Behavior Change Apps for Gestational Diabetes Management: Exploring Desirable Features

ABSTRACT Gestational diabetes mellitus (GDM) has considerable and increasing health effects as it raises both the mother’s and offspring’s risk for short- and long-term health problems. GDM can usually be treated with a healthier lifestyle, such as appropriate dietary modifications and engaging insufficient physical activity. While telemedicine interventions requiring weekly or more frequent feedback from health care professionals have shown the potential to improve glycemic control amongst women with GDM, apps without extensive input from health care professionals are limited and have not shown to be effective. We aimed to improve the efficacy of GDM self-management apps by exploring desirable features in a review. We derived six desirable features from the multidisciplinary literature and we evaluated the state of implementation of these features in existing GDM apps. The results showed that features for increasing competence to manage GDM and for providing social support were largely lacking.


Introduction
Approximately 16% of pregnant women globally are diagnosed with gestational diabetes (GDM) (Guariguata et al., 2014). The prevalence of GDM has been increasing primarily due to maternal obesity and a sedentary lifestyle. GDM and maternal obesity are associated with a range of adverse shortand long-term consequences for both mother and child (Owens et al., 2010;Persson et al., 2014). Although GDM is a temporary condition that lasts until the birth of the child, GDM increases the later risk of type 2 diabetes (T2D) for the mother (Bellamy et al., 2009). If GDM is left untreated, the costs of intergenerational obesity and diabetes are substantial.
For approximately 80% of the women diagnosed with GDM, the GDM is managed through lifestyle adjustments (Carolan et al., 2012). These adjustments require changes in behavior toward eating healthier and increasing exercising (Brown et al., 2017). Nutrition is the primary factor affecting glucose levels (Gilbert et al., 2019), but also physical activity (Halse et al., 2014;Harrison et al., 2016;Kokic et al., 2018), stress (Horsch et al., 2016), and sleep (Twedt et al., 2015) have an impact on glucose homeostasis. It is critical that women with GDM are supported in self-management. This paper was intended as a constructive step in the design and development of a mobile app to support behavior change required in the self-management of GDM. While over 1000 apps have been designed to support self-management of diabetes (Huang et al., 2018), they do not consider the design implications for GDM apps arising from temporality and pregnancy. In terms of duration, GDM is very different from the most common diabetes types (type 1 and type 2) that typically last a lifetime. Given that pregnancy usually lasts 40 weeks and GDM is diagnosed after 12 to 28 weeks of pregnancy, any health app designed for managing GDM is used for a limited time (for approximately 12-28 weeks). When considering pregnant women as a user group, the pregnancy itself has implications on design, as it causes several mental and physical changes (e.g., diet limitations and sleep disorders). On the other hand, pregnancy represents an exceptional opportunity for lifestyle changes (Arabin & Baschat, 2017).
According to a recent meta-analysis (Xie et al., 2020), telemedicine interventions requiring weekly or more frequent feedback from health care professionals have been shown to be effective in improving glycemic control amongst women with GDM. However, providing this feedback to each woman with GDM by healthcare professionals every day requires a lot of human work and, thus, is costly. GDM management apps with less input from health care professionals are limited and have not shown to be effective (Borgen et al., 2019;Mackillop et al., 2018;Rigla et al., 2018). In GDM, successful self-management requires knowledge of how one's activities and lifestyle (e.g., nutrition, physical activity, sleep, and stress) affect glucose levels. This is a significant challenge for women with GDM, who report the feeling of loss of control over their body and blood glucose (Evans & O'Brien, 2005).
Given the high potential of mobile apps for GDM management (Hewage et al., 2020) and the lack of evidence on the efficacy of different app features, approaches for improving apps should be investigated. The objective of this paper is to establish desirable features for GDM apps from the relevant multidisciplinary literature and to evaluate how desirable features are implemented in recent GDM apps. We believe that by incorporating the underutilized theories and results from the literature and interviews, we can increase the efficacy of apps designed for GDM self-management. We argue that the results are useful for various stakeholders that target effective mHealth interventions for GDM.

Background
In this paper, we sought to increase the impact of future GDM apps by providing desirable features. While we acknowledge that technical interoperability, standardization, and data security are still significant challenges for diabetes apps (Fleming et al., 2020), we aim to investigate desirable features that are important for supporting behavior change. First, we provide an overview of relevant themes related to behavior change amongst women with GDM and discuss why they are important when establishing desirable features for GDM apps.

Clinical outcomes of mobile apps for GDM
It is justified to begin with the clinical outcomes of existing GDM apps, as our ultimate aim is to improve the efficacy of GDM mobile apps. According to a recent meta-analysis, telemedicine interventions requiring weekly or more frequent feedback from health care professionals have shown the potential to improve glycemic control amongst women with GDM (Xie et al., 2020). Typically, in these interventions, women with GDM are able to communicate with each other and the study's interventionist on an online chat, such as WeChat. A study by Miremberg et al. (2018) revealed a statistically significant improvement in improving glycemic control among women with GDM. In addition to having a larger sample size (124 participants divided almost equally into the intervention and control groups), Miremberg et al. (2018) had a substantial feedback loop provided by study personnel; every evening the participants received individualized feedback via e-mail from the clinical team regarding their daily glycemic control. The feedback included "reassurance and positive messaging, dietary tips in attempts to optimize specific off-target measurements, modifications in insulin treatment, or alerts to reschedule an earlier appointment to the clinic." Moreover, participants were "encouraged to use the platform to ask questions and receive immediate answers regarding any aspect of GDM management" (Miremberg et al., 2018). Similarly, Yang et al. (2018) found statistically significant improvement in glycemic control when a chat platform between women with GDM and health care professionals was provided. However, GDM apps with less such input from health care professionals are limited and have not shown to be effective (Borgen et al., 2019;Mackillop et al., 2018;Rigla et al., 2018). Thus, the problem to be addressed lies in how these apps could be made more effective.

User experience of mobile apps for GDM self-management
While there are similarities between type 2 diabetes (T2D) and GDM, previous work indicates that there are differences that should be considered (Angueira et al., 2015). Fantinelli et al. (2019) conducted a review on experiences with telemedicine for managing GDM. They found that most of the telemedicine approaches were found to be acceptable. Nicholson et al. (2016) developed a web portal (GoodMomS) to support women with GDM that allowed users to record their daily weight, exercise, and glucose levels during pregnancy and to find information about GDM and a healthy lifestyle. While users found the system feasible, the largest obstacle was that the system was not usable as an app. Mackillop et al. (2014) designed an app that allowed women with GDM to track their diet and blood glucose and to engage in two-way communication with women with GDM and healthcare professionals. The healthcare professionals had access to the provided data through a website. The evaluation conducted by Hirst et al. (2015) revealed that the system was perceived as acceptable and convenient among women with GDM. Similarly, Rigla et al. (2018) found that an app coupled with a Bluetooth glucose measurement device was perceived as highly acceptable. These results indicate that a mHealth approach is applicable for women with GDM. However, the review by Fantinelli et al. (2019) concludes that more studies are required for investigating engagement and empowerment with telemedicine for GDM.
Knowledge on user experience with GDM apps is limited, although recent studies have provided relevant findings (Fantinelli et al., 2019;Garnweidner-Holme et al., 2015;Ming et al., 2016;Skar et al., 2018). Garnweidner-Holme et al. (2015) developed the Pregnant++ app, which was evaluated by Skar et al. (2018). Skar et al. (2018) found that the most important features of the GDM app were the overview of blood glucose values and real-time feedback. The information on nutrition concerning GDM was also found to be important. However, the app had some limitations. For example, physical activities were entered manually, which prevented the collection of physical activity data and thus did not support comprehensive self-tracking. Moreover, most users had to manually enter glucose values into the application, which also decreased data collection and usage over time. This result is consistent with the review by Fantinelli et al. (2019), that ways to improve engagement with the GDM apps should be explored.

Mobile apps for T2D self-management
As experiences with mobile GDM apps are limited (see the previous section), we can also build on previous work on T2D apps when applicable. Health literature provides examples of effective mHealth interventions for T2D self-management based on apps (Adu et al., 2018;Hou et al., 2016). Due to the similarities in treatment between GDM and T2D (such as appropriate dietary modifications and engaging insufficient physical activity (American Diabetes Association, 2016)), some of the design guidelines for self-management apps (such as those by Årsand et al. (2012)) can be derived from studies that have revealed statistically significant improvements in glucose levels using apps for T2D (Quinn et al., 2011). In addition, due to the similarities in treatment, multiple studies have had a mixture of participants with T2D or GDM (Desai et al., 2018;El-Gayar et al., 2013;Garabedian et al., 2015;Hoppe et al., 2017), which suggests that opinions from women with GDM have been considered at least to some extent in studies that investigate diabetes management using mHealth.
Although there exist many studies investigating the effectiveness of mobile apps for T2D (see a review by Adu et al. (2018)), the evaluations of the effectiveness of separate features are largely lacking. This makes constructing desirable features difficult, as knowledge on the efficacy of different features in mobile apps remains largely unknown (Hood et al., 2016). Deeper user studies on a particular feature in apps designed for diabetes management have focused on for example, visualization of glucose levels (Desai et al., 2018;Katz et al., 2018a). However, the efficacy of different features has not been investigated, as it is a typical approach in human-computer interaction (HCI) studies to address why the features in eHealth are perceived as useful but not their efficacy (Blandford et al., 2018;Klasnja et al., 2011).
Although the efficacy of different features in T2D applications are largely unknown, Årsand et al. (2012) provided the following four main design implications for T2D applications based on evaluating 10 different feature sets: 1) Automatic data transfer when possible; 2) Motivational and visual user interfaces; 3) Application should provide considerable health benefits in relation to the effort required; and 4) Dynamic usage, e.g., both personal and together with health care personnel. We reflect on these guidelines with respect to relevant findings with apps designed for pregnant women.

Mobile apps for pregnancy
A review of qualitative studies on the mental effects of GDM revealed "loss of normal pregnancy" as a common theme (Parsons et al., 2014). To avoid this, the feeling of a normal pregnancy should be supported, and thus features of normal pregnancy apps (such as information about pregnancy (Skar et al., 2018)) should be incorporated into GDM apps. A review by Overdijkink et al. (2018) revealed that a large body of research has been and is being performed in the field of pregnancy mHealth and eHealth using apps and web-based platforms. These solutions have different aims, such as making the pregnancy experience better and more pleasant for mothers, promoting the social dimension of pregnancy (e.g., getting support and sharing a baby activity with close people), or supporting a healthy pregnancy and countering the health risks both for the mother and the child (Overdijkink et al., 2018;Peyton et al., 2014). According to a study by Runkle et al. (2019), roughly half of pregnant women in the United States have downloaded a pregnancy-related mobile phone app. Despite supportive features, only a small number of studies have shown significant effective mHealth interventions for pregnant women (Daly et al., 2018).

Experiences with GDM
Qualitative studies report feelings of failure, anxiety, loss of control, and powerlessness after receiving a GDM diagnosis (Craig et al., 2020;Parsons et al., 2014;Persson et al., 2010). Also, quantitative studies have shown increased distress and reduced social life and quality of life (QoL) for women with GDM compared to healthy pregnant women (Kopec et al., 2015;Marchetti et al., 2017). However, women with GDM diagnosis take the diagnosis very seriously since the women want to protect the child from all harm (Evans & O'Brien, 2005). Women with GDM experience "a steep learning curve"; they go from the initial shock of the diagnosis to acceptance and active management of their condition (Draffin et al., 2016). Persson et al. (2010) describe the same kind of "curve"; they observed that women with GDM gradually adapted to the situation and found their own ways of managing their condition, achieving "a sense of balance, control and wellbeing." Evans and O'Brien (2005) also described a process of establishing a balance and going from a passive "victim of diabetes" to being an active agent in managing the condition. As such, there is a clear adaption to the situation and a transformation process from a passive victim into an active agent. Despite the learning process, QoL is also reduced in the later part of pregnancy, as Pantzartzis et al. (2019) reported reduced QoL compared to uncomplicated pregnancy in the third trimester of pregnancy. It is unclear how mobile apps should be designed to support this learning process and improve QoL.

Behavioral change theories
As we aim to provide constructive guidelines for women with GDM, the theories behind behavior change are important. Behavioral change interventions based on a theory are more effective at achieving behavior change than those that are not based on a theory (Glanz & Bishop, 2010). However, theoretical frameworks of behavior change have not been explicitly applied in previous research apps for diabetes management (Adu et al., 2018;Payne et al., 2015) or used widely in publicly and commercially available apps for diabetes management (Hoppe et al., 2017;Priesterroth et al., 2019). Thus, failure to incorporate theories of behavioral change is expected to be one obstacle for having a significant impact on the daily lives of the users.
From the large number of theoretical frameworks that explain constructs of behavior change (see a review by Kwasnicka et al. (2016)), in this paper, we focus on theories that have been empirically shown to be effective in describing the behavioral change. The first is a very popular theory on motivation, namely the self-determination theory (SDT) (Ryan & Deci, 2000). SDT differentiates between several types of motivation from purely extrinsic to purely intrinsic motivation and describes the implications of these different types of motivation to human behavior. In SDT, the most important division between different types is the one between controlled and autonomous motivation. According to SDT, the more autonomous types of motivation can be supported by fulfilling the three basic psychological needs of competence, autonomy, and relatedness (Ryan & Deci, 2000). With competence, Ryan and Deci refer to the need to achieve personally valued outcomes. Autonomy is the desire to be in charge of one's actions while acting in accordance with one's sense of self, while relatedness refers to the need to feel connected to others (Ryan & Deci, 2000).
In previous studies, autonomous motivation has been connected with improvements in glucose control in diabetes patients (Williams et al., 1998), higher levels of physical activity (Teixeira et al., 2012), and long-term behavioral outcomes (Silva et al., 2011). Moreover, SDT is often used as a theoretical framework for gamification (Sailer et al., 2017;Seaborn & Fels, 2015), which has been shown to increase the frequency of glucose measurements (Cafazzo et al., 2012). Over half of the studies in a review by Johnson et al. (2016) support the efficacy of gamification for improving health and wellbeing in general. Nevertheless, SDT has still not been explicitly widely used in apps designed for diabetes management (Adu et al., 2018;Payne et al., 2015).
To induce long-term behavioral change, the automatic or habitual level must also be considered in addition to supporting the reflective process by increasing motivation (Marteau et al., 2012;Wood & Rünger, 2016). According to the dualprocess theory, our actions result from the following two different processes: reflective but also automatic and habitual processes (Kahneman, 2011). Habits can be considered particularly important with women diagnosed with GDM, as their motivation for maintaining a healthier lifestyle typically decreases after the birth (Skar et al., 2018). Habits could help maintain a healthier lifestyle also after the pregnancy and thus decrease the probability of T2D. Paradoxically, although checking our mobile phones has become a habit (Oulasvirta et al., 2012), the creation of healthier habits using mobile phones has not been successful (Hermsen et al., 2016). This has been explained by the lack of implementing theories on habit formation in apps in general (Adams et al., 2015;Caraban et al., 2019;Stawarz et al., 2015) and in apps designed for diabetes management (Adu et al., 2018;Klonoff, 2019;Payne et al., 2015).

Material and method
We searched for articles related to each theme (see Section 2) using both health and HCI resources. The corpus was selected by searching online (the search terms are provided in the last column of Table 1) databases (Google Scholar, ACM Digital Library, PubMed, Scopus) for conference and journal articles on the themes described in Section 2. We chose to include very recent literature reviews around each theme (see Table 1) for achieving state-of-the-art knowledge. However, since literature reviews had a more restricted focus on selecting articles (either because of a more focused topic or because of using certain a database such as PubMed), we extended the search outside the reviews to search for relevant articles on themes that were not covered by the literature reviews (see Table 2). This also included articles that had appeared later than literature reviews. We used the same search terms as in Table 1 for each theme and citation tracking to complement the corpus. The inclusion criteria were relevance (based on the abstract), peer-reviewed research articles or wellestablished books, and written in English. Altogether our included corpus consisted of literature reviews (N = 6) and complementary research articles (N = 95).
During data analysis, one coder read all the articles and performed open coding. The coding was limited to corpus sentences that were related to the themes provided in Section 2. Codes were then clustered into the desirable features presented below. The desirable features presented below are indications of what should be considered when designing apps for gestational diabetes. However, it is worth noting that all of these characteristics may not always be needed in GDM apps. As such, these desirable features simply indicate possibilities in which GDM apps may be improved.

Desirable feature 1: Increase competence to manage GDM with automatic feedback and interactive exploration
Qualitative studies (Carolan et al., 2012;Craig et al., 2020;Draffin et al., 2016) indicate that self-discovery, i.e. learning causalities between lifestyle and blood glucose, is challenging and demanding, which takes a considerable amount of time. Nevertheless, the learning process could be expedited and facilitated. Based on the literature we found two approaches to support self-discovery that have been largely neglected and would be beneficial automated data collection and interactive learning environments.
Objectively and automatically measured, and constantly available data through wearable sensors data can be expected to support self-discovery MacLeod et al., 2013). In fact, multiple studies emphasize the importance of automatic data collection in diabetes apps (see (Hood et al., 2016) for a review), although this is rarely found in apps used in diabetes research (Hood et al., 2016). In particular, the measurement of blood glucose levels is the most important feature of the GDM app. Although the automatic transfer of blood glucose measurements is important (Skar et al., 2018), this feature has rarely been successfully implemented in diabetes apps found online (Chomutare et al., 2011). Moreover, the requirement of entering the information manually has significantly decreased the collection of physical activity data in a mobile GDM app (Skar et al., 2018). This is understandable, as pregnant women often have limited energy for monitoring their own behavior since they already have a lot to do and to deal with (Gao et al., 2014;Hewage et al., 2020;Peyton et al., 2014).
For an app designed for GDM management, one of the most important features is tracking and managing the diet, as this is the primary factor that affects glucose levels (American Diabetes Association, 2004). Collecting nutrition information cannot yet be fully automated reliably (Hassannejad et al., 2017). This leads to a situation where some effort is required from the user for collecting data. Some studies emphasize the Table 1. Themes and related reviews. The reviews are structured according to V. Smith et al. (2011). The most recent (before 1/1/2020) and comprehensive reviews were included. The used search terms are provided in the last column.
Theme and the review importance of taking a photo of the meal and combining it with glucose measurements (B. K. Smith et al., 2007). Furthermore, the time that a photo is taken also indicates the eating time (Aizawa et al., 2014; B. K. Smith et al., 2007). Peyton et al. (2014) suggest that self-monitoring of pregnant women, in addition to photographic journals, can be supported and encouraged by using simple designs with tips and tricks, reminders, and by keeping the techniques for user data input simple (e.g., checkboxes instead of long texts). The researchers also note that "immediate concerns," such as nausea prevention, could be used as "hooks" for app use and to tackle long-term issues. While physical activity is one of the cornerstones in diabetes management (American Diabetes Association, 2016;Harrison et al., 2016), automatic collection of physical activity data has gained minimal attention in GDM apps (Skar et al., 2018). Chan and Chen (2019) reported that interventions for increasing physical activity amongst pregnant women were more effective with wearable devices than without, and automatic self-tracking of lifestyle (e.g., nutrition, physical activity, and symptoms) has been argued to help in countering pregnancy-related health risks (Penders et al., 2015;Runkle et al., 2019). In T2D, data collection on physical activity has typically been based on steps recorded by a mobile phone, smartwatch, or a pedometer attached to the belt (Årsand et al., 2015, 2010). Although self-monitoring steps is motivating, Årsand et al. (2010) reported that these measurement approaches do not support the measurement of other activities (such as cycling, swimming, and skiing), which was the largest problem for users when tracking their physical activity.
As such, using a sensor that records physical activity (i.e., through heart rate) other than steps and can be worn ubiquitously would increase data collection on physical activity. In a recent study, users with T2D reported a high correlation between self-reported physical activity and physical activity measured with an activity bracelet (Weatherall et al., 2018), which have been found to be feasible amongst pregnant women (Grym et al., 2019). We provide example implementations for increasing competence to manage GDM with apps in Table 3.
These results indicate that automatic measurement and collection of physical activity with an activity bracelet has great potential also for women with GDM. Although automatic data collection can be expected to have multiple benefits as described above, it may increase the feeling of medicalization and lead to the feeling of losing a normal pregnancy (Parsons et al., 2014). To avoid this, it should be clearly stated that the purpose of the automatic data collection is to empower the mother to manage GDM and that the child is not monitored. Furthermore, as women with GDM perceive their diagnosis as "a stun," (Persson et al., 2010), starting selftracking on the mother's own terms would be a good strategy. This approach would also support autonomy, which has been shown to be an important motivational factor in diabetes management (Williams et al., 2004). The other approach for supporting self-discovery would be interactive learning environments. A survey on the features of diabetes apps by Adu et al. (2018) revealed that the teaching of diabetes self-management skills is underrepresented in diabetes applications, although recent guidelines emphasize  (Xie et al., 2020) Less involvement by health care professionals: (Borgen et al., 2019;Mackillop et al., 2018;Rigla et al., 2018), GDM App without self-tracking: (Kennelly et al., 2018) Mobile apps for GDM, user experience perience (Fantinelli et al., 2019) User experience with Mobile apps for GDM: Garnweidner-Holme et al., 2015;Mackillop et al., 2018;Miremberg et al., 2018;Nicholson et al., 2016;Nikolopoulos et al., 2019;Peleg et al., 2017;Skar et al., 2018) Mobile apps for T2D (Adu et al., 2018) Other reviews: (Chomutare et al., 2011;Hood et al., 2016;Hoppe et al., 2017;Payne et al., 2015), Applications for T2D: (Årsand et al., 2010El-Gayar et al., 2013;Hoppe et al., 2017), Data visualization specific: (Desai et al., 2018;Katz et al., 2018a;Katz et al., 2018b) Mobile Apps for Pregnancy (Overdijkink et al., 2018) (Abid & Shahid, 2017;Chan & Chen, 2018;Daly et al., 2018;Huberty et al., 2016;Hui et al., 2012;Lee et al., 2016;Muuraiskangas et al., 2016;Peyton et al., 2014;Sajjad & Shahid, 2016;Thomas et al., 2018;Wenger et al., 2014;Wierckx et al., 2014) Experiences with GDM (Craig et al., 2020) Other review: (Parsons et al., 2014), Qualitative studies on experiences with GDM: ( Review of using behavioral change theories in health: (Glanz & Bishop, 2010), Self-determination theorybased: (Deci & Ryan, 2012;Silva et al., 2011;Teixeira et al., 2012;Williams et al., 1998), Diabetes-specific behavioral change theories: (Williams et al., 2004), Gamification: Seaborn & Fels, 2015),  2015) noted that increased knowledge on diabetes during pregnancy increased perceived general health. Further Carolan-Olah et al. (2015) suggested that knowledge on GDM could be improved, particularly for women with multiethnic and low socioeconomic backgrounds.

Desirable feature 2: Increase autonomy by enabling personalization
In addition to competence, another important component of intrinsic motivation is autonomy. In general, mothers are often motivated to change their lifestyle habits because they are concerned about the condition of the child (Carolan, 2013;Draffin et al., 2016;Evans & O'Brien, 2005;Persson et al., 2010). This motivational factor is mainly extrinsic and is expected to last until birth. However, with intrinsic motivation behavior change will be better maintained (Williams et al., 1998). Williams et al. (2004) found that autonomy and competence measured with questionnaires have a positive impact on self-management of diabetes. The autonomy and competence were varied by how much the nurse gave freedom to manage diabetes. Regarding autonomy, Evans and O'Brien (2005) noted in their qualitative evaluation about the effects of GDM that part of their interviewed participants "cheated to retain their autonomy." The strict rules undermined these women's desired level of autonomy. Furthermore, in a synthesis of qualitative studies on GDM, Parsons et al. (2014) summarized that medicalization of pregnancy leads to feelings of loss of normal pregnancy and further to feelings of reduced control. We expect that autonomy can be increased by personalization, as exemplified in Table 4, 5, 6, 7, 8.
Personalization, also referred to as tailoring, is a wellestablished technique in behavior change apps. In the case of pregnancy and further GDM, personalizing the app could also be useful in accounting for the perceived uniqueness of the pregnancy, which Peyton et al. (2014) presented as a prominent belief among pregnant women in the USA. Personalization can be implemented in multiple ways, ranging from personalized greeting messages to personalized feedback algorithms (Goetz et al., 2017;Lee et al., 2016;Peyton et al., 2014). Thus far, personalized recommendations in GDM mobile apps have been created manually by health Table 3. Example implementations with justifications for desirable feature 1.

Implementation Justification
Continuous glucose monitoring for providing automatic feedback Manually entering the blood glucose values has decreased significantly collecting the glucose data amongst women with GDM (Skar et al., 2018). Glucose measurements can be done automatically and more frequently with continuous glucose monitors, which are found to be acceptable among women with GDM (Jovanovic, 2000;Lane et al., 2019;Scott et al., 2018;Voormolen et al., 2018). Activity bracelet for self-tracking physical activity Interventions for increasing physical activity amongst pregnant women are more effective with wearable devices than without (Chan & Chen, 2019) and activity bracelets are found to be feasible among pregnant women (Grym et al., 2019). An interactive learning environment in a GDM app that enables exploring the effect of lifestyle on blood glucose This allows creating causalities between health behaviors (diet, physical activity and sleep) and blood glucose, which are unclear for women with GDM, especially right after the diagnosis (Carolan et al., 2012). Providing feedback in the form of praise and suggestion Lentferink et al. (2017) showed that praise and suggestions are especially useful behavior change techniques in mobile apps supporting a healthier lifestyle. Personalization accounts for the perceived uniqueness of the pregnancy (Peyton et al., 2014) and the perceived usefulness of pregnancy apps (Muuraiskangas et al., 2016). Providing different options on how to achieve goals for physical activity Harrison et al. (2019) found that women with GDM wish to have clear goals for physical activity while still retaining autonomy. Partners support the women in terms of accompanying them on walks for exercise and encourage them to adhere to the GDM diet (Carolan et al., 2012). In general partner's support is seen as especially valuable in effecting a behavioral change of women with GDM such as increasing exercise (Downs & Ulbrecht, 2006). . Designing the application to be more personal can motivate users to use the apps, for example, by increasing the perceived usefulness of the pregnancy app (Muuraiskangas et al., 2016). Muuraiskangas et al. (2016) noted in their study on a pregnancy app that while their app got acceptable usability ratings, low perceived usefulness was a large barrier for actual app use due to lack of personalization. The developers of SDT, Deci and Ryan (2012) discuss the relationship between giving clear recommendations concerning lifestyle improvement and letting patients decide how they change their behavior. Their view is to offer advice while being "autonomy supportive." They argue that the directness of advising depends largely on the person and level of intrinsic motivation. In the case of GDM, it seems that women want to have clear and personalized recommendations for physical activity (Harrison et al., 2019), which indicates that women with GDM would benefit if they were provided multiple different suggestions for physical activity where they could choose the one they like to do or the one with least discomfort, respecting the autonomy of women with GDM.
In addition to competence (see 1. Desirable feature) and autonomy, relatedness can be considered important as well, as qualitative studies on women with GDM have emphasized the importance of social support from others, especially their partner (Carolan, 2013;Draffin et al., 2016). Wenger et al. (Wenger et al., 2014) also listed an opportunity to compare individual symptoms and physiology with other mothers as one motivational factor for self-tracking amongst pregnant women. We discuss this social support next.

Desirable feature 3: Provide social support, especially from the partner
The importance of social support in pregnancy (Huberty et al., 2016;Hui et al., 2012;Wierckx et al., 2014), and specifically the partner's central role, has been found in several articles (Carolan, 2013;Gao et al., 2014;Lee et al., 2016;Peyton et al., 2014;Prabhakar et al., 2017;Van Dijk et al., 2016) and in GDM (Carolan, 2013;Carolan et al., 2012;Craig et al., 2020;Downs & Ulbrecht, 2006;Draffin et al., 2016;Persson et al., 2010). Furthermore, Prabhakar et al. (2017) noted that the lack of or low partner support greatly increases the risk for postpartum depression. Other important sources for support mentioned were the women's own mothers, sisters, and friends (Prabhakar et al., 2017). It also seems that women turn to different people in search of different kinds of support. Based on interviews, in one study the participants mostly communicated their concerns regarding the physical part of the pregnancy to their own mothers and sought emotional support from their partners (Hui et al., 2012).
Some pregnancy mHealth solutions utilize social media and social networking features to provide support, but research on the effects of these components is limited. For Table 6. Example implementations with justifications for desirable feature 4.

Implementation Justification
A comprehensive information section about pregnancy with information similar to pregnancy apps (e.g., effects of pregnancy, development of the child) This section has been highly appreciated in GDM apps (Skar et al., 2018).
A comprehensive information section about GDM (e.g, when and how to measure blood glucose, what to eat, how to exercise) Women with GDM want to have reliable information about their disease (Carolan, 2013;Draffin et al., 2016;Skar et al., 2018).  Phelan, 2010), however, the motivation to a healthier lifestyle is decreased after birth (Skar et al., 2018). Simple visualizations (such as using traffic lights metaphor or arrows) for showing real-time feedback and line charts for providing detailed information.
Supporting the glanceable visualizations and in-depth information are needed to support the different needs of the users (Desai et al., 2018;Li et al., 2011) Visualization of trends in blood glucose The use of trends is important in creating simple rules between blood glucose and lifestyle. This reasoning operates between sensemaking and habitual reasoning (Katz et al., 2018b).

Implementation Justification
A chat to discuss with health care professionals Studies show that a two-way communication channel is an important feature for the effectiveness of GDM apps (Miremberg et al., 2018;Xie et al., 2020).

Use of app in the meeting with a nurse
The perceived usefulness of the GDM app is highly impacted by how it is used with health care personnel. (Muuraiskangas et al., 2016;Skar et al., 2018). example, Herring et al. (2016) developed a mHealth intervention with a private Facebook group for the participants with a function of social support. The intervention was effective in controlling gestational weight gain with overweight or obese women. However, due to the study design, it was not possible to determine if the social media component contributed to the intervention's effectiveness or not (Herring et al., 2016). In a meta-analysis on social media and mHealth apps on pregnancy care, Chan and Chen (2018) did not find interventions with social networking features to be significantly better in terms of intervention effectiveness compared to interventions without these features. However, there seems to be some demand for social features in pregnancy mHealth; Wenger et al. (2014) included a social networking feature in their app to offer social support and to make it "fun," based on user feedback that the application was too clinical. Based on insights from a user study, Wierckx et al. (2014) also included the possibility to connect with other pregnant women through a social media community. Peyton et al. (2014) noted instead that their participants were not really interested in using a social networking feature, although they too viewed social support as important and helped them get through the pregnancy.
The important role of social support could be considered in application design by providing an opportunity to connect with experts and other pregnant women and by also making the application suitable and interesting for the partner. This could be achieved by not making the application too feminine (Peyton et al., 2014) or by providing the partner their own version of the app. The application could also have functions such as bump notifications to offer support messages itself (for example, as proposed in Huberty et al. (2016)), However, more investigations are required to determine how this could be implemented such that women with GDM do not feel controlled but are supported in making healthier lifestyle decisions. Showing simple notifications to the partner, without accessing the woman's self-tracking data, would be a good starting point as the amount of data disclosed to the partner would be minimized. The user could then decide if they would be willing to share more data with their partner. Also, the partners' opinions about the intended functionalities for them should be studied more. These investigations have been conducted for pregnancy apps designed for men (Thomas et al., 2018), but are lacking in GDM apps.

Desirable feature 4: Support normal pregnancy and de-bunk myths about GDM
Qualitative studies on the experiences of women with GDM have indicated that the attention of healthcare providers is focused more on GDM than the pregnancy itself (Parsons et al., 2014). This further leads to the feeling of loss of normal pregnancy. To avoid this, supporting a normal pregnancy such as visualizing the development of the child should be one of the key features of the application. Skar et al. (2018) reported that providing information about normal pregnancy was one of the key features of the GDM app. However, emphasizing normal pregnancy should not ignore the fact that some women with GDM tend to neglect and forget their disease (Parsons et al., 2014). Thus, both pregnancy in general and GDM should both be incorporated into the app. Sharing baby activity with close people and supporting a healthy pregnancy by countering health risks (in addition to GDM-related risks) for the mother and the child are important (Overdijkink et al., 2018;Peyton et al., 2014).
Interviews with women with GDM revealed the high importance of providing reliable information about managing GDM. Skar et al. (2018) highlighted the findings (Carolan, 2013;Draffin et al., 2016) that the diagnosis is a surprise and that reliable information about how to manage GDM (particularly what to eat and how the child is developing) is one of the most important features of the GDM app (Skar et al., 2018). Although GDM lasts only until the end of pregnancy, providing knowledge on the postpartum effects of GDM during pregnancy is of high importance. There are two kinds of beliefs among women with GDM about what happens to GDM postpartum; one is "GDM is transient" and the other is "T2D is inevitable" (Parsons et al., 2014). Thus, reliable information about the postpartum effects of GDM, such as statistics about the incidence of T2D, should be provided.
In general, pregnant women have several pregnancyrelated uncertainties. These include appropriate nutrition, risky behaviors during pregnancy, the normal course of pregnancy, weight management, and safe physical activity (Goetz et al., 2017;Huberty et al., 2016;Peyton et al., 2014;Wierckx et al., 2014). For example, women with GDM wanted "clear and practical messages from credible sources" regarding physical activity (Harrison et al., 2019). Furthermore, it was noted that most women first turn to "Dr. Google" with their concerns, since such information is obtained easily and rapidly from the internet (Garnweidner et al., 2013). This also applies to women with GDM. While healthcare professionals are considered the most reliable source of information, searching for information on the internet is more frequent (Carolan, 2013). The other concern is that there is no guarantee that information from the internet is reliable, which can leave the women feeling even more unsure than they were before (Peyton et al., 2014). It is clear that reliable information about pregnancy-related concerns is needed, and apps would be a suitable way to provide that information as easily as "Dr. Google." Pregnancy-related myths seem to be one of the main concerns regarding healthy pregnancies in developing countries (e.g., the context of Pakistan (Abid & Shahid, 2017;Sajjad & Shahid, 2016)). Findings from user studies with pregnant women show that practices and beliefs regarding diet and exercise during pregnancy vary from culture to culture and that sometimes they are not compatible with advice the women receive from the apps or healthcare professionals (Greenhalgh et al., 2015). For example, in South Asia, physical exercise during pregnancy is uncommon because of the belief that exercise during pregnancy is not safe (Greenhalgh et al., 2015). According to Peyton et al. (2014), some false beliefs are persistent even in developed countries like the USA; these include the belief that every pregnancy is unique and therefore the universal guidelines can be seen as something that do not apply to the woman's own pregnancy. Peyton et al. (2014) also discuss the related issue of pregnant women's feeling that they lack control over their bodies. They state that women often feel like "the pregnant body is something to which they listen and submit, rather than something they actively manage." These myths or false beliefs about pregnancy can in the worst case be very harmful if for example, the "uniqueness myth" leads to pregnant women disregarding official recommendations and guidelines. Reliable information conveyed by mHealth or eHealth solutions can also help debunk these myths or false beliefs about pregnancy, GDM, and what is good for the mother and the child.

Desirable feature 5: Support dual processing as pregnancy is life-changing
We suggest applying dual processing in a more general way for supporting behavioral change. This approach can be considered particularly feasible for pregnant women as pregnancy provides a fertile time for changing habits (Arabin & Baschat, 2017;Phelan, 2010). However, the "teachable moment" (Phelan, 2010) for healthier habits is rather short for women with GDM, as creating a habit takes approximately 9 weeks on average (Lally et al., 2010) and the GDM app is designed to be used for 12 to 28 weeks. To support habit formation using mobile phones, Stawarz et al. (2015) investigated the effect of reminders and positive reinforcement on habit formation of reporting daily meals. They observed that reminders increased habit formation but decreased the automaticity of habit. Moreover, Skar et al. (2018) noted that motivation to manage GDM and to use the GDM app after giving birth decreases. These results indicate that phone-based cues (such as notifications and reminders) at certain times may not be the most optimal. For the women with GDM, the new habits could be coupled with actions they need to do anyway due to GDM, such as to measure blood glucose frequently.
Dual processing (sensemaking and habitual levels) can be also applied for interpreting glucose data visualizations (Katz et al., 2018b). As tracking glucose levels is the most important feature of the GDM app (Skar et al., 2018), the way to visualize this information should be carefully chosen. Research on how to present and visualize the glucose data has gained considerable attention recently (e.g., (Desai et al., 2018;Katz et al., 2018a, Katz, et al., 2018b), and the results indicate that simple metaphors such as traffic lights and arrows indicating trends are feasible for many. On the other hand, Li et al. (2011) argue for the importance of visualizing a lot of data at the beginning of the usage of a self-tracking app. This supports the discovery phase by allowing users to create an understanding of their data. Thus, both a glanceable overview and a more detailed visualization of glucose levels should be provided. This supports "fast" and "slow" reasoning mechanisms (Kahneman, 2011) and allows for the creation of simplifications regarding causalities between glucose levels and affecting factors (Katz et al., 2018b). Katz et al. (2018b) call the process where fast and slow reasoning mechanisms are combined as "contextual fluid reasoning" and argue that this is an important "middle ground area" in making decisions concerning diabetes self-management. "Contextual fluid reasoning" is more rational than habitual processes but requires less effort than "sensemaking" processes, thus being suitable for frequent everyday decision making. This approach would be especially useful for women with GDM who need to gain an understanding of the causalities as quickly as possible due to the rather short time span of GDM (approximately 12-28 weeks).

Desirable feature 6: Integrate the application with healthcare services
Although the aim is to design a GDM app without additional human power from health care professionals, integration to the current health care system is important. A review on technological support in diabetes (Greenwood et al., 2017) emphasizes the two-way communication between diabetics and health professionals, and among patients diagnosed with T2D, communication with healthcare professionals has been shown to increase knowledge and improve self-management of diabetes (Holmen, 2014;Orsama et al., 2013). Even oneway communication from healthcare professionals to patients through personalized messages delivered through an app is potentially beneficial in diabetes management (Holmen, 2017;Orsama et al., 2013) and in supporting a healthy lifestyle overall (Hawkins et al., 2008).
Tighter integration between health professionals and the app can be expected to increase the effectiveness (Xie et al., 2020) and the perceived usefulness of the GDM app (Greenwood et al., 2017;Muuraiskangas et al., 2016). A metaanalysis of the effectiveness of telemedicine for GDM studies (N = 32) showed significant improvements in glycemic control (Xie et al., 2020), where each study in the meta-analysis involved health care professionals to provide individualized feedback for women with GDM. In the studies included in the meta-analysis (Xie et al., 2020), the feedback frequency varied from every day (Miremberg et al., 2018) to once a week (Kim et al., 2019). When there has been less involvement of health care professionals, the effectiveness of telemedicine has been decreased. For example, in a study by Mackillop et al. (2018), the number of visits to maternal clinics was halved with the intervention group having a GDM app, no statistical differences were found between the control and intervention group in terms of glycemic control. Moreover, if the recommendations are not individualized by health care professionals, the effectiveness of GDM mobile apps seems to be reduced (Kennelly et al., 2018).
Qualitative studies on experiences of women with GDM have highlighted the importance of personal care with health professionals, especially in terms of social support and providing reliable information (Carolan, 2013;Draffin et al., 2016;Parsons et al., 2014;Persson et al., 2010). To support this, the GDM app should be considered more as a part of healthcare services rather than just an app. This is particularly important as pregnant women frequently visit a maternity clinic during pregnancy. A recent study with a GDM app has emphasized the importance of healthcare support for the app such that the app is part of the treatment path rather than a separate tool (Skar et al., 2018). For example, the requirement of entering Table 9. Desirable features and their existing implementations in recent GDM apps.
This was found to be important but technical difficulties decreased the usage. Real-time readings about the glucose level were requested by the women with GDM.
Only automatic transfer of fingertip glucose measurements over Bluetooth. Automatic transfer of fingertip glucose measurements over Bluetooth and an accelerometer to detect physical activity. Also, manual start/stop for recording physical activity.
2. Increase autonomy by enabling personalization Users were able to create a profile, which included information on their perceived level of physical activity before pregnancy; preferred food culture; and weight and height before pregnancy. However, it remains unclear how this information was used.
Not implemented.
Sending reminders at patient-specific mealtimes and context.
3. Provide social support, especially from the partner Not implemented.
Not implemented.
Not implemented.
4. Support normal pregnancy and de-bunk myths about GDM A question and answer section was provided and healthy eating was demonstrated with pictures and the help of smileys. Some reported an increase in knowledge about healthy eating, drinking, and physical activity. However, more detailed information was requested by women with GDM.
Not implemented.
Nutrition recommendations.
5. Support dual processing as pregnancy is life-changing Visualization showing the recent readings in a line graph and an overview graph.
Weekly overview and all glucose readings in a scatter plot. Sound and a reminder displayed at the time of the recommended postprandial reading.
Sending reminders of measurements.
6. Integrate the application with healthcare services Users can print their blood sugar values to discuss them with their health professional. However, a lack of further integration with healthcare professionals was found to be an issue.
Healthcare professionals had access to patients' glucose levels and contacted patients to provide feedback about blood glucose values.
Notifications and reminders to healthcare providers.
the glucose values into the app and writing down the glucose measurements with a pen and paper, and then reporting these values to a healthcare professional requires double recording of glucose values, which decreases the motivation to track glucose values in the GDM app (Skar et al., 2018). If healthcare professionals are involved in the GDM app, their views should be considered when designing it. In particular, the amount and quality of non-medical data provided by the GDM app (e.g., through an activity bracelet) should be agreed upon with healthcare professionals (West et al., 2016(West et al., , 2018. West et al. (2018) noted that automatic data collection could be one way of increasing the acceptability of using selftracking data amongst healthcare professionals by improving the completeness of the data and by reducing the need for recalling. Recently, Garnweidner-Holme et al. (2018) studied the healthcare professionals' perspective on the most important features of the GDM app. Garnweidner-Holme et al. (2018) found that cultural sensitivity, especially support for the mother's native language, is one of the most important features in the application. They also argued that providing information through a mobile phone is a very good delivery mechanism as the information is not lost as easily as printed papers and information in mobile phones could be referred to repeatedly after the consultations.

Suggested desirable features in existing GDM apps
In this section, we investigate how the presented six desirable features derived from the literature have been implemented in the research of GDM apps with the most extensive studies  GDm-Health (Hirst et al., 2015;Mackillop et al., 2018Mackillop et al., , 2014, and MobiGuide (Peleg et al., 2017;Rigla et al., 2018). These applications were also described with an adequate level of detail (such as screenshots) to evaluate how each of the desirable feature was implemented. The material for the analysis was gathered from the related publications listed above. The results of the evaluation are shown in Table 9. Providing feedback (1. Desirable feature), and social support (3. Desirable feature) in particular have been largely neglected in the chosen GDM apps. Most of the desirable features (3/6), namely support for personalization (2. Desirable feature), providing reliable information and debunking myths (4. Desirable feature), supporting habitual and reflective decision making (5. Desirable feature), and combining with existing healthcare practices (6. Desirable feature) are more commonly implemented. However, these recommendations are still missing from about half of the applications. None of the applications implemented all of the desirable features.

Discussion
Most of the studies on GDM apps have investigated the effects of these apps on medical outcomes or user experience. However, much less work has focused on constructive approaches for investigating the design features of GDM apps. This paper aimed to fill the gap by designing and providing examples of features that have justifications in the literature and by providing the relevant perspectives of women with GDM.
We propose six exploratory desirable features derived from a wide selection of papers from relevant areas for designing mobile GDM apps that support behavior change. While we believe that incorporating all of these desirable features into a GDM app is difficult and might not be feasible, considering and implementing some of them properly could be sufficient to improve the efficacy of apps designed for GDM management. In fact, Hood et al. (2016) found in a review of diabetes apps that the number of features available was negatively correlated with usability, indicating that simplistic approaches are preferred. Moreover, the relative importance of the proposed guidelines for improving efficacy is largely unknown. This is because the efficacy of different features of diabetes apps are rarely evaluated separately (Hood et al., 2016). However, recent studies have shown that providing daily feedback by health care professionals improves glycemic control of women with GDM (Xie et al., 2020). Healthcare professionals are key persons for providing knowledge and feedback on the situation of women with GDM (Craig et al., 2020;Parsons et al., 2014). Individualized feedback from healthcare professionals based on data provided by an application supports a wide range of the recommendations, in particular, desirable features on increasing competence with feedback (1. Desirable feature), supporting personalization (2. Desirable feature), debunking myths (4. Desirable feature), and integrating the app into the health care services (6. Desirable feature).
As our goal is to implement desirable features without the need for additional work from health care professionals, some of this feedback should be generated automatically, enabled by comprehensive self-tracking (1. Desirable feature) and machine learning approaches . The presentation of this feedback should support habitual decision-making (e.g., alarms) and reflective decision-making (e.g., detailed graphs) and thus support dual processing (5. Desirable feature). However, we acknowledge that personal contact with healthcare professionals is very important to women with GDM (Parsons et al., 2014) and thus feedback should not be fully automated.
Our results are in a line with Zhao et al. (2016), who found that features such as less time consumption and real-time feedback (related to 1. Desirable features), individualized elements (2. Desirable feature), detailed information (4. and 5. Desirable feature), and health professional involvement (6. Desirable feature) are key features in mobile apps for supporting a healthy lifestyle in general. However, the evaluation on how the desirable features were implemented in existing mobile GDM apps revealed that many of the desirable features were either not implemented at all or the implementations varied considerably between apps. Pregnant+ and MobiGuide provided some functionalities that were related to 5/6 Desirable features. However, these 5 desirable features were not fully achieved, for example, Pregnant+ required manual input for tracking physical activity, suffered from technical problems, and integration to existing healthcare practices was very limited (i.e., printing glucose levels for the nurse). The desirable feature of supporting dual processing in visualizations was implemented the furthest, as apps provided visualizations at varying levels of details and reminders.
The costs of implementing the presented desirable features are affordable, when considering the price of implications of GDM for the health care system (e.g., increased number of cesarean sections and increased prevalence of T2D postpartum). 1. Desirable feature requires the usage of CGM during the pregnancy (cost around $800 for 16 weeks (Hoskins, 2021)) and physical activity tracker (cost around $100 (Link, 2021)). However, given that only a cesarean section costs around $9 000 more than vaginal delivery (Hurst, 2021;Mathews et al., 2021), the costs of these devices are around one-tenth of the cost of one cesarean section. Not to mention the long-term costs of GDM, such as the increased prevalence of T2D. The following four Desirable features (2-5) can be expected to be even more cost-effective than Desirable feature 1 as they can be implemented without inducing other devices than a mobile phone for the user. Regarding 6. Desirable feature, the amount of reduced costs depends on how many working hours by health care personnel can be saved due to less frequent visits to maternal care by women with GDM. In general, cost evaluations are largely missing from the studies evaluating the efficacy of mHealth technologies for managing GDM (Ming et al., 2016;Rasekaba et al., 2015). One exception is a study by Mackillop et al. (2018), where participants with mHealth intervention had £1000 smaller delivery costs compared to the control group. However, the difference was not statistically significant. Nevertheless, cost evaluation should be considered as an important aspect of future studies on evaluating the efficiency of the presented features for the management of GDM.

Limitations and future work
As the literature for each of the themes described in Section 2 is large, conducting a completely comprehensive literature review for each of the topics is beyond the scope of this work. As we acknowledged this as a limitation, we included approximately 100 relevant papers for establishing the desirable features. While we believe that this was sufficient to gain insight for providing exploratory desirable features, we do not claim that the proposed desirable features are inclusive. Regarding the investigations on how the desirable features are implemented in current GDM apps, we chose four GDM apps based on the work by Nikolopoulos et al. (2019), who screened the apps based on the extent of evaluation and use. However, the number of GDM apps is estimated to be between 10 and 20 (Ming et al., 2016), some relevant GDM apps may be unintentionally excluded from current investigations.
In the future, these features should be prototyped, evaluated, and developed further with women with GDM. Further, they should be integrated into a functional GDM app to investigate their impact on GDM self-management. More research is required to understand the efficacy of each exploratory desirable feature provided here. In addition, the feasibility of implementing each desirable feature is largely unknown. However, the implementation of providing reliable information can be considered much easier than the implementation of habit formation and tracking, for example.
In this study, we investigated the design of a GDM app that would be adopted at diagnosis and used until birth. Future work should include investigations on how mobile apps could support a healthier lifestyle before diagnosis and postpartum. In fact, preventive actions should be prioritized when considering the most cost-effective ways to decrease the impact of GDM on the population. Individualized face-toface counseling at the beginning of pregnancy has been shown to decrease the incidence of GDM and gestational weight gain (Koivusalo et al., 2016). Providing this counseling through a mobile pregnancy app could be a cost-effective means to significantly impact GDM incidence. In addition to considering the intervention before the GDM diagnosis, it is important to examine interventions after pregnancy. With regards to supporting behavioral change postpartum with an app, the use of a GDM app during pregnancy may also lower the threshold to use the app after pregnancy. Thus, the use of the same application could be continued postpartum. However, the features would be different; for example, pregnancy-related information should be replaced with information related to children, such as feeding recommendations. Thus, the design of an app that considers issues before GDM diagnosis and postpartum should be studied. According to our knowledge, there is only one study that has investigated app design for the period after the GDM (O'Reilly & Laws, 2019). As relevant background material for the design, Dennison et al. (2019) provide 20 recommendations on how to support a healthier lifestyle postpartum in order to avoid T2D.

Conclusions
In this paper, we sought to explore desirable features that can increase the impact of GDM apps. Based on our literature review, which included studies on experiences with GDM apps and GDM itself, we synthesized six exploratory desirable features for future apps designed for GDM management. We found that these desirable features have been only partially implemented in GDM apps. We expect that the proper implementation of at least some of these recommendations will increase the efficacy of GDM self-management apps.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
This work was supported by Business Finland [860/31/2018] under Project eMOM GDM.