Our database search results yielded 3346 records from PubMed (1676), Scopus (1676), ScienceDirect (1050), and Cohrain (522). After the process of removing duplicate articles, 3201 records remained. Applying the inclusion and exclusion criteria reduced this number to 92. The researchers then excluded articles for the following reasons: (1) having different outcomes (n = 28); (2) different study designs (24); (3) different populations and no explanation of research criteria (n = 15); (4) inappropriate group comparisons (n = 2); and (5) untranslatable language (n = 3). Ultimately, 14 articles were included in the systematic review (Fig. 1).
Flow of studies found by the systematic review
Study characteristics
Description of intervention methods
The intervention approaches reviewed consisted of six types: CBT, mindfulness, counseling, and Motivational Interviewing (MI) [41]. CBT interventions were used in four studies. Two articles used an online platform as the digital innovation, and the other two used individual meeting sessions [42,43,44,45]. All of these studies involved tasks for pregnant women to complete at home (Tables 1 and 2).
The operational definition of study characteristics that we included consisted of the author being the name of the principal investigator or study author responsible for the study. Study location is the location of the study, including country and institution, if relevant. Subject (Gestational Age) and duration of Follow-Up is the gestational age at which participants were recruited for the study and the duration of observation or follow-up after the intervention. Number of Sessions (Hours) is the number of intervention sessions conducted in the study, measured in time units such as minutes or hours. A participant with Anxiety Risk is whether the participant has a risk of anxiety (yes/no); based on the criteria or diagnosis used in the study, anxiety risk was determined based on the presence of prior obstetric complications, as reported in the included studies. These complications included factors such as a history of preterm birth, chromosomal abnormalities, or other significant obstetric events, which have been identified in the literature as contributors to heightened anxiety during pregnancy.
Primary Intervention Approach (Name) Delivery Format is the primary method of intervention used, such as Cognitive Behavioral Therapy (CBT), Mindfulness, or Motivational Interviewing, and the delivery format, such as face-to-face sessions, online or via an app. Collaboration is the involvement of collaborators in the study, including the types of professionals participating (e.g., psychologists, midwives, nurses) and their roles in supporting the implementation of the intervention.
Three studies had a population of pregnant women with anxiety risk factors. The population in the study [42] consisted of pregnant women who had positive results on chromosome screening for chromosomal disorders in the first trimester, indicating that they have a higher risk for anxiety disorders. The population recruited in [46] were women with disorders during pregnancy and childbirth, such as antenatal complications, instrumental delivery, elective or emergency cesarean section, and postnatal complications. This study identified articles that explained the risk factors for anxiety because respondents faced situations that could increase psychological vulnerability, including postpartum anxiety and depression. The study population [43] involved women with preterm labor (PTL), which is a significant medical condition during pregnancy.
Counseling interventions were used in two studies, one using educational counseling and the other using supportive counseling [46, 47]. Mindfulness interventions were used in five studies, with most articles using digital innovations such as Smartphone-Based Mindfulness Training and WeChat to provide task guidance, notes, and daily mindfulness progress monitoring tools [18, 48,49,50,51].
The THP was used in one study conducted through online group sessions facilitated by midwives [52]. MI was used in two studies. The study by Abdollahi [41] used interview sessions with a psychotherapist and daily assignments recorded at home, and the study by Sapkota included an information booklet and a counsellor hotline that pregnant women could contact as needed [53]. Further details on the methods and results of each study can be found in Tables 1 and 2, and Table 3.
Description of interprofessional collaborations
There were five studies that involved collaboration between various health workers [18, 42, 43, 46, 48]. Two of these articles involved collaboration between midwives and clinical psychologists [42], one involved collaboration between nurses and obstetricians, another involved collaboration between psychologists, obstetricians, and clinical nurse specialists [39, 43, 44], and one involved collaboration between obstetricians, obstetric nurses, and two practitioners [18].
Nine studies did not involve collaboration or involved only one profession in implementing the intervention [41, 44,45,46,47, 49,50,51,52,53]. Seven studies involved psychotherapists in the intervention procedure, while two other articles involved nurses. These studies show a variety of intervention approaches used to improve maternal health through both cross-professional collaboration and interventions that focus on a particular profession. In this study, four articles used collaborative CBT interventions, and three studies did so without collaboration. Collaborative mindfulness interventions were recorded in five studies, while another five used mindfulness interventions without collaboration. All the studies that used MI were carried out without collaboration. Meanwhile, for the THP, only one study was carried out without collaboration.
The risk of bias in the studies included
An assessment of the risk of bias showed five studies with low risk [41, 42, 47, 48, 53], four studies with moderate risk [43, 45, 46, 50], and five studies with high risk (Fig. 2). In selection bias, nine studies were at low risk, two studies [49, 52] were at high risk, and three [43, 45, 46] were at moderate risk. All studies had a low risk of performance bias. Most studies had complete outcome data. However, three studies [18, 44, 51] had a high risk of attrition bias, and two studies [50, 52] had a moderate risk of bias. Only one study [51] had a high risk of measurement bias. All studies showed a low risk of reporting bias.
Risk of bias graph: reviewers’ domain assessments as study percentages
The highest risk of bias was found in the randomization process, which had two high-risk articles and three medium-risk articles, and regarding incomplete data, with three high-risk and two medium-risk studies. Adequate Respondent assistance can increase commitment to completing the task and reduce the risk of bias due to missing data.
Primary outcomes
Meta-Analysis
The total number of articles included in the meta-analysis was 13. Because there was one article that did not include post-test data during pregnancy, post-test data were available only after 32 weeks postpartum. These data did not match the PICOS of our meta-analysis, which only included anxiety measurements during pregnancy [46].
Intervention effect
Anxiety in the intervention group compared with the control group.
Forest plot showing anxiety outcomes of all interventions
The combined results of all interventions showed an overall negative effect size with SMD = − 0.64 (95% CI − 0.98–−0.31). This indicates that, on average, the psychological interventions significantly affected the intervention group compared to the control group. The high overall heterogeneity (I2 = 92%, p < 0,01) highlights the substantial variability in effect sizes across studies and interventions (see Fig. 3).
This variability highlights the importance of considering the context of individual studies when interpreting intervention effectiveness. Factors such as study location, implementation details such as intervention procedures, and participant characteristics (e.g., gestational age, age, and sociodemographic factors) should be considered to interpret the results better.
Subgroup analysis
Psychological intervention type subgroup
We used meta-analysis to examine several moderators of the observed effects. Clustering analyses covered a range of intervention approaches. Psychological interventions with a collaborative approach are efforts that involve the collaboration of multiple health professionals, such as psychologists, nurses, doctors, and midwives, to provide psychological support to pregnant women. Examples of these collaborative approaches include THP, MI Psychotherapy, Cognitive Behavioral Therapy (CBT) delivered by a multidisciplinary team, mindfulness facilitated by multiple professionals, and supportive counseling that involves various experts. Meanwhile, noncollaborative interventions are approaches in which psychological support is provided by only one type of health professional without the involvement of a cross-professional team. Two examples are CBT, delivered only by a psychologist without collaboration with other professionals, and mindfulness, facilitated by a single practitioner without the involvement of other health professionals.
Forest plots were used to show the comparative effectiveness of these psychological interventions, which included CBT, mindfulness, MI psychotherapy, supportive counseling, and the THP. SMD was used to measure effect size, which shows the difference in effectiveness between the control group and the experimental group in terms of the intervention (Fig. 4).
Forest plot showing anxiety outcomes from intervention subgroups in random effects model
The overall random effects model for CBT showed a negative SMD (− 0.80; 95% CI − 1.80–0.19) and high heterogeneity (I2 = 97%). This indicates significant variability in effect sizes across studies, which suggests that CBT may be more effective in specific contexts or populations. Therefore, further investigation is needed to identify moderating factors.
The random effects model for mindfulness showed a negative SMD (− 0.55; 95% CI − 0.78–−0.31). These results suggest that mindfulness has moderate benefits compared with the control group. Moderate heterogeneity (I² = 71%, τ² = 0.1031, p < 0.01) indicates significant variation among the studies included. This suggests that the effects of mindfulness may differ depending on the context or the population studied.
A study conducted by Abdollahi [41] and Sapkota [53] evaluated the effectiveness of MI therapy. The results showed that the SMD was negative (− 0.70; 95% CI − 1.08–−0.33). This means that MI therapy has a large and significant effect on improving the condition of participants compared with the control group. Since the confidence interval does not include zero, this result is statistically significant, strengthening the evidence that MI therapy is effective. The global weight of this study was 9.6%, which indicates its substantial contribution to the entire meta-analysis.
According to Esfandiari, the effectiveness of supportive counseling was evaluated. The results showed that the SMD was negative (− 0.73; CI 95% −1.19–−0.28) [47]. This means that supportive counseling has a large and significant effect on improving the condition of participants compared with the control group. Because the confidence interval does not cross zero, this result is statistically significant, strengthening the evidence that supportive counseling is effective. The global analysis weight of this study was 4.7%, which indicates its essential contribution to the overall meta-analysis.
A study conducted by Boran evaluated the effectiveness of the Thinking Healthy Program (THP) [52]. The results showed that the SMD was negative (− 0.35, 95% CI − 0.81–0.11). This SMD value indicates a small effect size and is not statistically significant because the confidence interval crosses zero. This means that although there is an indication that the THP may have an effect, the result is not strong enough to state the effectiveness of this program with certainty. The weight of this study in the global analysis was 4.7%, which indicates its contribution to the entire meta-analysis.
Collaboration subgroup
The studies included in the collaboration subgroup showed mixed results. The random effects model for the collaboration subgroup showed an overall SMD of − 0.88 (95% CI − 1.60–−0.15). Heterogeneity in this subgroup was very high, with an I² of 94%, a τ² of 1.1805, and p < 0.01, which indicates significant variation between studies.
The random effects model for the noncollaboration subgroup showed an overall SMD of − 0.47 (95% CI − 0.72–−0.22). Heterogeneity within this subgroup was also very high, with an I² of 87%, a τ² of 0.5514, and p < 0.01, which shows significant variation across studies. The SMD value in the collaboration group showed a more substantial effect size than the noncollaboration group and was statistically significant (Fig. 5).
Forest plot showing anxiety outcomes by collaborative and noncollaborative treatments
Length of follow-up
Duration = 8 (8 months follow-up) in Sun Y1, Sun Y2, Sun Y3 studies: SMD: 0.08 (CI: [-0.30, 0.46]), -0.17 (CI: [-0.64, 0.09]), and − 0.21 (CI: [-0.52, 0.09]) [18]. There are small and non-significant effects because the confidence interval includes zero. Moderate heterogeneity (I² = 47.5%), indicating moderate variation between studies. Duration = 4 (4 months follow-up) in Yang 1, Yang 2, Yang studies: SMD: -0.77 (CI: [-1.37, -0.17]), -0.80 (CI: [-1.24, -0.34]), and − 0.25 (CI: [-0.85, 0.34]) [48], Significant effects only in the first two studies; the latter effect was not significant. High heterogeneity (I² = 77.4%), indicating considerable variation between studies. Duration = 16 (16 months follow-up) Burger 1, Burger 2: SMD: 0.27 (CI: [-0.02, 0.52]) and 0.83 (CI: [0.39, 1.26]) [44], Small to medium effect, but only the second study was significant, Moderate heterogeneity (I² = 55.9%). Duration = 10 (10 Months Follow-Up) Forsell: SMD: -0.60 (CI: [-1.22, 0.02]), Medium effect but approaching significance [45]. Duration = 19 (19 months follow-up) Zhang Lin 2, Zhang Lin 3: SMD: -0.39 (CI: [-0.77, -0.01]) and − 0.30 (CI: [-0.56, -0.05]), Small but significant effect [51]. There is no heterogeneity between studies (I² = 0%). Duration = 25 (25 months follow-up) Zhang Xuan 2, Zhang Xuan 3: SMD: -0.85 (CI: [-1.18, -0.52]) and − 0.66 (CI: [-0.94, -0.37]) [50]. Significant and moderately large effect. Low heterogeneity (I² = 0%). Duration = 5 (5 months follow-up) Abdollahi: SMD: -0.64 (CI: [-0.84, -0.44]) [41]. Significant effect with moderate strength. Overall, the pooled SMD value was − 0.64 [95% CI: -0.96, -0.31], indicating a significant effect with a negative direction. The total heterogeneity was I² = 91.6%, indicating a high variation between studies. The subgroup difference test showed a significant difference between groups based on duration (p < 0.0001). The SMD values ranged from − 0.85 to -0.66, with the 25-week duration group showing a more significant effect size than the other groups and was statistically significant. The longest duration (25 months) follow-up durations tended to have more significant effects than moderate durations (Fig. 6).
Forest Plot showing the duration of observation or follow-up after intervention
Publication Bias analysis
The results of the funnel plot demonstrated a relatively asymmetrical distribution pattern, indicating the presence of publication bias. Visually, Fig. 7 reveals an imbalanced funnel plot, with a differing number of studies on each side (12 versus 8 studies). Moreover, several studies, both small and large in size, were outside the funnel triangle, reflecting varying and imprecise effect sizes. Notably, the number of studies outside the triangle was greater on one side. This visual indication of publication bias was further confirmed through statistical analysis using Egger’s test, which yielded a p-value of 0.0009.
link
