Early identification of postpartum depression using demographic, clinical, and digital phenotyping

Background Postpartum depression (PPD) and adjustment disorder (AD) affect up to 25% of women after childbirth. However, there are no accurate screening tools for either disorder to identify at-risk mothers and enable them to benefit from early intervention. Aims To evaluate combinations of anamnestic, clinical and remote assessments for an early and accurate identification of PPD and AD. Method Two cohorts of mothers giving birth were included in the study (N=308 and N=193). At baseline, participants underwent a detailed anamnestic and clinical interview. Remote assessments were collected over twelve weeks comprising mood and stress levels as well as depression and attachment scores. At twelve weeks postpartum, an experienced clinician assigned the participants to three distinct groups: women with PPD, women with AD, and healthy controls (HC). Combinations of these assessments were assessed for an early an accurate detection of PPD and AD in the first cohort and, after pre-registration, prospectively validated in the second cohort. Results Combinations of postnatal depression, attachment (for AD) and mood scores at week 3 achieved balanced accuracies of 93% and 79% for differentiation of PPD and AD from HC in the prospective validation cohort. Differentiation between AD and PPD, with a balanced accuracy of 73 % was possible at week 6 based on mood levels only. Conclusions Combinations of in clinic and remote self-assessments allowed for early and accurate detection of PPD and AD as early as three weeks postpartum, enabling early intervention to the benefit of both mothers and children.