Postpartum depression (PPD) is one of the most prevalent emotional disorders following childbirth, with implications for both the mother and the developmental outcomes of the infant, as well as the familial dynamics within the mother’s immediate social environment (1, 2). Although conditions such as the ‘baby blues’ (BB) are usually mild and self-limiting, they are recognized as potential precursors to more severe pathologies, including postpartum depression (PPD) and postpartum psychosis (PPP) (1). Given this situation, early detection becomes a public health imperative.

Although the term postpartum depression (PPD) is widely used in clinical practice and research, growing evidence suggests that it represents a partial and sometimes misleading label. Approximately 50% of depressive episodes diagnosed during the postpartum period have their onset during pregnancy, supporting a broader conceptualization within the perinatal (or peripartum) depression spectrum. This conceptual clarification, increasingly recognized in the literature, highlights the clinical relevance of preventive strategies and early detection during the antepartum period rather than an exclusive focus on the postpartum stage (3).

The prevalence of postpartum emotional disorders displays considerable variation on a geographical basis, depending on sociocultural and economic factors as well as the availability of healthcare services. Rogers et al. (4) demonstrate that in low- and middle-income countries, such as a significant portion of Asia, the prevalence of the condition can vary significantly from 3.5% to 63.3%, depending on the specific region. In contrast, Europe exhibits a more uniform prevalence, although it is also high, with a reported 55.2%. These disparities are associated with inequalities in access to perinatal care, stigma around mental health, and inadequate implementation of early detection protocols, which hinders timely diagnosis and effective treatment of these medical conditions. These figures underscore the necessity for preventive interventions initiated during pregnancy, particularly among women with a history of psychological issues or exposure to traumatic events (4, 5).

Postpartum emotional disorders, including anxiety, fears, and post-traumatic stress disorder (PTSD), manifest themselves through physiological and cognitive symptoms that affect maternal well-being and hinder the establishment of an emotional bond with the newborn (6, 7). Moreover, prenatal stress has been documented as a risk factor for autism spectrum disorders, attention deficit hyperactivity disorder, and psychotic disorders in children (8).

The field of artificial intelligence (AI), an interdisciplinary area of study that focuses on the development of intelligent systems capable of performing tasks that require analysis, reasoning, and autonomous decision-making, has emerged as a promising tool in the healthcare field. When applied to perinatal health, it facilitates enhanced prediction and monitoring of PPD through techniques such as machine learning (9). These technologies have proven useful not only in diagnosing PPD, but also in associated conditions such as gestational diabetes and premature birth (10).

Machine learning, a key component of AI, facilitates the analysis of substantial data sets to discern clinical patterns that are conducive to disease prediction. There are four distinct categories: supervised, unsupervised, semi-supervised, and reinforcement learning (11). These methodologies have already been employed with great success in the classification of medical images, prediction of clinical values, and improvement of hospital triage (12, 13).

The impact of PPD on a woman’s quality of life is notable, manifesting in feelings of worthlessness, guilt, fatigue, and breastfeeding problems. This situation has implications for the infant’s cognitive and physical development, as well as for family stability (14, 15). The implications of this dynamic extend to the father, who frequently articulates feelings of frustration and disengagement, stemming from his perceived inability to provide adequate support (16). Various biological factors (obesity, caesarean section, gestational diabetes), psychological factors (anxiety, chronic stress, low self-esteem) and social factors (insufficient support, poverty, migration) are identified as determinants of PPD (17, 18). Furthermore, obstetric violence and a lack of psychosocial preparation have been demonstrated to significantly increase the risk (19).

Healthcare professionals play a key role in the detection and monitoring of maternal emotional health issues. The utilization of tools such as the Edinburgh Postnatal Depression Scale (EPDS) facilitates the process of early screening (20). The provision of humanized and continuous care, in conjunction with health education, has been demonstrated to empower mothers and promote comprehensive recovery (21, 22).

Overall, the field of AI represents a disruptive advance in addressing postpartum emotional disorders by offering tools capable of identifying complex patterns in large volumes of data, which improves diagnostic accuracy and enables personalized interventions (5, 12). Rather than replacing traditional clinical assessment, it is more accurate to consider AI as a complement that optimizes screening and follow-up protocols, thereby promoting the integration of maternal mental health into healthcare systems (13, 23). The active involvement of healthcare professionals in the implementation and management of these technologies is vital to ensure their applicability and equity in different contexts, thereby guaranteeing comprehensive, evidence-based maternal care (2, 20).

The objective of this study was to evaluate the role of AI in the prevention and early detection of PPD.

A total of 1,857 articles were identified by searching the Scopus, Pubmed, WOS, and CINAHL databases.

After removing duplicate records (n=235), a total of 1,622 documents were obtained.

Then, by applying the exclusion criteria and implementing the PRISMA recommendations (24), 878 documents were excluded for various reasons that failed to meet the inclusion criteria: conducted more than 5 years ago (n= 333), type of study (n= 535), language (n=10) (Figure 1).

Following the initial screening of the title and abstract, the number of studies was reduced from 142 to 138. This was further reduced to 16 after reading the full text, and these were finally included in the review (Figure 1). Of the 16 articles, 4 were cross-sectional studies, 6 were retrospective cohort studies, 4 were prospective cohort studies, 1 was a longitudinal cohort study, and 1 was a comparative observational study. Six of the articles had been conducted in the United States, three in India, two in Sweden, one in Indonesia, one in Japan, one in China, one in Australia, and one in Italy (Table 3).

The present systematic review analyses the emerging role of AI in the detection and prevention of postpartum depression, addressing whether this technology can overcome the limitations of conventional methods. The findings indicate that machine learning models, such as Random Forest and XGBoost, have demonstrated high accuracy in early identification of women at risk of PPD, with markedly improved predictive performance over conventional approaches (34–37).

Despite this, there is a lack of consensus on the efficiency of AI, underscoring the need for optimized statistical techniques such as SMOTE and Bootstrapping to enhance prediction accuracy (28). Moreover, the significance of incorporating biological and psychosocial factors in conjunction with multimodal data into predictive models is underscored to facilitate a more comprehensive perspective (33).

An additional significant aspect to consider is the applicability of AI in real-world contexts. Despite the technology’s demonstrated efficacy in clinical settings and in the analysis of electronic medical records and social media, challenges persist regarding technological infrastructure, acceptance by healthcare personnel, and ethical concerns involving data privacy and algorithmic bias (31, 37).

Finally, emphasis should be placed on the fact that AI should not replace clinical work; rather, it should act as a complementary tool that optimizes screening protocols and facilitates more personalized and effective interventions. A comprehensive approach, encompassing both parental mental health and adaptation to diverse sociocultural contexts, is crucial for the future advancement of these technologies (30, 39).

These findings should be interpreted in light of the growing recognition that postpartum depression is a misnomer, as a substantial proportion (estimated at around 50%) of depressive episodes identified during the postpartum period originate during the antepartum stage. From this perspective, the potential value of AI-based detection models extends beyond postpartum screening to earlier, preventive identification within the broader perinatal mental health continuum (3).

Notwithstanding the methodological robustness of this systematic review and meta-analysis, it is imperative to acknowledge certain limitations that may influence the interpretation of the results. Firstly, considerable methodological heterogeneity was observed among the included studies, as evidenced by the high I² values found in both meta-analyses, which exceeded 90%. This variability may be attributable to differences in populations, algorithms employed, data sources, and evaluation metrics, leading to limitations in direct comparability across studies (29, 34, 35).

It is also worthy of note that, despite the rigorous quality assessment tools (JBI) that were applied, some studies exhibited minimal acceptable scores. This may have had an impact on the robustness of the integrated evidence (42). A number of studies were identified that exhibited potential publication bias, as evidenced by the absence of negative reports or insignificant results. This bias may have led to an overestimation of the effectiveness of the analyzed models, despite the non-significance of Egger’s test.

Several studies used secondary data from social media or electronic medical records, a methodology that, while innovative, may induce selection biases and restrict the generalizability of findings to other populations (26, 27, 40). Lack of standardization in performance metrics, including sensitivity, specificity, and accuracy, further complicates quantitative and qualitative comparisons between models.

Regarding the systematic review, despite an exhaustive search being conducted in international databases, studies in languages other than English or Spanish were not included. This may have introduced a linguistic bias and limited the inclusion of relevant evidence from other regions. Finally, the exclusion of grey literature, such as theses or technical reports, may have resulted in the omission of unpublished studies that could offer valuable data in this emerging field.

Recent advancements in the field of AI have led to significant progress in the prevention and detection of PPD. These developments have resulted in more accurate, personalized, and accessible diagnostic methods. The system’s capacity to analyze substantial quantities of clinical data and identify patterns on social media enables the early identification of risks, thereby facilitating timely interventions that complement conventional maternal healthcare services.

However, there are significant challenges associated with the implementation of AI, including the potential for algorithmic bias, a paucity of specific training for healthcare professionals, and ethical issues related to data protection. In order to overcome the barriers to the integration of such technologies, there is a need for an appropriate regulatory framework, awareness-raising strategies, and robust technological infrastructure.

In conclusion, the integration of AI as a support to clinical judgement, particularly in conjunction with nursing care, is essential for providing comprehensive and equitable maternal care tailored to individual needs. Future research should focus on developing models that are more explainable, accessible, and ethically responsible. Such models should be promoted to facilitate their adoption within health systems, with the aim of enhancing the quality of postpartum care.

Although this review focuses on postpartum depression, the findings should be interpreted within the broader framework of perinatal mental health. Expanding AI-based models to the antepartum period may enable earlier, preventive identification of depressive symptoms and facilitate timely implementation of psychoeducation, psychological support, structured follow-up, and referral for appropriate therapeutic interventions. Such an approach ensures that early detection translates into meaningful improvements in maternal and infant outcomes, rather than remaining a purely predictive exercise.