a protocol for design and development. JMIR Res Protoc, 2017; 6:e62.

39.

Modave

, Bian

, Rosenberg

, et al.: DiaFit: the development of a smart app for patients with type 2 diabetes and obesity. JMIR Diabetes, 2016; 1:e5.

40.

O'Neil

, Cocker

, Rarau

, et al.: Using digital interventions to improve the cardiometabolic health of populations: a meta-review of reporting quality. J Am Med Inform Assoc, 2017; 4:867–879.

41.

Bantle

, Wylie-Rosett

, Albright

, et al.: Nutrition recommendations and interventions for diabetes: a position statement of the American Diabetes Association. Diabetes Care, 2008; 31:S61–S78.

42.

Bell

, Barclay

, Petocz

, et al.: Efficacy of carbohydrate counting in type 1 diabetes: a systematic review and meta-analysis. Lancet Diabetes Endocrinol, 2014; 2:133–140.

43.

Carrington

, Stewart

, Gerber

, Cohen

: Bittersweet Findings of Blood Glucose Levels in 467 955 Patients in Primary Care. Melbourne: Baker IDI Heart and Diabetes Institute, 2015.

44.

National Diabetes Services Scheme: Diabetes Map. www.diabetesmap.com.au (accessed August 1, 2017 ).

The Use of Mobile Applications Among Adults with Type 1 and Type 2 Diabetes: Results from the Second MILES—Australia (MILES-2) Study

Abstract

Background:

Methods:

Results:

Conclusions:

Introduction

Methods

Participants

Measures

Statistical analyses

Results

Sample characteristics

App usage among adults with T1D

App usage among adults with T2D

Univariate and multivariate correlates of app usage among adults with T1D

Discussion

Footnotes

Author Disclosure Statement

References