Detecting psychosocial types for mothers and partners in perinatal depression: a study of machine learning

Perinatal depression (PD) impacts parental relationships and child development. Most studies focused on maternal depression, leaving the psychosocial experiences of partners less explored. Furthermore, earlier studies frequently relied on univariate statistics, oversimplifying group-level outcomes. Through unsupervised Machine Learning (ML) techniques, this study aims to identify specific psychosocial profiles for mothers and partners, and to test their relevance in modulating PD levels. Expectant parents were recruited (N=120; WomenAge=34.9±4.3; PartnerAge=36.7±4.8) and completed questionnaires assessing psychosocial factors such as social support, affectivity, prenatal attachment style, and childhood trauma. Depression was measured using the Edinburgh Perinatal Depression Scale (EPDS). Two unsupervised K-Means clustering ML algorithms were fed with 15 variables to identify psychosocial types for both women and partners. ANOVA and t-tests were conducted to verify the effect of the identified psychosocial types on PD levels. Three psychosocial types were identified for women: (1) “Reassured mothers” (n=39), characterising women who had high support during pregnancy, no childhood trauma, and positive prenatal attachment; (2) “Anxious-Protective mothers” (n=64) characterizing women with substantial support but who were overly focused on their relationship with the fetus; (3) “Mothers marked by trauma” (n=17) describing women lacking support and experiencing traumatic events such as emotional abuse and neglect during childhood. For partners, two types were identified: (1) “Reassured partners” (n=72) and (2) “Partners marked by trauma” (n=48), analogous to the mothers’ profiles. Higher depressive symptoms were observed in all participants with a history of childhood trauma (all p<.05). The study identifies similar psychosocial types for both mothers and partners, emphasising the importance of support during pregnancy and the role of parent-fetus attachment. Furthermore, the study reinforces the impact of past trauma on the onset of perinatal depression through multivariate analytical techniques that highlight the subjectivity of experience. These findings might orient clinicians towards personalized interventions for preventing perinatal depression at the dyadic level, and not only at the single-subject level.