This study was conducted as a systematic review and meta-analysis to examine the contents and delivery methods of video-based prenatal interventions and to evaluate their effects on health-related outcomes among pregnant women. The research question guiding this review was: “What is the effect of video-based interventions on health-related outcomes among pregnant women?” The study was conducted in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses) reporting guideline (Page et al., 2021).

From March 2 to 8, 2025, a comprehensive literature search was conducted across 5 electronic databases, including core databases (Higgins et al., 2022): PubMed, the Cochrane Library, Embase, the Cumulative Index to Nursing and Allied Health Literature (CINAHL) Complete, and the Research In formation Sharing System (RISS). The search strategy employed advanced search techniques, including MeSH (medical subject headings), Emtree terms (Elsevier's controlled vocabulary), natural language keywords, and relevant synonyms. (Supplement Material 1). All records were managed manually using database advanced search functions; reference management software (e.g., EndNote) was not used. In addition, hand searching was conducted using Google Scholar and by screening reference lists of eligible studies, resulting in the identification of 11 additional records (Supplement Material 2).

The inclusion criteria were defined as follows: (1) fully accessible articles published in English or Korean; (2) studies

published in peer-reviewed journals; (3) studies evaluating video-based interventions targeting health-related outcomes among pregnant women; and (4) experimental study designs. The exclusion criteria were: (1) protocol studies, theses, degree dissertations, preprints, conference presentations, reports, magazine articles, books, or letters; (2) studies that did not report outcome data; and (3) interventions conducted exclusively during the postpartum period.

The search strategy was structured according to the participant, intervention, comparison, outcome, setting, time, and study design (PICOST-SD) framework (Bidwell & Jensen, 2003). The criteria were defined as follows:

The study selection and data extraction procedures were demonstrated and implemented in a structured graduate-level classroom setting to enhance transparency and methodological rigor. The corresponding author presented each step of the screening, eligibility assessment, and data extraction process to 12 master's-level graduate nursing students during a research methodology course. Following this demonstration, the author conducted the study selection and data extraction, while the students independently applied the same procedures using their own approaches for comparison purposes. Data extracted included the general characteristics of the studies, intervention characteristics, and intervention effects. After independent completion, the extracted results were compared. When discrepancies or uncertainties were identified, they were addressed through a structured process of question-and-answer clarification, discussion, and revision. Final decisions regarding study inclusion and extracted data were confirmed by the corresponding author based on predefined criteria.

Data were extracted using Microsoft Excel, including general characteristics of the studies (first author, publication year, country), intervention characteristics (intervention content or theme, comparison content, study design, video duration and number of sessions, and sample size of experimental and control groups), and intervention effects (outcomes, measurement scales, statistical effects between experimental and control groups [p-values], and conclusions).

The methodological quality of the 12 included studies was assessed using standardized risk-of-bias tools (Sterne et al., 2019): Risk of Bias 2.0 (ROB 2) for 9 RCTs and Risk of Bias in Non-randomized Studies of Interventions (ROBINS-I) for 3 non-RCTs. The ROB 2 tool evaluates bias arising from the randomization process, deviations from intended interventions, missing outcome data, outcome measurement, and selective reporting. The ROBINS-I tool assesses similar domains, with additional consideration of confounding and participant selection. For both tools, judgments were made at the domain level and subsequently synthesized into an overall risk-of-bias judgment, classified as “low risk,” “some concerns,” or “high risk” for ROB 2 and as “low,” “moderate,” “serious,” or “critical” risk for ROBINS-I. Risk-of-bias assessments were visualized usingfu summary plots and traffic-light charts gene rated with the robvis (Risk-of-bias VISualization) tool (McGuinness & Higgins, 2020).

Statistical analyses were conducted using the meta-analysis module in jamovi (version 2.3.28; The jamovi project, Australia). The standardized mean difference (SMD) was used as the primary effect size measure. A random-effects model was applied to account for between-study variability. Between-study heterogeneity (τ²) was estimated using the maximum-likelihood estimator and further assessed using Cochran Q-test, with heterogeneity quantified using the I² statistic. When heterogeneity was detected (τ²> 0, regardless of the Q-test result), a prediction interval was calculated to estimate the range in which the true effects of future studies are expected to lie. Influential studies were identified using Cook's distance, with values exceeding the median plus 6 times the interquartile range considered influential. Publication bias was evaluated through visual inspection of funnel plots. For outcomes that could not be pooled due to heterogeneity in outcome measures, study designs, or insufficient statistical data, a narrative synthesis was conducted. These findings were summarized descriptively in the “Effects of video-based interventions” section using reported effect estimates, including percentages, means and standard deviations, and p-values, as appropriate.

Between March 13 and 21, 2025, a total of 325 records were identified through database searches, including PubMed (n=144), the Cochrane Library (n=1), Embase (n=148), CINAHL (n=5), and RISS (n=27). After removal of 43 duplicate records, the titles and abstracts of 282 articles were screened for eligibility based on the predefined PICOST-SD criteria. During this initial screening, 260 articles were excluded for failing to meet the inclusion criteria. Full texts of the remaining 22 articles were retrieved and assessed for eligibility. All 33 full-text articles were reviewed in detail. Of these, 17 were excluded due to nonexperimental study designs, and 4 were excluded because the interventions were not video-based. Ultimately, 12 studies met all inclusion criteria and were included in the systematic review and meta-analysis (Fig. 1, Supplement Material 2).

Overall, based on the ROB 2 assessment, 2 RCTs were judged to have a low risk of bias, 3 showed some concerns, and 4 were rated as having a high risk of bias. Using the ROBINS-I tool, 2 non-RCTs were judged to have a serious risk of bias, and one was judged to have a critical risk of bias (Supplementary Fig. 1).

Twelve articles published between 2001 and 2024 (Adam et al., 2021; Björklund et al., 2012; Chaudhary et al., 2023; Hewison et al., 2001; Hughes et al., 2017; Kellams et al., 2018; Metin & Baltacı, 2024; O'Sullivan et al., 2019; Parobek et al., 2022; Potharst et al., 2022; Stortz et al., 2023; Tsai et al., 2018) were included in this systematic review. Four studies were conducted in the United States of America (Hughes et al., 2017; Kellams et al., 2018; Parobek et al., 2022; Stortz et al., 2023). Of the included studies, 9 were RCTs (Adam et al., 2021; Björklund et al., 2012; Chaudhary et al., 2023; Hewison et al., 2001; Hughes et al., 2017; Kellams et al., 2018; Metin & Baltacı, 2024; Parobek et al., 2022; Stortz et al., 2023), and 3 were non-RCTs (O’ Sullivan et al., 2019; Potharst et al., 2022; Tsai et al., 2018). The number of participants per group ranged from 17 to 1,007 (Table 1).

Analysis of the 12 included studies showed that prenatal video-based interventions addressed a broad range of maternal health topics. Several studies focused on genetic screening and testing, including Down syndrome screening (n=2) (Björklund et al., 2012; Hewison et al., 2001), prevention of cytomegalovirus (CMV) infection (n=1) (Hughes et al., 2017), and decision-making related to genetic testing and data sharing (n=2) (Parobek et al., 2022; Stortz et al., 2023). A second major thematic area involved breastfeeding and lactation support, with interventions aimed at promoting breastfeeding, preventing early cessation (n=3) (Adam et al., 2021; Kellams et al., 2018; Metin & Baltacı, 2024), and teaching colostrum extraction techniques (n=1) (O'Sullivan et al., 2019). Another group of studies focused on maternal self-care knowledge and practices during pregnancy and the early postnatal period, including web-based antenatal education (Tsai et al., 2018) and video-based interventions designed to improve maternal postnatal care practices (Chaudhary et al., 2023). Finally, some interventions targeted maternal and infant well-being by strengthening maternal-child bonding and providing emotional support (Potharst et al., 2022), as well as by reducing stress, enhancing self-efficacy, and improving satisfaction during pregnancy (Tsai et al., 2018) (Table 1).

The duration of video-based interventions ranged from 2 minutes to 90 minutes, with delivery spanning one to 13 sessions. Intervention platforms included clinic-based viewing, online websites, mobile applications, and social media platforms. Several interventions were delivered during early pregnancy (n=5) (Björklund et al., 2012; Hewison et al., 2001; Hughes et al., 2017; Parobek et al., 2022; Stortz et al., 2023), whereas others were implemented across all gestational stages and focused on breastfeeding, maternal-infant bonding, or postnatal care (n=7) (Adam et al., 2021; Chaudhary et al., 2023; Kellams et al., 2018; Metin & Baltacı, 2024; O'Sullivan et al., 2019; Potharst et al., 2022; Tsai et al., 2018). Comparison groups received usual care or standard educational information only, such as counseling (n=1) (Hewison et al., 2001), infographics (n=1) (Parobek et al., 2022), brochures (n=1) (Hughes et al., 2017), without any video-based components (Table 1).

Overall, video-based interventions were associated with improvements in maternal knowledge across multiple domains, including Down syndrome screening, breastfeeding, colostrum extraction, genetic testing, and postnatal care (Björklund et al., 2012; Chaudhary et al., 2023; Hewison et al., 2001; Metin & Baltacı, 2024; O'Sullivan et al., 2019; Parobek et al., 2022; Stortz et al., 2023). These interventions also enhanced self-efficacy and confidence, particularly with respect to antenatal care, colostrum expression, and breastfeeding (Metin & Baltacı, 2024; O'Sullivan et al., 2019; Tsai et al., 2018). Positive psychological effects were reported, including reductions in stress, depression, worry, decisional conflict, and regret (Potharst et al., 2022; Stortz et al., 2023; Tsai et al., 2018). In addition, some interventions supported behavioral change, such as adoption of safer practices to prevent CMV infection (Hughes et al., 2017) and more informed decision-making regarding genetic data sharing (Parobek et al., 2022). In contrast, no significant effects were observed for breastfeeding duration (Kellams et al., 2018), anxiety (Hewison et al., 2001), or screening uptake (Björklund et al., 2012). Across all studies, knowledge (n=6) and self-efficacy (n=2) were the most frequently reported outcomes, followed by worry (n=2) and screening uptake (n=2). Other outcomes—including satisfaction (p<0.001), confidence (p<0.001), decisional conflict (p<0.001), regret (p<0.001), depression (p=0.017), bonding (p=0.046), risk-reduction behavior (p=0.020), anxiety (p=0.910), attitude (p=0.275), and feeding cessation (not effective)—were each reported in a single study (Table 2).

Among the 6 studies that assessed knowledge as an outcome, one employed a single-arm design (O'Sullivan et al., 2019) and 2 did not report standard deviations (Chaudhary et al., 2023; Stortz et al., 2023); therefore, 3 RCTs (Adam et al., 2021; Björklund et al., 2012; Tsai et al., 2018) were eligible for inclusion in the meta-analysis. These 3 studies included a total of 1,600 participants in the intervention group and 1,714 participants in the control group. The random-effects model yielded an average SMD of 0.22 (95% confidence interval, 0.15-0.29), indicating a small but statistically significant positive effect. This pooled effect differed significantly from zero (z=6.43, p<0.001), indicating a small but statistically significant positive effect of video-based interventions on knowledge outcomes. Although the pooled effect was statistically significant, the wide prediction interval (-0.71 to 1.15) suggests considerable uncertainty in the effect size that may be observed in future studies. The Q-test indicated no significant heterogeneity among studies (Q²=1.63, p=0.44), with τ²=0.00 and I²=0%. Studentized residuals were all within ±2.39, suggesting no outlying studies. Visual inspection of the funnel plot did not suggest substantial publication bias (Fig. 2, Supplementary Fig. 2).