The estimated global premature birth rate per 100 live births increased slightly from 9.8% (13.8 million premature births) in 2010 to 9.9% (13.4 million premature births) in 2020 [1]. This indicates an average annual increase rate in premature birth prevalence of 0.14% [1]. Complications of premature birth are the leading cause of childhood mortality, which refers to the probability of dying between birth and exactly 5 years of age, expressed per 1000 live births. This encompasses 35% of neonatal deaths and 18% of deaths in children aged <5 years [2]. Compared with their term-born peers, premature babies are more likely to develop respiratory distress syndrome, sleep apnea, necrotizing enterocolitis, and intraventricular hemorrhage in the neonatal period [3] and have worse cognitive, language, and motor development and social-emotional competence as children aged <5 years [4,5]. In addition, young adults born preterm are more susceptible to psychological fragility (in terms of anxiety and insecurity) and tend to exhibit lower cognitive ability [6]. Therefore, reducing the incidence of premature births is of utmost importance for the short- and long-term health and development of children.

Sociodemographic, nutritional, medical, obstetric, and environmental factors can increase the risk of premature birth. Consequently, interventions have been developed to prevent or mitigate known modifiable risk factors for premature birth [7]. Primary and secondary prevention encompasses a wide range of interventions, including medication, surgical procedures, cervical devices, targeted diets, physical exercise, smoking cessation programs, nutritional supplementation, education, and various special tests or investigations [7,8]. For women to participate in preventive activities for primary health promotion; secondary activities; and tertiary activities, including lifestyle modification, health screening uptake, treatment compliance, and participation in rehabilitation programs, understanding and continuous motivation are required. Recently, internet-based interventions have been developed to provide health-related information to individuals who may not have direct access to medical facilities [9], offering accessibility and availability regardless of time and location [10]. In addition, internet-based interventions can enhance engagement with self-monitoring, promote health-related understanding, and increase knowledge and risk perception of disease as well as bolster self-efficacy in disease management [11].

Previous systematic reviews have thoroughly examined interventions aimed at preventing premature birth, evaluating the overall evidence for such interventions [7,8,12,13]. These reviews covered a range of specific interventions, including infection treatment [14], pharmacological methods [15-17], nutrient supplementation [18], cerclage [19], cervical devices [20], and social support [21]. However, to date, only 2 systematic reviews have explored internet-based interventions in this context, confirming the effectiveness of technology-supported lifestyle interventions [22] and telemedicine [23] specifically for pregnant women with gestational diabetes mellitus (GDM). Although the effectiveness of internet-based interventions for smoking cessation among pregnant women [24] and prenatal interventions for maternal health [25] has been established, their effectiveness in preventing premature birth remains unconfirmed.

To address this gap, we conducted a systematic review of randomized controlled trials (RCTs) and quasi-experimental studies focusing on internet-based interventions for premature birth prevention. The objectives of this review were three-fold: (1) to describe the general characteristics of the studies included, (2) to identify the study designs of internet-based interventions pertinent to premature birth prevention, and (3) to evaluate the effectiveness of internet-based interventions in achieving outcomes related to premature birth prevention among the target population.

This systematic review was conducted in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [26] and registered in PROSPERO (CRD42021278847). Our review focused on 2 specific research questions formulated using the Population, Intervention, Comparison, and Outcome strategy: (1) What is the efficacy of internet-based interventions in reducing the risk of premature birth among pregnant women compared to standard prenatal care? (2) How do internet-based interventions impact maternal health outcomes, such as stress, anxiety, and gestational diabetes, in pregnant women at risk of premature birth? These questions aimed to evaluate the effectiveness of internet-based interventions in both reducing premature births and improving crucial maternal health outcomes. Our systematic search targeted several electronic databases, including MEDLINE, Embase, CINAHL, and the Cochrane Library, focusing on studies published up to February 2023. To augment our database search, we manually reviewed the reference lists of the included publications.

Our inclusion criteria encompassed published RCTs, quasi-experimental studies, and experimental studies on the prevention of premature birth. We imposed no restrictions regarding the country or language of publication. The target population included all pregnant women, including those with normal or high-risk pregnancies, pregnancy complications, or at risk of premature birth. The interventions were internet-based and used various devices, such as computers and mobile phones. We excluded cross-sectional, case-control, retrospective, and prospective cohort studies; noncomparator experimental studies; animal experiments; reviews; qualitative studies; case reports; unpublished data; and gray literature, such as conference abstracts, letters, editorials, dissertations, and unavailable full texts. Studies targeting prepregnant women, women in the postpartum period, women outside childbearing age, and men were also excluded.

Adapted search terms for each database included a combination of terms related to population (eg, “women”), pregnancy (eg, “premature birth” and “pregnancy”), information and communication technology (eg, “computer”), treatment (eg, “internet” and “online”), and study design (eg, “randomized controlled trial”). These terms were used to search titles, abstracts, keywords, or text words. The exact search terms are detailed in Multimedia Appendix 1.

All identified studies were first imported into a reference manager for deduplication. The titles and abstracts were then independently screened by 2 of the 3 reviewers, working in rotating pairs (ie, A and B, B and C, and A and C). Following this initial screening, relevant studies underwent a full-text review. Disagreements at this stage were resolved through discussion or consultation with the third reviewer, ensuring a consensus on the inclusion of studies. Any studies found to be irrelevant after full-text review were excluded from further consideration. Simultaneously, a data extraction form was collaboratively developed and pretested by the reviewers to systematically collect review characteristics and outcome data from the selected studies. This process of data extraction was also conducted independently by 2 pairs of reviewers. In cases of discrepancies in the extracted data, the reviewers engaged in discussions to reach a consensus or consulted the third reviewer for an objective resolution.

The extracted data included study characteristics (eg, authors, year, country of origin, research design, and sample size), study results (primary and secondary findings for outcome measures, including effect sizes), and intervention details (eg, name, method, timing, duration, and group type). Because of the variation in methodologies across studies, conducting a meta-analysis was considered unsuitable. Instead, information was synthesized narratively, categorizing outcomes into psychosocial, physiological, and wellness health outcomes. Effect sizes were calculated using means and SDs or frequencies and percentages depending on the study design.

Two pairs of reviewers independently assessed the methodological quality using the revised Cochrane RoB tool for randomized trials [27]. This tool evaluates 5 domains: randomization process, deviation from intended interventions, missing outcome data, outcome measurement, and reported result selection grouped into 3 levels of RoB (low risk, some concern, and high risk). Studies were categorized into 2 groups: intention-to-treat (ITT) and per-protocol (PP), with disagreements resolved through discussion or consultation with a third person.

Owing to the heterogeneity in interventions and participant characteristics, we opted for a narrative synthesis instead of a meta-analysis. When available, effect sizes were calculated using data from the studies, using various metrics such as Cohen d, Cohen f, Morris d, Hedges g, Cohen h, odds ratio (P value and 95% CI), and relative risk (95% CI and P value) [28]. Of the 26 papers reviewed, 5 (19%) lacked sufficient statistical data to calculate the effect size of the intervention. We reached out to the authors of these papers for additional information. However, responses were not received for several of these inquiries, limiting our ability to calculate effect sizes for all studies. Consequently, effect sizes were calculated for 21 articles. In instances where additional data from the original authors were not obtained, our evaluations were based on the information available in the study. In cases where studies presented results solely in graphical form, we calculated effect sizes where possible, specifically if the graph provided measurable mean and SD. However, for graphics lacking detailed data, such as missing SDs, effect size calculation was not feasible.

Initially, a total of 2748 articles were retrieved from the 4 databases. After excluding 761 duplicate articles, 1987 remained. During the initial screening stage, 1959 papers were excluded after reviewing the study title and abstract. The full texts of the remaining 28 studies were reviewed, and 2 studies that were not controlled comparative experimental studies—they were 1-group pre- and postintervention comparison studies—were excluded. Finally, 26 studies were selected for the systematic review. Figure 1 illustrates the study selection process.

In the subset of studies that used ITT analysis, the overall RoB was categorized as low for 17% (2/12) of the studies, exhibiting some concerns in 25% (3/12) of the studies, and high in 58% (7/12) of the studies. Among the studies using PP analysis, only 7% (1/14) of the studies were assessed as having a low RoB, whereas 57% (8/14) of the studies had some concerns, and 36% (5/14) of the studies were deemed to have a high risk. The detailed outcomes of the RoB assessment for the 12 ITT and 14 PP studies are shown in Figure 2.

The 26 studies included in this review were conducted across various countries, with the largest number from the United States (6/26, 23% studies [29-34]), followed by China (4/26, 15% studies [35-38]), Taiwan (3/26, 11% studies [39-41]), Switzerland (2/26, 8% studies [42,43]), and the United Kingdom (2/26, 8% studies [44,45]). In addition, 1 study each was conducted in Australia [46], Canada [47], Iran [48], Ireland [49], the Netherlands [50], Norway [51], Spain [52], Sweden [53], and Thailand [54]. The publication years spanned from 1996 to 2022, with most studies (14/26, 54%) [32-34,40-46,49,51-53] published between 2010 and 2019, 42% (11/26) of the studies [29,31,35-39,47,48,50,54] published from 2020 to the time of the review, and 4% (1/26) of the studies [30] published before 2000. Among these, one study was published in Chinese [35] and another in Persian [48], with the remaining studies all written in English.

With regard to research design, 81% (21/26) of the studies used a randomized controlled experimental design [30-39,42-46,49-54], whereas 19% (5/26) of the studies used a quasi-experimental design [29,40,41,47,48]. The focus of these studies varied, with 8% (2/26) of the studies targeting pregnant women diagnosed with anxiety or depression [46,53], 27% (7/26) focusing on pregnant women with GDM or at risk of GDM [29,35,37-39,47,52], and 8% (2/26) involving women experiencing preterm labor [42,43]. Of the 26 studies, 1 (4%) was dedicated to pregnant smokers [44], 2 (8%) were dedicated to pregnant women using drugs or alcohol [30,34], and 1 (4%) was dedicated to those experiencing insomnia [31]. The remaining 42% (11/26) of the studies targeted nonspecific pregnant populations [32,33,36,40,41,45,48-51,54]. Detailed information on these studies is provided in Tables 1 and 2.

The outcomes of the interventions, as detailed in Multimedia Appendix 2 [29-54], were divided into 3 main categories: psychosocial, physiological, and wellness health outcomes. Each category encompasses several subcategories, as shown in Figure 3.

This systematic review presents a comprehensive evaluation of web-based intervention studies focused on the prevention of premature birth in pregnant women. Various web-based interventions and diverse groups of pregnant women were included in the analysis; however, a significant gap was noted in studies that directly confirmed the effects of these interventions on premature birth. The measurement variables used to ascertain the direct effects of the interventions varied, and there were few well-designed interventional studies. These findings echo those of a meta-analysis of web-based educational interventions in 2022 [55]. Among the analyzed studies, only 12% (3/26) of the studies had an overall low RoB. Many of the included studies had moderate to high risk, primarily owing to nonblinding, adherence issues, and selection bias. This highlights the need for more high-quality intervention studies in this field. Blinding in internet-based intervention studies poses a challenge because of the active participation requirement and common attrition. Consequently, RCTs may not always unfold as intended. Thus, statistical analysis requires adjustment for participant bias, and participant characteristics should be carefully considered during interpretation.

This study found that most web-based interventions primarily used websites, followed by mobile instant messaging, aligning with common methods used in web-based health education interventions [56]. Only 1 intervention in this review used a mobile app [31], although it predominantly relied on mobile instant messengers for information delivery and encouraging participation. This approach is prevalent in countries such as China, Taiwan, Thailand, and Iran [35-39,48,54]. Recent studies, including 1 study using WhatsApp with a chatbot for health promotion messages, have shown higher uptake rates for interventions [57]. Mobile apps, being more accessible than websites, have shown greater effectiveness for glycemic control [11]. Further studies should thus compare the efficacy of websites and mobile apps and explore the use of mobile instant messengers and apps in more depth.

Premature births have many unexplained causes, with preeclampsia, eclampsia, intrauterine growth restriction, spontaneous preterm labor, and preterm premature rupture of membranes being the common causes [58]. Nonspecific pregnant women were the most common target group in the reviewed web-based interventions, followed by pregnant women with GDM, preterm labor, substance use, depression, and anxiety. However, experimental research on women with specific health issues is lacking. Although 7 experimental studies focused on pregnant women with GDM, highlighting the importance of blood glucose (BG) self-management, none targeted women with hypertensive disorders, which is crucial for early detection and management. Furthermore, despite spontaneous premature labor and preterm premature rupture of membranes being leading causes of premature birth [58], only 2 experimental studies [42,43] focused on these conditions. Hence, there is a critical need for further experimental studies targeting pregnant women with specific health conditions.

In most of the reviewed studies, interventions spanned the entire duration of pregnancy, likely because of the increasing risk of premature birth as pregnancy progresses. These web-based interventions typically lasted for 4 to 8 weeks, with a few studies using longer durations. This contrasts with web-based interventions for nonpregnant adults, where longer durations are more common [59]. Given the evolving physical conditions during pregnancy, there is a growing need for long-term interventions that cover the entire pregnancy period, aligning with the emphasis on a holistic approach to pregnancy health care [60]. This approach encompasses not only the pregnancy period itself but also the early and prepregnancy stages as well as the postpartum period. Web-based interventions offer the advantage of accessibility and flexibility, making them suitable for long-term implementation compared to in-person interventions. However, the conclusive impact of the intervention duration remains to be determined, indicating the need for further research.

Our research found that the interventions had varying effects on stress or distress among different groups of pregnant women. Studies involving nonspecific pregnant women [40,45,48,49] and those focusing on women diagnosed with anxiety or depression [46] reported significant reductions in stress or distress following the interventions. This success can partly be attributed to the interventions being tailored to the specific needs of these groups. For instance, the study by Loughnan et al [46], which applied self-guided cognitive behavioral therapy, was specifically designed to target anxiety and depression, closely aligning with participants’ conditions. In contrast, interventions aimed at pregnant women with preterm labor did not achieve a similar reduction in stress or distress [42,43]. This disparity may be owing to the interventions in these studies, which were led by Scherer et al [43] and Urech et al [42], not being adequately customized to meet the unique needs of patients experiencing premature labor. Given that the condition of premature birth was not directly addressed [42], it is possible that the programs were less effective for these participants. In addition, as gestational age increased and fetal maturation progressed, the health risks for the newborn decreased [42], which might have influenced the perceived stress levels and the efficacy of the interventions.

The studies included in this research used various interventions, such as cognitive behavioral management [42,43,45,46,49], unified transdiagnostic treatment [48], and antenatal care system [40]. Each of these approaches has its own theoretical basis and methodological implications, which could affect the outcomes. Furthermore, the studies used different stress assessment tools, including the Perceived Stress Scale, Pregnancy Stress Rating Scale-36, and Prenatal Distress Scale. It is important to note that these are self-report screening questionnaires and not diagnostic assessment tools evaluated by trained professionals. This raises questions about the generalizability and applicability of the findings to individuals clinically diagnosed with stress or distress.

The interventions showed significant effects on anxiety in studies targeting women with anxiety [46], GDM [37], and insomnia [31]. However, no significant effects were noted in studies targeting women with preterm labor [42,43]. This aligns with the earlier findings regarding stress. The diversity of the interventions and measurement tools used across the studies, particularly as only 2 studies used the same tools [45,53], indicates the need for further research to reliably estimate these effects on anxiety. In addition, concerning depression, the cognitive behavioral interventions did not significantly impact pregnant women with anxiety or depression [46] or those with preterm labor [42] assessed by the Edinburgh Postnatal Depression Scale (EPDS). Similarly, women with depression [53] showed no significant changes when evaluated using the EPDS, though different results emerged from the Montgomery-Åsberg Depression Rating Scale. Furthermore, 2 studies targeting depression in general pregnant populations [50,51,53] used the EPDS but reported no significant effects [50,51]. The EPDS, although effective in detecting postpartum depression, might be less sensitive in identifying severe or prenatal depression [53]. Notably, in women with insomnia [31], depression significantly decreased when assessed with the EPDS and treated with cognitive behavioral methods. This raises questions about the validity of the EPDS during pregnancy, especially for those at a high risk for depression. Furthermore, most internet-based interventions for depression in pregnant women did not yield effective outcomes as measured by the EPDS. Ashford et al [61], however, claimed its effectiveness in perinatal depression. This discrepancy could be owing to differing methodological qualities; the studies in this review had a high overall RoB, whereas Ashford et al [61] included studies with average to high methodological quality.

This review underscores a notable inconsistency in the types of interventions, target audiences, and assessment tools used across the studies examined. Dennis [62] emphasized the importance of preventive interventions for mental health issues in pregnant women, particularly advocating for targeting women with identified risk factors [63]. In our analysis, the limited effects of interventions aimed at reducing anxiety or depression among nonspecific pregnant women [50,51] could be attributed to the absence of specific risk factors for these conditions. This observation suggests that the effectiveness of preventive psychosocial health interventions may be enhanced by tailoring them to the needs of women with identifiable risk factors.

Most studies in our review focused on verifying the effectiveness of BG control in women with GDM. The research demonstrated significant effects on both fasting BG and 2-hour BG levels in several studies [11,22,23,64], aligning with the findings from previous meta-analyses. However, variations were observed, such as in 1 study using a telemedicine system where HbA_1c levels were significantly lower in the experimental group [65]. Conversely, a study that provided educational information on GDM did not show a significant impact. These discrepancies might be because of factors such as the duration of the interventions, sensitivity of the indicators used, and the limited number of trials conducted. The physiological changes during pregnancy, such as iron deficiency and reduced life span of red blood cells, can affect the sensitivity of HbA_1c assays [65]. In contrast, glycated albumin, with a shorter half-life, may offer a more accurate measure of short-term glucose fluctuations. However, research on glycated albumin in pregnant women is still limited [11], suggesting an area that warrants further investigation in the future.

In terms of preventing premature birth, interventions targeting women with GDM or those at risk for GDM did not demonstrate a significant effect on this outcome [35,37,39,42,47,50,52,54]. These studies primarily assessed the role of internet-based intervention in addressing the causes of premature birth, analogous to managing diabetes for blood sugar control. The complexity of factors influencing premature birth makes it challenging to ascertain the direct preventive effects of these interventions. Hence, premature birth was not the primary outcome in any of the reviewed studies. Given the insufficient number of studies specifically examining each intervention type, it becomes necessary to continue evaluating premature birth as a secondary outcome in future research. This approach will further our understanding of how interventions can mitigate or eliminate the causes of premature birth, thereby enhancing maternal and neonatal outcomes.

This systematic review contributes to the literature by methodically analyzing internet-based interventions aimed at preventing premature birth. Its main strength lies in its extensive focus on a broad spectrum of outcomes, covering a wide range of maternal and neonatal health aspects. Another key strength is the inclusivity of the review, encompassing a diverse range of populations. The absence of geographical or language restrictions in the selection criteria enhances the comprehensiveness and global applicability of the review. This broad and inclusive approach not only illuminates the potential of internet-based interventions in the pregnancy context but also delineates their limitations, thereby laying a foundation for future research endeavors and informing clinical practice with a more global perspective.

However, the review has several limitations that need to be considered. Its reliance on studies from only 4 databases may have missed relevant research from other sources, potentially limiting the scope of the findings. The exclusion of specific types of publications, such as conference abstracts and dissertations, coupled with limited responses from original authors for additional data, could have introduced publication bias. In addition, 2 studies translated from non-English language using Google Translator may have inaccuracies in translation, affecting the interpretation of these studies. Although backward and forward citation tracking of the final included articles was initially planned, it was not conducted because of resource constraints and the comprehensive nature of the initial search. The underrepresentation of studies with a low RoB in this review suggests the need for caution in generalizing the results. Moreover, the focus on studies predominantly from countries with high internet use may limit the applicability of the findings to regions with different internet access levels and use patterns. Therefore, the results should be interpreted with an understanding of these contextual differences.

This systematic review uncovered a wide array of internet-based interventions that target risk factors associated with premature birth, with notable efficacy in the realms of psychosocial health and blood sugar management. However, interventions addressing other risk factors have shown a considerable diversity in measurement methods and a range of experimental effects. This variability points to an ongoing challenge in accumulating robust evidence. These findings underscore the critical necessity for future clinical trials to not only develop and test but also widely disseminate internet-based interventions that are both safe and effective. In addition, there is a pressing need for the creation of standardized measurement tools. Rigorous evaluation processes should be established to enhance the effectiveness and reliability of these interventions in clinical settings. Such efforts are essential for ensuring that these digital health solutions can effectively contribute to the prevention of premature births and improve maternal health outcomes on a broader scale.