Empirical examination of early maternal–infant bonding disturbances: Latent class analysis and development of a clinical decision rule

Background Early maternal–infant bonding is important for infant development and maternal caregiving, and disturbances in bonding are linked to adverse outcomes. Although Brockington et al. (2006) proposed a clinical framework for bonding disturbances, these have not been examined using statistical classification methods. Objective To evaluate the construct validity of early maternal–infant bonding disturbances using latent class analysis (LCA) and to develop a clinically applicable decision rule. Methods A clinical sample of 211 postpartum mothers attending a specialised perinatal mental health unit was assessed using the Stafford Interview, the Postpartum Bonding Questionnaire (PBQ), and the Edinburgh Postnatal Depression Scale (EPDS). Thirteen dichotomous estimators derived from the Stafford Interview were entered into an LCA to identify latent classes. Associations with Brockington et al. (2006) categories and with clinical, psychosocial, and infant factors were examined. A decision rule was derived from the LCA results. Results Four latent classes were identified: normal bonding (55.5%), mild bonding disturbance (26%), mild bonding disturbance with anxiety (10%), and severe bonding disturbance (8.5%). Latent classes partially overlapped with Brockington et al. (2006) categories. Concordance between decision rule-based categories and LCA classes was good to very good (κ = 0.75–0.93). Classes were associated with PBQ scores, depressive symptoms, and difficult infant temperament. Conclusions This study provides preliminary empirical support for the proposed classification of early maternal–infant bonding disturbances and suggests that a clinically applicable decision rule derived from LCA may be useful for their identification in specialist clinical settings.