The research questions addressed in this review were as follows: (1) What is known from the existing literature about the availability, functionality, and effectiveness of mobile phone apps on GDM prevention and management? (2) What is the role of health literacy in these apps?

We selected 7 databases in consultation with a reference librarian: PubMed, Cochrane Library, Web of Science, CINAHL Complete, Communication & Mass Media Complete, Inspec, and Google Scholar. We identified articles by conducting searches using a combination of 2 sets of keywords: (1) gestational diabetes and (2) mobile, app, digital, technology, mHealth, wearable, wireless, smartphone, cell phone, telemedicine, or telecare. The combination of the first 2 sets of keywords and a third keyword, literacy, was searched in all 7 databases to identify extra articles. To retrieve the most relevant results, titles and abstracts were searched in PubMed; title, abstract, and keywords were searched for in Cochrane Library; topics were searched for in Web of Science; and abstracts were searched in CINAHL Complete, Communication & Mass Media Complete, and Inspec. The Google Scholar search was based on title. We also conducted a backward search of references of all articles that met the inclusion criteria.

We retrieved articles for further analysis according to the following inclusion criteria: the targeted study population had to be women with GDM or women who were at risk of GDM; we considered overweight and obese women to be at risk for developing GDM. Studies had to describe a mobile phone app, and the mobile phone app had to focus on health promotion or disease prevention, or both. Exclusion criteria were studies focused solely on women with T1DM or T2DM. We excluded studies if mobile devices were used only to communicate between patients and health care providers (for data transmission, short messages, talk, or counseling). Other exclusion criteria were reviews or editorials, studies not in English, and studies for which the full text was not available.

According to these criteria, we selected the articles by title, abstracts, and then full text. Figure 1 presents the flow diagram of the search strategy. The initial searches were carried out in July 2016 and were not limited by date. The same searches were performed again in April 2017 to identify newly published studies.

After a full-text review, we classified studies into 2 categories based on the following content: (1) description of the development, feasibility, or usability of the apps or systems, and (2) trial protocols. To answer the research questions, we created a data charting form in Excel for Mac Version 15.25.1 (Microsoft Corporation) with the following elements: authors, year of publication, country of the study, category of the study, features of the app, behavioral theories, personalization features of the app, health literacy-related features, study design, sample characteristics, usability, feasibility, intervention components, and outcome measures.

We used information from the data charting form to summarize the overall number of studies, years of publication, characteristics of the study populations, countries where studies were conducted, and the focus and purpose of the studies. We report results of the review as categories and elements identified in stage 4 to answer the research questions, make comparisons among the studies, and identify research gaps.

The final 12 articles were published between 2014 and 2017 and were conducted in 7 countries: Switzerland, Spain, Norway, the United Kingdom, South Korea, Ireland, and Malaysia. Among the 12 articles, 7 described the development and feasibility or usability of the app or system [28-34]; 4 proposed RCT protocols using an app or system [35-38]; and 1 described the development of an app [16]. A total of 7 distinct systems or projects containing an app were described in the 12 articles.

Table 1 summarizes specific features of the apps and systems. Table 2 summarizes the results from development, usability, and feasibility studies.

Garcia-Saez et al described the development of a telemedicine system called MobiGuide [16]. This system used a decision support system (DSS) to generate personalized feedback for GDM management based on an expert-approved GDM guideline and patients’ data. MobiGuide includes a body area network, which provides real-time monitoring of biosignals such as BG by a Bluetooth-enabled glucometer, blood pressure by a blood pressure monitor, and activity level by an accelerometer within the smartphone. The MobiGuide system generates advice directly to both patients and health care providers. For the patients, advice regarding therapy, monitoring, and clinical assessment was generated based on their compliance with the therapy prescribed by the doctors to reinforce their behaviors. For the doctors, recommendations for changing diet and exercise, or insulin prescriptions, were generated based on patients’ compliance and BG control. In the feasibility test of the MobiGuide system, 20 women with GDM were initially instructed to measure their BG (4 times/day), ketonuria (once/day, manually entered), blood pressure (twice/week), and weight (once/week). One participant dropped out after 1 week. Recommendations were modified based on patients’ control and compliance. For example, good glycemic control could switch the BG measurement recommendation from 4 time per day to twice per week with 4 measures per day. Patients had high compliance (proportion of performed to recommended numbers of measurements) for BG (0.87±0.11), ketonuria (0.98±0.03), and blood pressure (0.82±0.24). When compared with the data of a historical cohort of 247 GDM patients, MobiGuide users had significantly better compliance to follow at least 4 BG measurements on indicated days (1.01±0.10 vs 0.87±0.28; P=.03) and better blood pressure control (98.6/64.7 vs 119.3/72.8 mmHg; P<.001). All 19 patients who completed the study received the message “High BG (2 abnormal per week). Did you eat more than you should?” The compliance rate (number who responded to the questions) was 0.31. Other recommendations related to ketonuria were generated to fewer patents (1-4 patients) with 0 to 1 compliance. Most of the users (12/17, 71%) thought the system improved their confidence in GDM management, 15/17 (88%) thought it did not complicate their lives, 12/17 (71%) liked the system’s ability to adapt to their daily life and context change, and 16/17 (94%) would recommend it to other patients. All 6 clinicians thought the system helped them to identify treatment priorities, 5/6 (83%) agreed the system increased patient safety, and 4/6 (67%) thought the system made it easier to manage patients [34].

As part of the expert personal health system (PHS) developed by Bromuri et al, an Android smartphone was given to GDM women to input their BG measures once they read it from the glucometer that they were provided [28]. Autonomous software entities, or agents, were programmed based on the American Academy of Family Physicians’ monitoring rules [39] to serve as the “experts.” The PHS was able to monitor BG readings and generate text-based alerts of hypoglycemic and hyperglycemic events to the caretaker (nurses, dietitians, doctors) on a Web interface. For example, if 2 BG values were less than 3 mmol/L within 1 hour, a hypoglycemia warning would be generated. With the automated alerts, caretakers were able to initiate an in-person or phone-based consultation with the patient based on the BG measures that triggered the alerts. An RCT was conducted with 24 GDM women to compare effects between a control group (standard care, n=12) and an intervention group (n=12) using the PHS in addition to standard care [28]. All women were asked to record their BG values 6 times a day for 2 to 4 months: fasting, postprandial breakfast, preprandial lunch, postprandial lunch, preprandial dinner, and postprandial dinner. The intervention group recorded significantly more BG measurements than the control group (235±86 vs 135±80; P<.001). Women in the intervention group had overall better BG control (5.4 vs 5.7 mmol/L or 98 vs 102.4 mg/dL; P<.001) than the control group. Among the 6 daily BG measurements, 4 were significantly better controlled in the intervention group (all 4 P<.001). In the intervention group, all patients rated the smartphone app easy to use and were satisfied with the care provided by the system. Caregivers in this study considered the system appropriate for GDM management. Although the time needed for patients’ consultation remained the same, the caregivers thought the PHS was more time efficient because they were able to focus on the hyperglycemic or hypoglycemic events based on the alerts instead of going through patients’ BG records manually. In addition, consultations could be initiated within 1 to 3 days with the PHS instead of 1 to 3 weeks with the standard care after the hyperglycemic or hypoglycemic events. The PHS system also allowed for the possibility of daily consultation on patients’ BG readings.

Pregnant+ is an app developed to monitor GDM women’s BG level by Bluetooth transmission or manual input [29]. It generates immediate feedback, provides information on healthy diets based on the cultural background of the user (eg, using food items preferred in users’ cultures), provides physical activity information based on level of activity, and provides general information about GDM. The content of the app was designed to emphasize patients’ perceived severity of their disease, emphasize perceived benefits to treatment and management, and provide cues to action based on the health belief model. Although BG records cannot be transferred automatically to the health care providers due to medical data security, users can print their BG records at the clinics [29]. In a user involvement study for the Pregnant+ app, most participants were able to perform tasks related to the 4 major functions of the app; namely, finding (1) where to register BG levels (11/11, 100%), (2) information about healthy eating (10/11, 91%), (3) information about physical activities (10/11, 91%), and (4) general information about GDM (10/11, 91%) [29]. Fewer participants were able to find other functions, such as entering appointments for medical consultations (9/11, 82%) and finding how to register body weight (5/11, 45%). Users of this app believed it would make it easier for them to register and control their BG level than with standard care. They also reported favorable reviews for the features that provided real-time feedback and information about GDM, diet, and physical activity.

Jo and Park developed an app for Korean women with GDM [32]. This app generates recommendations about the risk factors of GDM, importance of GDM management, and management of BG, diet, physical activity, and body weight to patients based on their initial assessment and lifestyle data, including caloric intake and physical activity level. Algorithms using patients’ data and clinical guidelines [40-42] were developed to generate individually tailored recommendations. A total of 5 GDM patients participated in the usability test of this app. The average usability score was 69.5 out of 100 as measured by a Korean version of the System Usability Scale [32]. User acceptance was measured using Wilson and Lankton’s model of patients’ acceptance of provider-delivered eHealth [43]. The user acceptance score with behavioral intention to use was 5.5, intrinsic motivation score was 4.3, perceived ease of use score was 4.6, and perceived usefulness score was 5.0, out of 7 for all measures.

The GDm-Health system is a real-time BG monitoring management system for women with GDM that consists of a smartphone app and a website [33]. This smartphone app allows women to automatically synchronize their BG levels from their glucose meter through Bluetooth and provides immediate feedback based on the BG readings. BG levels are sent to a central server where health care professionals can access the data on a website. Another function of this system is to allow 2-way communication where health care professionals give advice or change medication and users can request a callback from the team to address their concerns. A total of 7 women were involved in the beta test phase of the GDm-Health system, and 50 women with GDM tested the system until delivery. On average, the women used the system for 13.1 weeks, 46 of 54 (85%) submitted the minimal requirement of 18 BG readings per week, and 19,410 of 19,686 (98.6%) readings were manually tagged with additional information indicating when it was measured (pre- or postprandial) [33]. The Oxford Maternity Diabetes Treatment Satisfaction Questionnaire was developed and used to assess the acceptability of the system [30]. Overall, 45/49 (92%) women were satisfied with the care delivered by the system, and 46/49 (94%) agreed they had a good relationship with their care team. Most agreed that the equipment was convenient (47/49, 96%), reliable (43/49, 88%), and fit into their lifestyle (42/49, 86%). Birth outcome data were available for 41 women, of whom 12 (29%) delivered large for gestational age (LGA) babies. Mothers of LGA versus non-LGA babies had significantly higher mean (6.3 vs 5.6 mmol/L; P=.004), fasting (5.8 vs 5.1 mmol/L; P=.004), and 2-hour postprandial BG readings (6.9 vs 6.0 mmol/L; P=.001). A 1-SD increase (0.7 mmol/L) in mean BG increased the odds of delivering an LGA baby by fivefold (odds ratio 5.5, 95% CI 1.4-21.2) [31].

A total of 4 ongoing RCTs are using a mobile phone app or using an app as part of the intervention component to prevent or manage GDM [35-38]. Table 1 summarizes characteristics of the apps and Table 2 summarizes characteristics of the RCT protocols.

Mackillop et al are testing the efficacy of using the GDm-Health system compared with standard clinic care [37]. A total of 200 women with abnormal glucose tolerance between 14 and 34 weeks of gestation have been randomly assigned to 1 of 2 groups: GDm-Health system and clinic visit every 4 to 8 weeks; or normal clinic care (visit the clinic every 2 to 4 weeks). The primary outcome is BG control, as determined by mean BG readings from recruitment until delivery compared between the intervention and the control group. The secondary outcomes are compliance with the allocated BG monitoring regimen, maternal and neonatal outcomes, glycemic control using hemoglobin A_1c and other BG metrics, and patient attitudes toward care.

Borgen and colleagues are evaluating the efficacy of the Pregnant + app [35]. A total of 230 pregnant women with GDM who own a smartphone, understand Norwegian, Urdu, or Somali, and are before 33 weeks of gestation were recruited. Women will randomly receive either the Pregnancy+ app and standard care or standard care until 3 months postpartum. The primary outcome is glucose tolerance after the intervention, measured by 2-hour oral glucose tolerance test. Secondary outcomes are birth weight, mode of delivery and complications for mother and child, change in diet and physical activity from baseline to 36 weeks of gestation (measured by a modification of the Fit for Delivery questionnaire and the Pregnancy Physical Activity Questionnaire), and quality of life (measured by a short version of the Edinburgh Postnatal Depression Scale and by health-related quality of life during pregnancy and postpartum) [35].

Kennelly et al are conducting a pregnancy, exercise, and nutrition research study with smartphone app support (Pears) targeting low glycemic index dietary and physical activity promotion among overweight and obese pregnant women [36]. The Pears healthy lifestyle package includes a 75-minute in-person education session, biweekly emails, 2 follow-up appointments, and an app. The app provides behavior, dietary, and physical activity tips, physical activity benefits, and a database of low glycemic index recipes. Control theory and social cognitive theory were applied to set up patients’ personal SMART (specific, measurable, achievable, relevant, and time-specific) goals, and to overcome personal and environmental barriers to change. The authors are randomly assigning 506 overweight or obese women between 10 and 15 weeks’ gestation to the intervention or the control arm to assess the impact of the Pears healthy lifestyle package [36]. The primary outcome is the incidence of GDM at the 29th week. Secondary outcomes will be gestational weight gain, maternal physical activity levels, and glycemic index and glycemic load of the mothers’ diet in the third trimester.

Skau et al developed a behavior change intervention, Jom Mama, targeting young Malaysian couples to promote women’s health prior to pregnancy [38]. This project includes 3 in-person and 3 phone communications with community health promoters, and an eHealth platform. The eHealth platform includes an app for the couples and a Web-based interface for the community health promoters. The couples can set personal goals, track their progress, and access general information on healthy lifestyles from the app. Women and spouses can select different challenges for obtaining a healthy diet (eg, avoid soft drinks), increasing their physical activity, or decreasing their sedentary behavior during the intervention period (eg, cycle for 30 minutes). Community health promoters follow the progress of the couples and interact with them during the in-person and phone communications using the information from the Web-based interface. Skau and colleagues are testing the efficacy of the Jom Mama project on preconception health promotion [38]. They are recruiting a total of 660 nulliparous women between 20 and 39 years of age who own a smartphone and are free of diabetes to randomly assign to an intervention or control group. The planned follow-up duration will be 8 months. The primary outcome is change in waist circumference. Secondary outcomes will be changes in other anthropometric (eg, body mass index, waist-to-hip ratio), biochemical measures (eg, hemoglobin A_1c, lipid profile), health literacy, dietary intake, physical activity, and stress level. They will also measure the incidence of GDM proposed as an outcome in women who completed the intervention and become pregnant after the trial.

Health literacy was taken into account in 2 of the 7 final apps and systems. The Pregnant+ app was the only system that incorporated user literacy level in the development phase [29]. The researchers checked the app against Kreuter’s message checklist, which includes checking the content, writing, literacy, and elements of visual communication [44] and administering the Suitability Assessment of Materials instrument [45] to make sure the app was appropriate for their targeted audience. In addition, after the second stage of the user involvement study with 11 GDM patients of varying literacy levels, a diabetes wordlist was added to the app to explain medical jargon [29]. The Jom Mama intervention was designed to measure change in the level of health literacy using the European Health Literacy Survey Questionnaire (a 47-item scale covering 3 domains: health care, disease prevention, and health promotion) as one of the outcomes of the intervention [38]. PHS, Pregnant+, and GDm-Health all used a data visualization strategy to present normal and abnormal BG data in figures [28,29,33]. Although Jo and Park involved users in the development of the app, they did not mention whether the app met the health literacy level of their targeted users [32]. Mackillop et al assumed that their target users for the GDm-Health system being recruited from a large single-center tertiary referral unit in southern England would have high rates of literacy and low levels of social deprivation [37].

There is a limited body of published data on the use of mobile phone apps for GDM. In this review, 12 articles focused on the development, usability, feasibility, and trial protocols of mobile phone apps or interventions including an app. DSSs were used to connect patients’ data to tailored feedback for both patients and health care providers using clinical guidelines. Health literacy was considered a feature of 1 app during the development phase [29] and was measured as an outcome by another app [38].

Figure 2 presents the common framework combining all the characteristics of the app or system with automated features. The findings suggest that DSSs have the capacity to generate real-time personalized feedback based on users’ input and existing clinical guidelines. From the studies identified, a variety of data from the users have been collected, recorded, and saved in the app or system, or used to develop the app or system. These data include health records, biosignals collected by body area network, user preferences, culture, lifestyle data, clinical data, and personal health goals. However, not all of the information was used in the DSSs to generate output to the patients or health care providers. The information that was used most frequently was BG levels collected by body area network or entered by the users. The MobiGuide system had the highest level of personalization among all the apps and systems we identified [16]. This system feeds the DSS with historical clinical data, the personal health record, body area network sensor data, and manually entered data to generate personalized feedback. However, deciding on the right amount of advice to reinforce users’ behaviors without overwhelming them can be challenging. The primary functions of the apps and systems included providing information, promoting lifestyle change, assessing and monitoring user status, and managing medication and complication with approval from providers. Also, protecting the security and privacy of patients’ data is a common feature of the apps and systems. Similar to the findings of this review, in a review of commercial apps for diabetes self-management, El Gayar found that 12 of 71 (17%) had decision support capabilities and all of them were related to insulin dosage suggestions as opposed to lifestyle changes [14]. Internet-based interventions that promoted lifestyle modifications for diabetes management, were based on theory, included interactive components and personalized feedback, and provided peer support were most successful [46]. In our review, only 2 apps incorporated behavior change theories [16,29], and 2 RCTs used theories in their proposed interventions, but not specifically in their app or system [36].

Even if the apps provide high-quality, evidence-based content, the value is limited if the information does not adequately match and address the usability, accessibility, readability, and health literacy needs of target audiences [47]. Although users were frequently involved in usability and feasibility studies to inform the development and finalization of the app or system, their literacy level was discussed in only 2 studies. Caburnay et al analyzed the health literacy-related features of more than 100 diabetes apps with a specific focus on using plain language, displaying content clearly, organizing and simplifying the user interface, and engaging users [48]. A total of 84% of the apps employed at least one of the plain language strategies, such as using common everyday words; avoiding undefined technical or medical terms; and using active voice, action words, and present tense [48]. Involving users in the development phase and evaluating users’ attitudes, knowledge, beliefs, and behavior related to health information are helpful strategies to improve usability of an app [49].

Overall, the users involved in the usability and feasibility studies found it easy to navigate the apps and systems, and were satisfied with the technology. Apps and systems have the potential to improve compliance with BG monitoring and treatment prescriptions, and improve communication between users and health care providers. El-Gayar et al categorized the technology for diabetes self-management into the Internet, cellular phones, telemedicine, and decision support techniques [50]. However, our review found that DSSs can be embedded in mobile phone apps to generate real-time feedback.

Only 1 feasibility study [28] with a sample of 24 women showed that patients who received PHS care had better BG control than did patients who received standard care. Better blood pressure control was reported in another feasibility study; however, no difference in BG control was observed [34]. Large RCTs are needed to confirm the system’s impact on BG control and other clinical outcomes. In this review, we identified 3 RCT protocols: only 1 protocol, by Mackillop et al [37], evaluated the efficacy of the GDm-Health system, and the other 2 protocols evaluated complex lifestyle programs with an app as part of the intervention [36,38].

We applied no evaluation criteria to the articles due to the nature of this scoping review. This review only searched abstract, title, and topics in most databases, which may not yield a complete pool of relevant articles. Articles published in languages other than English were not included. However, this is, to our knowledge, the first known review of mobile phone apps on GDM to provide an overview of the literature.

This scoping review describes the literature on mobile phone apps for GDM prevention and management. We identified and described 12 articles that discussed the design and development, usability, feasibility, and RCT protocols of GDM-related apps. Findings from this scoping review suggest that mobile phone apps have the potential to prevent GDM and improve GDM management. Future research should focus on large RCTs of the impact of these apps. In addition, health literacy levels of the potential audience should be taken into consideration when developing and evaluating the usability of apps for this audience.