This cross-sectional observational study used two surveys (see Multimedia Appendices 1 and 2), one for people with diabetes attending a secondary care diabetes outpatient clinic and the second for HPs (who treat people with diabetes) attending a national diabetes conference. Both surveys were multi-choice format, collected, and managed using REDCap electronic data capture tools. REDCap (Research Electronic Data Capture) is a secure, Web-based app designed to support data capture for research studies [24]. The survey questions were derived from criteria in the Mobile app rating scale [25] to address attitudes and practices of both the people with diabetes and HPs. The list of apps was compiled by searching Apple and Android App stores and included the first consecutive ten diabetes apps. We eliminated any apps not specific to diabetes by reviewing app store descriptions. We reviewed the main features from these apps to develop the list of app features. The patient survey asked responders to select any useful app features from a list. Responders could select more than one useful app feature. The HP survey listed app features and used a scale to assess usefulness of app features (from 1 [not at all useful] to 5 [extremely useful]) and their confidence in recommending apps (from 1 [not at all confident] to 5 [extremely confident]).

The 1177 people with diabetes attending clinics at Capital and Coast District Health Board (CCDHB), Wellington, New Zealand over a 12-month period (10th September 2014 to 10th September 2015) were the sample population. Out of the total patients, 521 patients with an email address in the hospital management system were invited to participate via email. To include a representation of people without a recorded email address in the sample (n=656), every 5th person was telephoned (up to twice) and invited to provide an email address. Of the 131 patients telephoned, 54 (41.2%) were reached, of whom 49 (91%) agreed to participate. Patients without phone numbers or unable to provide an email address were excluded. This generated a sample population of 570 people.

The survey was piloted with the first 30 patients with an email addresses (chronological order of clinic visits). Responses were reviewed after response rate reached 50%. As 4 questions were unanswered by some participants, a “none of the above” option was added. The invitations were sent out to the remaining 540 participants. A further 31 participants were excluded (4 email address errors, 13 gestational diabetes, 10 deceased, 4 did not have diabetes) resulting in a final total of 539 participants. This survey remained open for 3 weeks, with reminders sent to non-responders at one week and two weeks.

Additional data on all patients were collected from the hospital management system, including age, and the most recent values within the previous 12 months from date of survey for blood pressure (BP), glycated hemoglobin (HbA_1c), urinary microalbumin to Creatinine ratio (ACR), low density lipoprotein cholesterol (LDL), and total cholesterol to HDL ratio (C:HDL). Prescription of lipid lowering drugs, anti-hypertensive drugs, insulin, or other hypoglycemic medication were also extracted from the medication list from the last visit within the sample period. Type of diabetes was self-reported in the survey (type 1 [T1DM], type 2 [T2DM], other or unknown) and in four participants who had selected ‘other’ or ‘unknown’ diabetes type was determined by examination of the clinical records. For categorization of participants by app use, 4 responders who did not indicate if they had a mobile phone or not were included in the non-app group.

To obtain data on HPs’ knowledge and recommendation of apps to people with diabetes, a second survey was conducted of the HPs attending the annual scientific meeting of the New Zealand Society for the Study of Diabetes (NZSSD) in May 2016. Immediately prior to the meeting all registered attendees (n=286) were invited to participate in the online survey via email. The data from the patient survey was presented at the conference in a 15-min oral presentation and attendees were encouraged to complete the survey. Paper copies of the survey were also available at the meeting. This survey remained open for 2 weeks, with a reminder sent at 1 week.

Data were imported into SPSS version 24 (IBM). Incomplete responses were included in the analysis. In the patient survey, independent sample t tests were conducted to compare mean clinical variables (age, BP, C:HDL, LDL, HbA_1c) by type of diabetes, method of recruitment, and whether the responder used a diabetes mobile phone app. Adjustment was made for unequal variances. Normal distribution was assumed for all variables, apart from urinary microalbumin to creatinine for which a Wilcoxin test was used. No statistically significant differences in these variables or in mobile phone app use were found between patients with recorded email addresses and patients phoned for their email address. Therefore, all 189 responses were combined for further analysis. Chi-square tests were used to compare medications and survey responses by type of diabetes. Statistical significance was determined by exact 2-sided P values less than .05. In the HP survey, mean values on the usefulness and confidence Likert scales were calculated to compare app features.

In this large sample of people with diabetes attending a secondary care clinic in NZ, 19.6% (37/189) of patients reported using diabetes apps to support their self-management. Diabetes app users were younger and more often had T1DM. The most used app feature in current app users was a blood glucose diary (87%, 32/37). The most desirable feature of a future app was an insulin dose calculation function in app users (46%) and a blood glucose diary in non-app users (64.4%). A Scottish survey has reported similar results and observed that people with T1DM were more likely to desire insulin calculators in an app [23].

Almost two-thirds of HPs responding had recommended a diabetes app to patients. Dieticians were more likely to recommend an app than others. Blood glucose and carbohydrate diaries were considered the most useful feature and HPs were most confident to recommend blood glucose diaries. HPs are the least confident recommending insulin dose calculation functions. Over one-third of HPs desire guidance with app recommendations.

Similar to a national American mHealth survey, a large proportion of patients are not using health apps [26]. However, there was a higher rate (20%) of diabetes app use in this patient group compared to the 4% found in a survey of diabetes app use in the USA in 2015 [14] and 7% in Scotland in 2016 [23]. Our findings are consistent with previous surveys showing people using apps are more likely to be younger [26]. It has been suggested that people who are more in need of diabetes care are less likely to use apps [27]; however, we found no significant difference in HbA_1c between app users and non-app users. The most favored feature being the blood glucose diary is not surprising given it is the most common feature included in the apps available [5,14]. However some responders are also using health apps that are not specific to diabetes, such as apps for dietary advice.

In contrast with the extensive app problems presented in the literature, over half of the responders with an app reported no problems [5,11-13,15]. This discrepancy may be due to false self-report or responders may have tried multiple apps before finding the one they like. Our study is unable to add significantly to literature about insulin dose calculation problems [15], as only 7 responders reported using their app for insulin calculation. However it is notable that this feature is desired by users and reinforces the importance of having a regulated environment to ensure safety.

The 60.2% of HPs in our survey who had recommended a diabetes app is significantly higher than previously documented amongst physicians across a range of specialties [28], although it is similar to HPs’ recommendation for any type of health app [19]. We did not observe any effect of HPs’ age on app recommendation, although it is previously well established that younger HPs are more likely to adopt mHealth for diabetes [28].

A large patient sample size was obtained by contacting all patients seen in the last 12 months with an email address. The risk of overrepresentation by more technology-literate responders through recruitment via email was minimized by also recruiting via telephone and by providing paper surveys at the HPs’ conference. The demographic and clinical data of responders and non-responders were compared, and most variables showed no difference. Responders were actually older than non-responders and had better glycemic control. This study focused on the beliefs and opinions of people with diabetes (potential app users) and HPs (potential app prescribers) rather than simply describing apps for diabetes . It is one of the first papers to describe app use in people with diabetes in New Zealand.

This patient sample came from patients in secondary care diabetes clinics, and therefore, app use may be different amongst patients managed in primary care. Similarly, findings may not generalize to patients with poorer glycemic control as responders had statistically significantly lower HbA_1c than non-responders. This was a cross-sectional survey that is useful to assess app use at one point in time, but it is likely that people vary their app use and recommendations over time. It was therefore not possible to assess whether the introduction of an app has significant effect on clinical outcomes. Our study did not address the difference in needs in app features between responders on insulin and those not on insulin. Overall the response rates for both surveys were low and responses were limited by self-report and therefore liable to responder bias.

This study shows app usage is relatively low among people with diabetes, while 60.2% of HPs have recommended an app to patients. There is, however, interest amongst people with diabetes and HPs to use diabetes apps, with strong interest in an insulin dose calculator. Apps with this feature have the potential to improve diabetes control. However, the critical problem of app safety remains a barrier to the prescription and use of insulin dose calculators. Further work is needed to ensure apps are safe and provided in a regulated environment. An app assessment process would provide HPs with confidence in the apps they recommend and would ultimately ensure app quality and safety for app users. At present, however, app users and HPs must remain cautious with diabetes apps, especially those in the insulin dose calculator category.