Effectiveness of mobile health interventions for pregnant women with gestational diabetes mellitus: a systematic review and meta-analysis

Abstract Gestational diabetes mellitus is a growing global health problem. Inadequate management during pregnancy can lead to maternal and foetal complications. Currently, mobile health (mHealth) delivers healthcare services, playing an increasingly important role in the management of blood glucose in GDM. This study aimed to systematically evaluate the effectiveness of mHealth intervention in pregnant women with GDM. Based on randomised controlled trials of mHealth application in GDM patients searched from the database, literature screening, data extraction, and quality evaluation were conducted independently by two researchers. Statistical analysis was performed using Review Manager 5.4 software. The review included 27 studies with a total of 3483 patients. The results showed a significant improvement in glycemic control. In addition, mHealth interventions could reduce the occurrence of adverse pregnancy outcomes and improve self-management ability. In a subgroup analysis, recording of delivery mode and WeChat combined phone call indicated significant differences with mHealth interventions. It was suggested that mHealth interventions imposed a positive effect on glycemic control and reduction of adverse pregnancy outcomes in GDM patients. Our results demonstrated that the application of mHealth interventions can act as an effective and feasible approach to self-management to promote the self-management level and awareness of GDM patients.


Introduction
Gestational diabetes mellitus (GDM), one of the common complications of pregnancy, is defined as the first occurrence of diabetes due to maternal abnormal glucose metabolism during pregnancy (McIntyre et al. 2019).The rapid increase in the number of people living with GDM is a global crisis that places a huge burden on public health systems.In China, the overall prevalence of GDM increased from about 12.5% from 2015 to 17.5% in 2020 (Diabetes Branch of Chinese 2022).Globally, about 1.64 billion pregnant women were affected by GDM with an overall prevalence rate of 14.0% and a prevalence rate of 7.1% � 31.5% in all continents (Wang et al. 2022;Paulo et al. 2021).According to the report of the International Diabetes Federation (IDF), the standardised prevalence of GDM in low-, middle-and high-income countries were respectively 12.7%, 9.2% and 14.2% (Wang et al. 2022).Generally, GDM is associated with poor pregnancy outcomes, which not only increases the risks of pregnancy infection, polyhydramnios, premature delivery, birth injury and postpartum infection, but also leads to foetal hypoxia, foetal weight exceeding the normal range, neonatal hypoglycaemia and macrosomia (Giannakou et al. 2019).Besides, GDM also brings long-term health consequences for both mother and offspring, it was reported that women with prior GDM encountered a tenfold increased risk for type 2 diabetes compared to those without GDM history (Liang et al. 2021).
The number of smartphone subscriptions worldwide today surpasses six billion and is forecast to further increase by several hundred million in the next few years (Taylor 2023).Notably, the rapidly growing and large population of mobile phone users, especially in low-and middle-income countries, has prompted smartphones to function as powerful tools for enhancing efficiency in the health sector (Peprah et al. 2019).With the increasing popularity of mobile phones, a growing body of evidence is available to support the use of mobile phones to improve various medical conditions.
Mobile health (mHealth) as a new development direction of innovative health care is defined by the Word Health Organisation as 'medical and public health practice supported by mobile devices' (Risling et al. 2017) and developed with solutions including health education, inquiry of medical information, electronic health records, disease risk assessment, online disease consultation, electronic prescription, remote consultation, remote treatment and rehabilitation (Liu et al. 2018).In recent years, mHealth interventions have been proposed as promising strategies for delivering health interventions to people suffering from chronic diseases (Lunde et al. 2018), such as hypertension, diabetes, chronic heart failure, and asthma (Echarri et al. 2020).Given these advantages, mHealth also plays an increasingly important role in glycemic control and self-management of pregnant women with GDM (Mackillop 2018;Kim 2019) .
In recent years, more and more mHealth applications have been developed for GDM patients (Tian et al. 2021;Rasekaba 2021).Up to now, some mHealth programmes on the management of gestational diabetes have been implemented.Nevertheless, the resulting evidence is always limited and inadequate.An updated review of 32 studies revealed that mHealth interventions contributed to beneficial impacts on the glycemic level as well as certain maternal and neonatal/foetal complications in patients with GDM compared to the effects of standard care.However, this study did not explore the impact of mobile health on self-care management behaviours such as weight, diet, exercise, blood glucose monitoring in pregnant women with GDM (Xie and Dai et al. 2020).Furthermore, another meta-analysis manifested that mHealth interventions are mainly used to record and transmit blood glucose values to clinical nursing teams.No sufficient evidence suggests that mHealth interventions can produce significant effects on glycemic control of pregnant women with gestational diabetes mellitus (Ming et al. 2016).Laursen et al. assesses the effects of mHealth solutions for managing gestational and pregestational diabetes, the meta-analysis reveal no evidence that mHealth as an alternative to usual care when considering maternal and foetal outcomes (Laursen et al. 2022).
Given the limited evidence available for the advantages of mHealth interventions in women with GDM aforementioned, this systematic review and meta-analysis aimed to evaluate the effectiveness of mHealth in pregnant women with GDM.We also analysed the self-management behaviour of pregnant women with GDM.In addition, subgroup analysis was performed according to the different mobile health intervention types to provide a better management model for GDM.

Materials and methods
This study was performed by the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (Moher et al. 2009).This systematic review was prospectively registered with PROSPERO International Prospective Register of Systematic Reviews (CRD42022315689).Because no human participants were involved, the review was exempt from institutional review board approval.

Search strategy
We used the PubMed, Web of Science, Medline, Scopus, Cochrane Library, Chinese National Knowledge Infrastructure, Wan fang, China Biology Medicine (CBM), Weipu Database and International Clinical Trials Registry Platform (ICTRP) to search for relevant randomised controlled trials (RCT) comparing mHealth intervention to standard care from the database inception to Dec 31st, 2021.The appropriate materials were obtained by searching titles, abstracts, keywords, and Mesh terms, only using English and Chinese languages (Supplementary Table 1).There was no restriction on language or publication status.Meanwhile, we also searched the reference list of the selected articles manually to ensure that all eligible studies were included in this review.

Inclusion and exclusion criteria
The inclusion criteria, based on the PICOS framework, were as follows: (1) Participants: patients aged �18 years old and diagnosed with GDM; (2) Intervention: The study evaluated studies that analysed mHealth interventions for gestational diabetes.Mobile phone text messaging, wearable or portable monitoring devices, and applications running on smartphones were all classified as mHealth interventions; (3) Control: the control group received standard care; (4) Outcomes: the primary outcomes were indexes of glucose control including fasting blood glucose (FBG), 2h-postprandial blood glucose (2hPBG), and glycated haemoglobin (HbA1c).The second outcomes contained caesarean delivery, pregnancy-induced hypertension (PIH), postpartum haemorrhage, polyhydramnios, premature rupture of membranes, macrosomia, the preterm delivery, neonatal hypoglycaemia, and intrauterine foetal distress; (5) Study type: randomised controlled trials.The excluded criteria were: (1) Any other types of diabetes including type 1 or type 2 diabetes.(2) The studies lacked available data or the full texts were inaccessible.(3) Any other type of paper (e.g.review, meta-analyses).

Data extraction
Two researchers (WHX and LTY) independently conducted the literature search and screening according to both inclusion and exclusion criteria, and then cross-checked the screening results using the document management software Notexpress.The extracted data mainly included the author, country/region, sample size, intervention measure, control measure, outcomes, and details of mHealth interventions (assessment, monitoring, education, interaction, and reminder).Differences in assessment were resolved through mutual exploration among the reviewers until consensus was reached.Some data were transformed into mean and standard deviation using a data inference algorithm according to the needs of the meta-analysis.Differences between researchers were resolved by a third researcher through discussion or arbitration (YYL).

Risk of bias assessment
The two reviewers dependently evaluated the risk of bias of the included research according to the Cochrane Risk of Bias Assessment Tool.The evaluation contents included 7 components: random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias), and other potential biases.The risks of Cochrane standard determination bias were: (1) high risk; (2) low risk; (3) unclear risk (Jackson and Waters 2005).

Data synthesis and analysis
Revman 5.4 software was used to analyses data from all included studies.Study results were reported as Standardised Mean Differences (SMD) for continuous outcomes and Relative Risks (RR) for binary outcomes, accompanied by 95% confidence intervals (CI).Summary effects were calculated using fixed effects meta-analysis or random effects meta-analysis if there was high heterogeneity (I 2 statistic > 50%).The significance level of the meta-analysis was set to 0.05.Sensitivity analysis was performed using the software, excluding one trial at each time according to the methodological quality of evidence synthesis recommended in the JBI Manual (Zeng et al. 2015).The subgroup analysis was conducted based on delivery mode of mHealth interventions.

Search results
The PRISMA flow chart of study selection was introduced in Figure 1.A total of 1468 studies were initially retrieved, of which 736 were excluded because of duplicate records, and additional 619 unrelated studies were removed by screening the titles and abstracts.After verifying the full text of the remaining 113 studies, 86 articles were excluded for the following reasons: the study design and participants did not satisfy the inclusion criteria (9); the outcomes were inconsistent (35); the interventions were inconsistent ( 16); or we were unable to access the data (9).Ultimately, 27 studies met the inclusion criteria for our study.

Basic characteristics of the included studies
Table 1 presented information of the included trials involving 3483 patients with GDM.The majority of studies were published after 2019.27 studies were conducted in China, 2 studies in USA, 2 studies in Korea, 1 study in UK, and 1 study in Israel (Chen et al. 2020, Chen et al. 2021;Chen, Lin, et al. 2020;Guo et al. 2019;Homko et al. 2007;Homko et al. 2012;Hong 2021;Huo 2021;Jiang et al. 2019;Jiang and Wei 2019;Sung et al. 2019;Kang 2021;Kim et al. 2019;Liu et al. 2020;Lu et al. 2020;Mackillop et al. 2018;Miremberg et al. 2018;Wu 2021;Xie 2019;Xue et al. 2018;Xue and Du 2018;Xue and Zheng 2019;Yang et al. 2018;Yang and Lu 2021;Zeng and Zeng 2021;Zhang 2021;Zhu et al. 2020).A total of 3473 participants were included in this meta-analysis.The intervention group was assigned 1729 participants, and the control group comprised 1754 participants.

Characteristics of the mHealth program
The characteristics of mHealth interventions were summarised in Table 2.The characteristics of the included studies were identified and classified as follows: details of the mHealth program, delivery mode, communication type, frequency, and leader.The details of the mHealth program included assessment, monitoring, education, interaction, and reminding.Notably, almost all of the studies conducted health education through mHealth, including at least two details.The mHealth interventions of the included studies were delivered by means of websites, apps, WeChat (a type of social media that has gained considerable popularity among the young population in China), and mobile phone calls.The delivery methods were mainly self-developed mHealth app and WeChat group or official account based on WeChat platform, and a few studies utilised GDM management platform (n ¼ 4) based on Internet.Through adopting both one-way and two-way communication types, mHealth interventions were implemented mainly by nurses, doctors, nutritionists and psychologists who provided guidance.The frequency of communication varied from once a day to once a week.Some studies determined the intervention time based on patients feedback.

Risks of bias of the studies
Figure 2 showed the graph of risk of bias in included studies.Among the 27 RCTs, 7 trials showed a high risk of randomised selection, and 24 trials were unclear about the description of allocation concealment.25 trials had no reporting bias and none of the trials appeared to have attrition bias.Only four trials had sufficient blinding for patients, researchers and evaluators.6 trials implemented adequate blinding for outcome assessors.More than half of the studies were at high risk due to the lack of blinding.

Fasting blood glucose
24 studies with a total of 2946 patients were enrolled to compare fasting blood glucose in the two groups (Figure 3).Due to high heterogeneity among the studies, (p < 0.00001, I 2 ¼95%), a random effects model was selected for analysis.The result reflected that the fasting blood glucose level of the mHealth intervention group was lower than that of the standard care group [SMD=-1.02,95%CI (-1.30, 0.75), p < 0.0001].The sensitivity analysis demonstrated that the pooled effect and I 2 statistic changed minimally after itemby-item exclusion.

2-hours postprandial blood glucose. 19 studies of a
total of 2744 patients were enrolled for meta-analyses to compare 2-hour postprandial blood glucose in the two groups (Figure 4).Given high heterogeneity among the studies (p < 0.00001, I 2 ¼96%), a random effects model was selected for analysis.The sensitivity analysis demonstrated that the pooled effect and I 2 statistic changed minimally after item-by-item exclusion.The result showed that the 2hPBG of the telemedicine intervention group was lower than that of the standard care group [SMD=-1.15,95%CI (-1.42,À 0.88), p < 0.00001].

Maternal and neonatal/foetal complications
(1) 14 studies evaluated the effect of mHealth interventions on the caesarean section rate of GDM patients, with mild heterogeneity (Table 3) (p < 0.005, I 2 ¼56%).Through the sensitivity analysis, it was concluded that the study by Yangping was the main source of heterogeneity, and the heterogeneity dropped to 30% after removal.The result revealed that mHealth guidance can reduce the caesarean section rate of GDM patients compared with standard care [RR ¼ 0.67, 95% CI¼(0.52,0.88), p ¼ 0.004].(2) The results of meta-analysis of 5 trials using a fixed-effect model indicated that the incidence of the postpartum haemorrhage in TM group was significantly lower than that in the standard group [RR ¼ 0.22, 95%CI¼(0.11,0.47), p < 0.01], with no heterogeneity among studies (I 2 ¼0, p ¼ 0.43).
(3) The results of meta-analysis of 5 trials using a fixed-effect model showed that the incidence of the polyhydramnios in TM group was significantly lower than that in the standard group [RR ¼ 0.23, 95% CI¼(0.12,0.43), p < 0.01], with no heterogeneity among studies (I 2 ¼0, p ¼ 0.76).( 4) The results of metaanalysis of 6 trials using a fixed-effect model suggested that the incidence of the premature rupture of membranes in TM group was significantly lower than that in the standard group [RR ¼ 0.46, 95% CI¼(0.24,0.88),p < 0.01], with no heterogeneity among studies (I 2 ¼0, p ¼ 0.53).( 5

Self-management behaviour of blood glucose monitoring
3 studies reported the self-management behaviour of blood glucose monitoring in patients with Summary of Diabetes Self Care Activities (SDSCA) scale (Figure 6).On account of high heterogeneity among the studies (p < 0.0001, I 2 ¼93%), the random effect model was selected for analysis.The sensitivity analysis showed that the study of Lu XL was the main source of heterogeneity, and the pooled I 2 statistic dropped to 27% after excluding the trials.The result revealed that the self-management score of blood glucose in the mHealth intervention group was higher than that in the standard group [MD ¼ 1.61, 95% CI¼(0.48,2.75), p ¼ 0.005].

Subgroup analysis
We conducted a subgroup analysis in FBG based on delivery mode of mHealth interventions (Figure 7).The results demonstrated that patients receiving WeChat combined phone interventions benefitted more that those who received other mHealth interventions (such as use of a health app or WeChat only or a web-based platform) with respect to the FBG outcome.

mHealth interventions impose a positive effect on improving patients' blood glucose levels
The prevalence of gestational diabetes mellitus (GDM) is increasing, which poses a serious threat to the health of mother and infant.With the rapid development of science and technology, mHealth applications are increasingly employed in areas such as diabetes management and health promotion, especially in the management of GDM.Improving glycemic control is the cornerstone of diabetes management, as improvements in blood glucose levels can decrease the risk of diabetes-related complications (Qiu et al. 2016).In traditional blood glucose management, doctors can provide lifestyle guidance for pregnant women with GDM through telephone follow-up or outpatient service according to the daily blood glucose monitoring data recorded by pregnant women (Tian et al. 2021).However, outpatient management not only increases the cost of transportation for patients, but also leads to excessive investment in hospital manpower.This study indicated that mHealth interventions have a positive effect on improving patients' blood glucose levels.The mHealth platform can provide dynamic and whole-process standardised blood glucose management for GDM patients, which is conducive to reinforcing patients' awareness and attention to reasonable and scientific control of blood glucose, thereby mobilising their initiative to participate in blood glucose management, as well as blood glucose monitoring.Additionally, it was found from the subgroup analysis that WeChat combined phone call interventions were more effective than other types of mHealth tools on FBG.Currently, WeChat is widely used in daily life and easy to use.Patients can record and upload their blood glucose data, realise mutual communication and discussion, obtain professional guidance, reduce psychological burden and acquire more health information in various forms (Chen, Zhou, et al. 2020).Therefore, mobile phone reminders further strengthen patients' compliance with selfmonitoring of blood glucose.

mHealth interventions decrease metal and foetal complications
Gestational diabetes mellitus increases the risk of pregnancyrelated complications, such as premature birth, preeclampsia, and caesarean section births, which is also related to the subsequent recurrence of GDM and the occurrence of type 2 diabetes after childbirth (Kim et al.2019).Consequently, effective control of blood sugar is especially important to improve the prognosis of pregnant women and foetuses.Compared with guidebook and verbal information, mHealth interventions are more flexible, characterised with more variable modes of communication (text, pictures, sound, interactivity) and functionality (response on blood sugar levels).Information about healthy diets, physical activity and gestational diabetes is always near at hand on womens smartphones (Borgen et al. 2017).
Pregnancy women can monitor their condition and receive prompt guidance from doctors at any time.More importantly, the availability of essential information on GDM, hospital routine and relevant phone numbers may decrease worries, distress and hospital visits.This meta-analysis proved that mHealth reduced the incidence of adverse pregnancy outcomes and promoted maternal and foetal health.At present, the majority of studies only focus on reporting perinatal complications, lacking long-term maternal and offspring outcomes.Therefore, more studies are needed in the future to explore the effects of mHealth interventions on maternal and offspring long-term outcomes.

mHealth interventions improve the selfmanagement level of GDM patients
Effective treatment of GDM requires rigorous and scientific self-management.Self-management of diabetes is still considered as the cornerstone of diabetes care.mHealth apps have been confirmed to enhance self-management of chronic conditions such as diabetes (Echarri et al. 2020).Considering pregnant women are generally motivated to self-monitor their condition remotely and are often adept in using smartphones (Alqudah et al. 2019), individualised management, including exercise, dietary guidance and encouragement for physical activity can be suggested in order to improve glycemic control in patients with GDM (Sung et al. 2019).Twenty-one articles reported on self-management, including diet, exercise, blood sugar monitoring, knowledge of diseases and self-management attitude.At present, in the studies related to the self-management of pregnant women with GDM, the self-management evaluation tools for chronic diabetes or the self-compiled questionnaire are mostly utilised (Harrington et al. 2021).This meta-analysis of studies using the SDSCA scale to evaluate glycemic control showed that mHealth interventions contributed to glycemic control in pregnant women with GDM.GDM patients are especially expected to benefit from telemedicine, because their mature childbearing age and familiarity with the use of mobile phone technology guarantee that they can exhibit a high degree of acceptance of GDM disease-related knowledge, diet, exercise, drugs, psychology and other related guidance, which can promote the change of self-management behaviour and implementation of scientific intervention to reasonably control blood glucose.In conclusion, using mHealth interventions could be advantageous for self-management of glycemic control in patients with GDM.In the future, more studies should pay attention to the role of mHealth in the self-management of GDM.

Implications for clinical practice
mHealth management of GDM has become an effective measure and future development trend.The application of mHealth in the management of GDM patients can realise data sharing, instant individualised guidance, and long-term monitoring of blood glucose, which has good social benefits (Surendran et al. 2021).However, there are still deficiencies in clinical nursing practice.First, there is no complete mHealth system.Some studies only obtain information through the way of inquiry without timely feedback, and some studies have short intervention time, leading to poor compliance of pregnant women.As a consequence, improvements in technology and frequent cooperation between health care personnel and pregnant women should be greatly strengthened to improve participation.In addition, studies have shown that current mHealth places more emphasis on the impact of pregnant women's behaviour change and the medical accuracy of app content, ignoring the psychological needs of pregnant women (El- Gayar O et al. 2021).Since GDM can cause depression and anxiety in pregnant women, projects to increase the emotional experience of pregnant women should be considered in future practice to enhance their real emotional requirements.Moreover, mHealth management takes the Internet as the carrier.Due to the current imperfect laws on Internet use, as well as personal privacy involved, time and money costs, the construction and maintenance of a more secure network environment are indispensable.Health belief models suggest that behavioural triggers may play an important role in the impact of apps on behaviour change (Skar et al. 2018).It has also been shown that the use of Behaviour Change Theories is beneficial to the development of smartphone apps (West et al. 2017).Given existing mobile intervention programs lack theoretical basis, future research should consider increasing behavioural motivation through theoretical support.Currently, there is no scale specifically designed to assess adherence to selfmanagement behaviours among pregnant women with GDM and there is a need to develop valid and reliable tools to assess the compliance of GDM pregnant women with selfmanagement behaviours.An Arabic study translated the SDSCA scale into a simple, effective, and culturally appropriate self-management tool for pregnant women with GDM (Al Hashmi et al. 2022).Based on the advantages of the Internet, future mHealth should use the Diabetes self-management scale to improve the design of the self-management program for pregnant women with GDM, so that pregnant women receive more support from healthcare providers and families to engage in and adhere to self-care activities.

Limitations of the study
Several limitations exposed in this systematic review are as follows.(1) There are few studies on some outcome indicators, and the available evidence is insufficient.(2) There is significant heterogeneity in many outcome indicators in this study, which may be caused by the differences in specific intervention methods, intervention frequencies and intervention contents included.(3) Included studies did not report long-term outcome indicators for mothers and children (e.g.postpartum diabetes, childhood obesity).( 4) Since the sample size of this study involves multiple countries and regions, cultural and physical differences may also lead to the existence of bias.

Conclusions
In conclusion, compared with standard care, Internet-based mHealth interventions is more effective in controlling blood glucose levels and improving maternal and infant outcomes in patients with gestational diabetes.This proves that applying mHealth to the clinical treatment of gestational diabetes mellitus is feasible and has certain practical significance.
Considering the rapid advancement of information and communication technologies and the unbalanced development among regions, further research is needed on the effect of mHealth on gestational diabetes mellitus.

Figure 1 .
Figure 1.Flow diagram of the study selection process.
that the study by Mackillop was the main source of heterogeneity, the pooled RR and I 2 statistic dropped to 0.45 and 0 after excluding the trials.The result indicated that mHealth guidance can reduce the neonatal hypoglycaemia rate of GDM patients compared with standard care [RR ¼ 0.44, 95%CI¼(0.31,0.62) p < 0.001].(8) 8 studies evaluated the effect of mHealth

Figure 5 .
Figure 5. Forest plot of mHealth interventions VS standard care for glycated haemoglobin.

Figure 7 .
Figure 7. Subgroup analysis in FBG based on delivery mode.

Figure 6 .
Figure 6.Forest plot of effect of self-management.

Table 1 .
Basic characteristics of the included studies.

Table 2 .
The characteristics of mHealth interventions.