Mobile Phone–Based Behavioral Interventions in Pregnancy to Promote Maternal and Fetal Health in High-Income Countries: Systematic Review

Background Chronic diseases have recently had an increasing effect on maternal-fetal health, especially in high-income countries. However, there remains a lack of discussion regarding health management with technological approaches, including mobile health (mHealth) interventions. Objective This study aimed to systematically evaluate mHealth interventions used in pregnancy in high-income countries and their effects on maternal health behaviors and maternal-fetal health outcomes. Methods This systematic review identified studies published between January 1, 2000, and November 30, 2018, in MEDLINE via PubMed, Cochrane Library, EMBASE, CINAHL, PsycINFO, Web of Science, and gray literature. Studies were eligible for inclusion if they included only pregnant women in high-income countries and evaluated stand-alone mobile phone interventions intended to promote healthy maternal beliefs, behaviors, and/or maternal-fetal health outcomes. Two researchers independently reviewed and categorized aspects of full-text articles, including source, study design, intervention and control, duration, participant age, attrition rate, main outcomes, and risk of bias. Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines were followed, and the study was registered in PROSPERO before initiation. Results Of the 2225 records examined, 28 studies were included and categorized into 4 themes: (1) gestational weight gain, obesity and physical activity (n=9); (2) smoking cessation (n=9); (3) influenza vaccination (n=2); and (4) general prenatal health, preventive strategies, and miscellaneous topics (n=8). Reported sample sizes ranged from 16 to 5243 with a median of 91. Most studies were performed in the United States (18/28, 64%) and were randomized controlled trials (21/28, 75%). All participants in the included studies were pregnant at the time of study initiation. Overall, 14% (4/28) of studies showed association between intervention use and improved health outcomes; all 4 studies focused on healthy gestational weight. Among those, 3 studies showed intervention use was associated with less overall gestational weight gain. These 3 studies involved interventions with text messaging or an app in combination with another communication strategy (Facebook or email). Regarding smoking cessation, influenza vaccination, and miscellaneous topics, there was some evidence of positive effects on health behaviors and beliefs, but very limited correlation with improved health outcomes. Data and interventions were heterogeneous, precluding a meta-analysis. Conclusions In high-income countries, utilization of mobile phone–based health behavior interventions in pregnancy demonstrates some correlation with positive beliefs, behaviors, and health outcomes. More effective interventions are multimodal in terms of features and tend to focus on healthy gestational weight gain.


Background
Pregnancy and the postpartum period are times of rapid medical, social, and behavioral changes for women and their families.This period is perceived to be a window of opportunity for health interventions because many women have enhanced access to health care during pregnancy and may have increased motivation to improve their health during this time.Healthy maternal behaviors have been shown to improve the risk of pregnancy-related morbidities [1].For example, smoking cessation, exercise, and healthy weight gain in pregnancy have all been linked to better maternal and fetal health [2][3][4].
Chronic disease is a particularly important arena.Per the Centers for Disease Control and Prevention, although the rate of maternal death related to traditional risk factors such as hemorrhage, hypertensive disorders of pregnancy, and anesthesia complications in the United States is decreasing, mortality related to cardiovascular disease (CVD), cerebrovascular accidents, and other medical conditions continues to increase [5].Cardiovascular conditions were responsible for more than one-third of all pregnancy-related deaths in the United States between 2011 and 2016.Thus, pregnancy is an important time to improve health behaviors, such as promoting healthy gestational weight gain and managing chronic disease.However, changes to health behaviors often require intensive provider support, consistent follow-up, and frequent counseling that are difficult to maintain during short outpatient visits.These requirements may be supported by technology.
Many health behavior and lifestyle interventions have incorporated technology in various areas of chronic disease management [6,7].In particular, the field of mobile health (mHealth) has recently seen rapid growth.mHealth refers to the use of mobile technologies including mobile phones, personal digital assistants, and even tablet computers to improve patient health.Outside of pregnancy, a growing amount of literature suggests that mHealth and other digital interventions are feasible, acceptable, and may promote improved health behaviors [8][9][10].An estimated 76% of people in high-income countries own a mobile phone, and 87% use the internet [11,12].Furthermore, a more focused study of pregnant women in the United States showed that 88% had access to a mobile phone, and 89% had access to the internet [13].These data suggest both are promising media for use with pregnant women in the management of chronic conditions in high-income environments.
Past studies suggest women are interested in receiving health information on the Web and are comfortable with using their mobile phones [14].However, research on the use of mHealth in pregnancy has been broad and heterogeneous.Much of the research done is with small groups in low-or middle-income countries or utilizes a telemedicine format, defined as technology-facilitated direct communication with medical professionals [15][16][17][18][19][20][21][22].There remains a lack of organized discussion on mHealth interventions in pregnancy that are tailored or self-maintaining as well as on studies of women in high-income countries, where access to mobile phones is the greatest and women are highly affected by chronic disease

Objective
The objective of this study was to systematically evaluate mHealth interventions used in pregnancy in high-income countries and their effects on maternal health behaviors and maternal-fetal health outcomes.

Study Registration
Before performance of this search, information about the study proposal was published electronically in the University of York PROSPERO register of systematic reviews [23].The authors followed all guidelines for the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [24].

Eligibility Criteria, Information Sources, and Search Strategy
We conducted a systematic review of studies on mobile phone-based mHealth interventions designed for pregnant women.A research librarian (PS) was primarily responsible for a comprehensive literature search.We included English-language articles with a patient population that included pregnant women who utilized pregnancy-related mobile phone interventions during their pregnancy.In addition, we limited our studies to those performed in developed or high-income countries as defined by The World Economic Situation and Prospects 2012 of the United Nations [25].Study types included meta-analyses, systematic reviews, randomized controlled trials (RCTs) including randomized crossover trials and cluster randomized trials, and nonexperimental observational studies.

Study Selection
Titles and abstracts of studies were read by 2 independent reviewers (TH and LY) on two online abstract organizers (abstrackr: [26] and Rayyan [27]).Discordant assessments were resolved by discussion between reviewers or with the involvement of a third author (PS) when necessary.
Studies were then divided into 4 subgroups based on their primary clinical focus: (1) gestational weight gain, obesity and physical activity; (2) smoking cessation; (3) influenza vaccination; and (4) general prenatal health, preventive strategies, and miscellaneous topics.For all study types, data extraction was standardized to include source, study design, number of participants in the intervention and control groups, intervention and control descriptions, duration, participant age and other details if available, attrition rates, and main outcomes.

Data Extraction
Two authors (TH and LY) simultaneously reviewed all abstracts for inclusion using Abstrackr and Rayyan, as described above.EndNote X7.2 (EndNote, Clarivate Analytics, Philadelphia, Pennsylvania, United States) was used to identify and remove duplicate records.Two searches were conducted; the initial search reviewed literature to 2016 and an updated search reviewed more recent literature until November 30, 2018.Once relevant abstracts were agreed upon, full-text analysis of included abstracts was then performed by the same authors.In addition, review of the bibliographies of included full-text articles were reviewed for additional eligible articles.Relevant articles meeting the final inclusion criteria were then abstracted in-depth for bias, study quality, and overall findings.

Assessment of Risk of Bias
Bias was evaluated by 2 independent reviewers (TH and LY).We applied specific tools for assessment of risk of bias tailored to each study type.For observational studies (not randomized controls), we used 1 of 2 National Institutes of Health (NIH) Quality Assessment Tools (NIH QAT), which consisted of 12 items to assist raters in formulating a holistic final quality assessment [28].If a study had a control but was not randomized, the Quality Assessment of Controlled Intervention Studies was used; if no control was available, the Quality Assessment Tool for Before-After (Pre-Post) Studies with No Control Group was used.For RCTs, we used the Cochrane Risk of Bias tool [29].Studies were rated independently by 2 reviewers (TH and LY).Disagreements were resolved by discussion between reviewers or with the involvement of a third author (PS) when necessary.

Data Synthesis
Data were collected to be primarily presented descriptively.We considered a meta-analysis or pooling of data if sufficient homogeneity in measured outcomes were to be observed, but the evaluation of data demonstrated heterogeneity that precluded such analyses.

Study Selection
An electronic search as described previously revealed a total of 2225 titles and abstracts after the removal of duplicates.After full-text evaluation, a total of 28 studies met the criteria for inclusion.An adapted PRISMA study flowchart is shown in Figure 1.

Study Characteristics and Synthesis of Results
Included studies fell into 4 categories: (1) gestational weight gain, obesity, and physical activity (n=9); (2) smoking cessation (n=9); (3) influenza vaccination (n=2); and (4) general prenatal health, preventive strategies, and miscellaneous topics (n=8).Reported sample sizes ranged from 16 to 5243 with a median of 91.All participants in the included studies were pregnant at the time of study initiation.Tables 1 and 2 outline studies focused on gestational weight gain, obesity, and physical activity [28,[30][31][32][33][34][35][36][37].Of the 9 eligible studies, 2 used exclusively text messages, 4 utilized text messages in conjunction with other technology, and 3 utilized mobile phone apps without text messages.One included study was not randomized, whereas the remainder were RCTs.Outcomes varied widely among studies.Two studies showed that intervention participants were significantly less likely to exceed healthy gestational weight gain during pregnancy (37% vs 66%; P=.03) [31] , and (58% vs 85%; P=.04) [36], but one found no such difference [33].Three interventions were associated with less overall gestational weight gain in intervention users over the study period [31,33,37].Notably, each of these interventions were multimodal and incorporated at least one additional communicative technology (Facebook or emails) alongside its main intervention (text messages or an app).The 2 studies that evaluated gestational weight gain and utilized interactive text messages or an app alone exhibited no difference in gestational weight gain compared with controls [34,36].Studies also differed regarding behavior change.Although some showed improvements in behavior analogs such as increased self-reported exercise [37] and less reduction in physical activity during pregnancy compared with prepregnancy [33], others showed no such relationship [32,34,35].There was no difference in incidence of gestational diabetes in any study [31,37].In terms of cost, 1 study did find that a mobile app compared with a parallel intervention requiring in-person counseling by health coaches was significantly less expensive (US $97 vs US $347) [36].In this study, both remote and in-person interventions were associated with lower proportion of excess gestational weight gain when compared with controls.[35] Until delivery 2 intervention groups: In person-dietary intake advice, exercise advice, paper weight graph and counseling provided by health coaches; Remote-same information as above provided in a mobile app format with electronic data capture; both compared with usual care from obstetrician 2 intervention groups: In person (18), Remote (19), and control (17) RCT Various clinics/United States/BMI=25-39.9;first trimester of pregnancy Redman (2017) [36] Until 34 weeks' gestation A mobile phone app with low glycemic index recipes, an exercise advice section, and a home page with tips and encouraging thought of the day.Also received emails every 2 weeks and two face-to-face hospital visits vs usual care Intervention vs usual care: 278 vs 287 RCT Maternity hospital/Dublin, Ireland/BMI=25-39.9; 10-15 weeks' gestation Kennelly (2018) [37] a RCT: randomized controlled trial.h ROBT: risk of bias tool.
Tables 3 and 4 outline interventions to address smoking cessation during pregnancy [38][39][40][41][42][43][44][45][46].Of the 9 studies, 7 used exclusively text messages, and the remaining 2 studies used mobile phone apps.Overall, outcomes were sparse regarding the ability of interventions to affect smoking cessation.Although 2 small uncontrolled studies showed a decrease in cigarettes smoked over the course of intervention [38] and more than 70% achievement of nonsmoking by the end of the intervention [39], the studies that employed control arms showed no difference in outcomes.These outcomes varied but included self-reported abstinence, biochemically reported abstinence, and number of smoke-free days [40][41][42][43][44][45][46].In 1 study, using text messages as the intervention mode was associated with increased self-efficacy, determination to quit smoking in pregnancy, and setting a quit date [42].
Tables 5 and 6 highlight the 2 studies of interventions to improve influenza vaccination rates [47,48].Both utilized text messages alone.There was no difference in influenza vaccination rates in intervention vs control groups in either study.
Tables 7 and 8 outline the remaining 8 studies, which focused on general prenatal health, preventive strategies, and miscellaneous topics [49][50][51][52][53][54][55][56].Four studies employed text messages alone, and 4 used mobile phone apps.In this sphere, interventions were most associated with improvements in health beliefs [49] and behaviors including self-reported attempts to eat more nutritious food [50], belief that taking prenatal vitamins will improve the health of the fetus [54], and belief that the participant is prepared to be a new mother [54].There was also a significant association between intervention use and attending a prenatal visit at least 6 months before delivery in 1 controlled study [51].In 1 study without formal controls, there was a higher rate of clinic attendance in intervention users (84%) compared with that for the general clinic population (50%) [52].In this study, attendance was even higher (89%) than in those who scheduled transportation through a free rideshare service facilitated through the app.Among these studies, there was no difference in any measured health outcomes including cesarean delivery and neonatal intensive care unit admission [51], hypertensive disorders of pregnancy, gestational weight gain, delivery outcomes [53], and beliefs and behaviors around smoking and alcohol [54,55].One unique study employed a mobile phone app to improve rates of perineal massage in Japan; this intervention was not associated with any difference in rates of practice of perineal massage, perineal lacerations, or episiotomy rates [56].

•
No difference in practice of perineal massage, perineal lacerations, or episiotomy rates ment by unblinded participants, unclear randomization

Risk of Bias in Included Studies
Bias ratings for all studies are included above in Tables 1-8.In total, 5 studies received a low risk rating, 11 studies received a fair risk rating, and 12 received a high risk rating.Reasons for the ratings were varied, and several studies had multiple reasons for increased risk of bias.Most commonly, studies with fair or high risk scores had issues with blinding (10 studies), high attrition (9 studies), or randomization (7 studies).Blinding issues most commonly revolved around patients and/or providers knowing a patient's allocation during the study.Randomization issues were varied and included unclear randomization schemes and lack of true randomization (being allocated by study staff).
Other less common issues included low sample size (5 studies) and high rates of participant-reported outcomes (5 studies).

Principal Findings
The findings from this systematic review suggest that available stand-alone mobile phone interventions show some positive changes in behavior and health outcomes in pregnant patients.
Although findings were limited and some studies had high risk of bias, these early data suggest such interventions may have some ability to improve behaviors and health outcomes.Further research in this area could take multiple forms.For example, medication adherence for patients with diabetes and hypertension could be explored; a study of adolescents and young adults with sickle cell disease found that of proposed mobile phone app features for improving adherence, daily medication reminders were ranked first most frequently; this sentiment may be shared by pregnant patients in the same age group [59].Further economic data may also be beneficial.One study of various mHealth interventions with reported economic evaluations found that 74.3% of interventions were cost-effective, economically beneficial, or cost saving [60].This was briefly noted in one of our reviewed studies [36], but additional data on the topic are necessary.In addition, future work may investigate cross-platform technologies, such as those that are both stand-alone mobile phone platforms and available via the Web.

Strengths and Limitations
The primary strength of this review is its inclusion of a wide variety of studies that investigated changes in behavior and clinical outcomes, both critical pieces necessary to evaluate novel behavioral technologies.Our study was focused on stand-alone interventions that did not require intensive clinical support and also reviewed data from high-income countries, which may provide more specifically applicable data for patients and physicians.
However, limitations include insufficiency of existing data and lack of granular clinical outcomes data in existing reports.Many included studies also exhibited high levels of bias with an unclear effect on results.In addition, no systematic evaluation of the interventions was performed (for example, using a specified taxonomy), which would allow more formal organization of intervention features themselves.Finally, intervention designs varied widely as did the measured outcomes and time frames of studies.All of these factors precluded the completion of a meaningful meta-analysis.

Comparison With Existing Literature
A systematic review recently published in April 2018 by Overdijkink et al [61] described a similar review of studies employing text messages and mobile phone apps in pregnancy.
Although there was an overlap in included studies, their methodology differed most notably because of the inclusion of telemedicine-based approaches.Most notably, at least five of their studies addressed gestational diabetes telemedicine and remote monitoring systems in which glucometers were coupled with mobile phone communications for nurse or physician feedback.In contrast, we aimed to find studies with minimal clinician input, preferring automated systems, and self-tracking technologies that supported or enhanced behavior changes without added clinician burdens.Furthermore, several of their included studies utilized primarily email and Web-based approaches, whereas we aimed to limit our review to mobile phone app and text-based technologies that could be implemented with use of phones or other primarily mobile technology.Despite these differences, we identified similarly that results are heterogeneous and that additional research is required to evaluate the effects of mHealth interventions on long-standing positive health outcomes.

Conclusions
In

Figure 1 .
Figure 1.Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow sheet.
The assessed interventions were stand-alone mobile phone-based interventions including, but not limited to, mobile phone apps, text messaging, games, and information services.We excluded studies that used technology interventions aimed solely at communication between patients and clinicians without a stand-alone educational, motivational, or interactive component (such as telemedicine portals or electronic medical record-based portals for use with mobile phones), and interventions that were not primarily intended for mobile phone use, eg, websites.Studies were excluded if they focused solely on neonatal health, such as neonatal feeding support interventions or growth tracking tools.Studies were also excluded if they were exclusively published as abstracts or conference proceedings without a full peer-reviewed manuscript.Finally, studies were excluded if they were solely meant to evaluate feasibility or desirability of hypothetical interventions or supplied outcomes with fewer than 2 weeks of intervention use.We searched MEDLINE via PubMed, EMBASE, Web of Science, Cochrane Database of Controlled Trials, CINAHL, and PsycINFO databases from January 1, 2000 to November 30, 2018.We began with the MEDLINE search and translated to the appropriate syntax for each of the other databases, using controlled vocabulary when possible.Search terms related to pregnancy, mobile interventions, and select behaviors (including smoking cessation, weight loss, and diabetes management) were included.Full search strategies can be found in Multimedia Appendix 1, and a completed PRISMA checklist is found in Multimedia Appendix 2.

Table 1 .
Design of trials with a focus on gestational weight gain, obesity, and physical activity.

Table 2 .
Outcomes and bias of trials with a focus on gestational weight gain, obesity, and physical activity.
a IG: intervention group.b CG: control group.c GWG: gestational weight gain.d IOM: Institute of Medicine.e NIH QAT: National Institutes of Health Quality Assessment Tool.f CES-D: Center for Epidemiologic Studies Depression Scale.g N/A: not applicable.

Table 3 .
Design of trials with a focus on smoking cessation.

Table 4 .
Outcomes and bias of trials with a focus on smoking cessation.
b IG: intervention group.cCG: control group.dPPA: point prevalence abstinence.e ROBT: risk of bias tool.f N/A: Not applicable.g eCO: exhaled carbon monoxide.

Table 5 .
Design of trials with a focus on influenza vaccination.

Table 6 .
Outcomes and bias of trials with a focus on influenza vaccination.

Table 7 .
Design of trials with a focus on general health, preventive health, health beliefs, and other topics.

Table 8 .
Outcomes and bias of trials with a focus on general health, preventive health, health beliefs, and other topics.
high-income countries, utilization of mobile phone-based health behavior interventions in pregnancy demonstrates some correlation with positive beliefs, behaviors, and health outcomes.More effective interventions are multimodal in terms of features and tend to focus on healthy gestational weight gain.As mHealth interventions become increasingly available, future work must aim to maximize the clinical effectiveness of such interventions.As researchers, we should aim to broaden the scope of effective and sustainable interventions and continue to augment our care with appropriate evidence-based technologies.