The protocol for this systematic review was reported according to PRISMA guidelines (15). The search strategy was designed by a biomedical information specialist (WMB) from the Medical Library of Erasmus MC, University Medical Center Rotterdam (16). We searched Embase, MEDLINE, Web of Science, Cochrane Central Register of Controlled Trials, CINAHL and, PsycINFO for published, peer reviewed original articles. The search combined terms for (1) perinatal mental health problems, serious mental illness (i.e. schizophrenia, mood disorders, personality disorders, anxiety), eating disorders, substance use disorders or autism, and (2) midwives’ knowledge, attitudes, skills or training needs, as well as existing training programs for midwives on PMH. We included only published articles in English or French. No time restriction was set. The search was updated prior to publication on 21 June 2023. We hand-searched the reference list of three systematic literature reviews (7, 10, 11) for additional relevant articles. The full search strategy, search terms and syntax are presented in online Supplementary Table 1.

To be included, articles had to meet all the following criteria: 1) focus on midwives (included midwives, nurse-midwives, registered midwives, registered midwives tutors, registered midwives prescribers and registered advanced midwives practitioners - referred as “midwives” in this review); 2) reporting on midwives’ training needs in PMH, PMHPs or related conditions or existing training programs that focus on the use of screening tools to detect PMHPs, on PMH in general or specific aspects of PMH; 3) using quantitative, qualitative or mixed-methods design. For training programs, we included uncontrolled and controlled studies (placebo, TAU or active comparators).

Our exclusion criteria were: 1) no full text available or studies published in languages other than English or French; 2) grey literature because the aim of this systematic review was to guide the development of future interventions; 3) training programs on psychological interventions (e.g. cognitive behavior therapy) because this review focused on interventions aiming at improving midwives’ training on essential competencies related to PMH (e.g. PMH assessment, detection, referral and support of parents with PMHPs).

The screening process was conducted in two separate stages: 1) Two authors (M.D. and J.D) independently screened the title and abstracts of all non-duplicated papers excluding those not relevant. Potential discrepancies were resolved by consensus; 2) Two authors (M.D. and J.D) independently applied eligibility criteria and screened the full-text papers to select the included studies. Disputed items were solved discussing together and reading further the paper to reach a final decision. Supplementary Tables 2, 3 present the list of included/excluded studies. Inter-rater reliability was calculated (kappa=0.90).

Two authors (MD and JD) performed independently the data extraction. For each study, we extracted the following information: general information (author, year of publication, country, design, type of study, population considered, period), assessment tools or methods, cultural aspects, the main findings and variables relating to quality assessment. For studies reporting on training programs, we also extracted information about the intervention (nature, type, length, targeted skills or outcomes, format), outcome measures and effectiveness on midwives’ knowledge, attitude, skills or routine use of screening tools to detect PMHPs or parents’ outcomes (e.g. depressive symptoms). Tables 2–6 present the factors associated with knowledge, skills, confidence and decisions about screening, referral or support. Supplementary Tables 4, 5 present the detailed characteristics of the included studies.

Quality assessment was realized using the Mixed Methods Appraisal Tool (MMAT) (61). MMAT is a validated instrument to assess the methodological quality of qualitative, randomized controlled trials, non-randomized trials, descriptive studies, and mixed methods studies. It is comprised of five 5-item subscales assessing different aspects of quality (e.g. appropriateness of the selected design/methods/measurements, integration of quantitative and qualitative parts for mixed-methods studies). Two researchers (MD and JD) independently assessed methodological quality using the MMAT and extracted MMAT scores for each article. Discrepancies were resolved through consensus. The MMAT overall quality score and detailed scores are provided in Supplementary Tables 4, 5. The study protocol was registered on PROSPERO on November 1, 2021 (CRD42021285926).

The characteristics of the 66 included studies are presented on Tables 7, 8. Most studies were conducted in high-income countries (89.4%) and published after 2015 (50%). Study designs were quantitative (n=33; 50%), qualitative (n=22; 33.3%) or mixed-methods (n=11; 16.7%). Samples included qualified midwives (n=37; 56.0%), qualified midwives and other perinatal health providers (n=17; 25.8%) and student midwives (n=11; 16.7%). Qualified midwives had a variable level of training in PMH ranging from none to 90% (specified in 24 studies; most covered topics: general information about PMH and PMHPs; least covered topics: interviewing/counseling skills, psychopharmacology and suicide risk assessment). Eight studies (12.1%) reported on midwives’ mental health nursing experience (ranging from 0.8% to 30%) or placement experience in a mental health setting or a mother-baby unit during their studies (ranging to 9% to 23.2%). Four studies (6%) mentioned family or personal experience of mental health problems ranging from 25% to 66.3%. Most studies covered the entire perinatal period (n=44; 66.7%) and reported on PMHPs (n=32; 48.5%). The definition of PMHPs was highly variable across the studies (e.g. inclusion of conditions usually not considered as PMHPs, such as schizophrenia, bipolar disorder, personality disorders, self-harm, suicide eating disorders or SUD in 16 studies; definition restricted to anxiety, depression, postpartum psychosis and/or posttraumatic stress disorder in 9 studies; unspecified in 7 studies). One third of the included studies used validated instruments to assess outcomes (n=16; 36.4%). Five studies (7.6%) investigated the influence of cultural aspects on the detection and management of PMHPs.

Of 15 studies reporting on a training program using a quantitative or a mixed-methods design, three used a waiting-list control group (20%; one randomized controlled trial (RCT)) and 13 (86.7%) were uncontrolled. Sample size was small in most studies (< 50 participants; n=9 studies). Nine studies (47.3%) reported contact with persons with lived experience when designing their training program. The training programs were heterogeneous in nature (initial training, n=6, 31.6%; continuous education, n=13, 68.4%), type, format and duration (ranging from 2 minutes to a fifteen-week module). All studies assessed training outcomes either immediately after (n=15; 79%) or up to 3 months after the intervention is delivered (n=4; 21%).

The overall assessment score ranged from low (n=30, 45.4%; n=13, 68.4% for training programs) to high (n=11, 16.7%; n=2, 10.5%). For quantitative or mixed-methods studies, the reasons were convenience sampling (n=61 studies, 92.4%), sample size, low response rate (n=18 studies > 60%), limited use of validated outcome measures (36.4%), use of self-reported measures, absence or short duration of the follow-up period, limited integration of the results in mixed-methods studies and lack of controlled/RCT studies to evaluate the effectiveness of training programs. For qualitative studies, the reasons were interpretation bias (e.g. no investigator triangulation, the data being analyzed by only one researcher), absence of data saturation and lack of reflexivity.

Many studies found that midwives felt ill equipped to care for parents with PMHPs [e.g. ranging from 69.2% of 815 midwives in Jones et al., 2011 (17) to 82.2% of 157 midwives in Noonan et al., 2018 (28)]. The reasons included insufficient initial training/continuous education on PMH (n=2 studies), perception that PMH assessment is not part of their role (n=2 studies), lack of knowledge about the detection, referral and management of PMHPs (n=12 studies). Compared with other perinatal health providers (GPs, health visitors, maternal child health nurses; n=11 studies), midwives had lower knowledge on PMH (n=2), felt less confident in the detection, referral or management of PMHPs (n=3) and had more negative attitudes toward their role in perinatal mental healthcare (PMHC) (57) or suicide prevention (40). Self-reported barriers to discuss PMH issues or self-reported interviewing skills did not differ between nurses and midwives (25). Student midwives’ knowledge, skills and attitudes in PMH did not clearly differ from those of qualified midwives (n=5 studies). On the job experience, learning from peers and attending to workshops/conferences were midwives’ main sources of knowledge (n=3 studies).

The factors positively associated with knowledge about PMHPs included the perception to be well equipped to provide PMHC (66.7% significance), previous training in PMH (50% significance), younger age (17), shorter work experience in general and as a midwife (20% significance), frequent contact with parents with PMHPs (50% significance) and type of practice (33.3% significance). Mental health nursing experience was positively associated with the perception to be well equipped to provide PMHC, but not with higher knowledge about PMH (8). No significant association was found between confidence in providing PMHC and other factors [e.g. age, personal experience of mental health problems, frequent contact with parents with PMHPs (29)], except for PMH education and case identification (8). Compared with suicide risk assessment and other conditions (e.g. postpartum psychosis, SMI, eating disorders or posttraumatic stress disorder; n=4 studies), midwives reported higher knowledge, better skills and more confidence in detecting and managing perinatal depression and anxiety. Midwives felt in general ill equipped to care for postpartum psychosis, eating disorders, posttraumatic stress and SMI (n=10 studies) and reported ambivalent or negative attitudes toward parents with these conditions (n=7 studies). Knowledge about PMHPs varied according to the assessment method [i.e. higher self-report knowledge than researcher-rated knowledge (19, 43)] and the timing of perinatal period (i.e. higher in the postpartum than during pregnancy, n=5 studies).

While the need to improve midwives’ initial and continuous education in PMH is now well established (7, 10), student midwives, midwives and even specialist midwives continue reporting to feel ill prepared to care for parents with PMHP in particular in case of co-occurring SMI (9, 21, 24, 33, 34, 36). Moreover, the proportion of midwives who received education in PMH - in particular in topics such as mental health/suicide risk assessment - remains consistently low. Given suicide is the leading cause of maternal mortality in the 1^st year postpartum in high-income countries, this is concerning (1, 80).

Aligning with previous research (7, 10, 11), this systematic review found that education/training programs had positive effects on proximal outcomes (e.g. midwives’ knowledge, skills, attitudes and confidence in providing PMHC) and contrasted effects on distal outcomes (e.g. screening in maternity wards, the number of referrals or depressive symptoms). This could be related to methodological bias (e.g. lack of RCT or quasi-experimental studies). There is a need for high-quality studies on interventions designed following the Medical Research Council framework for complex interventions (81), which proposes among other core elements to: 1) take into account the context of delivery; 2) use a clear theoretical basis (e.g. how the intervention is expected to produce positive effects and under which conditions) and; 3) promote a meaningful engagement of persons with lived experience among other relevant stakeholders.

According to Wadephul et al. (2018) (82) framework for assessing midwifery practice in PMH, knowledge, confidence, attitudes and organizational factors influence midwives’ ability to detect and manage PMHPs. However, higher knowledge about PMH does not necessarily translate into higher confidence in providing PMHC and the opposite (8). As reported in one of the articles included in this review (42) and aligning with the theory of planned behavior (82), additional factors such as individual values, e.g. personal interest in PMH, and behavioral intent (e.g. the intention to open discussions about PMH) could influence detection and decision-making about referral and support in PMHPs and thus be relevant for midwifery education.

To improve midwives’ engagement into PMHC, training programs should put PMH in context (e.g. the positive outcomes that could be achieved with appropriate support) before covering topics related to specific knowledge or skills (5, 38, 49, 50, 53, 54). Instead of focusing only on biomedical aspects (e.g. the signs, risk factors, consequences and treatments of PMHPs), programs should propose a continuum approach of PMH that covers the positive aspects of the person’s life including wellbeing and personal recovery (83–86).

Extending the findings of previous reviews (7, 10, 11), training programs should target student midwives, midwives and specialist midwives and cover interviewing and distress management skills with a focus on specific aspects (e.g. opening discussions without feeling intrusive, using flexibly screening tools and reacting in case of a positive answer) (5, 21, 38, 45, 49, 50, 53, 54). In addition, training programs should include clinical supervision by mental health providers during and after intervention delivery (14). Future studies should include a longer follow-up period, as the embedding of practice change requires a minimum of nine months after the intervention is delivered (87).

Finally, while contact with persons with lived experience is one of the most effective strategies to reduce mental illness stigma in the general public and in frontline health providers (88, 89), this review found a very low proportion of training programs that engaged persons with lived experience in the conception and delivery of the intervention. Initial and continuous midwifery education curriculums on PMH should involve persons with lived experience - co-design and co-intervention - and include content about personal recovery/person-centered care (72, 73, 81, 84, 90–92).

There are limitations. First, despite a growing number of published studies on midwives’ training needs in PMH and training interventions designed for this population (n=66 studies in this review vs. n=22 (7); n=17 (10); n=43 (11);), the quality of the included studies remains low to moderate, a concerning finding given the clinical relevance of this topic that is also a considerable limitation. Among other methodological bias, the absence of a clear theoretical basis for designing interventions (81), the small or unjustified sample sizes, the lack of RCT/quasi-experimental studies, the absence of control groups (or active comparators in controlled studies) and the absence or short duration of follow-up makes unclear whether interventions have positive effects on proximal or distal outcomes. Future high-quality studies on this topic are therefore needed. Despite these limitations, the inclusion of quantitative, qualitative and mixed-methods studies provided a complete synthesis of the available evidence and consistent messages emerged across studies. Second, relevant studies may have been missed since we excluded studies published in other languages than English or French and did not include the grey literature in our searches.