Association between maternal overprotection and premenstrual disorder: a machine learning based exploratory study

Premenstrual disorder (PMD), which includes premenstrual syndrome and premenstrual dysphoric disorder, has a complex pathogenesis and may be closely related to emotional cognition and memory. However, the mechanisms underlying these associations remain unclear. Therefore, this study used machine learning to explore the roles of various factors that are not typically considered risk-factors for PMD.

A predictive model for PMD was constructed using a dataset of questionnaire responses and heartrate variability data collected from 60 participants during their follicular and luteal phases. Based on the Japanese version of the Premenstrual Symptom Screening Tool, the binary objective variable (PMD status) was defined as "PMD" for moderate-to-severe premenstrual syndrome and premenstrual dysphoric disorder and other conditions as "non-PMD." The contribution of each feature to the predictive model was assessed using the Shapley Additive exPlanations (SHAP) model-interpretation framework.

Of the 58 participants (providing 117 data points), 17 (34 data points) were in the PMD group and 41 (83 data points) were in the non-PMD group. The area under the receiver operating characteristic curve was 0.90 (95% confidence interval: 0.82-0.98). Among the top 20 features with the highest SHAP values, six were associated with maternal bonding. Four of the six mother-related characteristics were associated with overprotection.

Based on these findings, parental bonding experiences, including maternal overprotection, may be associated with the presence of PMD.