Direct Annotation of Unstructured Clinical Progress Notes Describes Postpartum Healthcare Needs and Encounter Characteristics of Women Whose Newborns Are Immediately Hospitalized in Intensive Care

Background: Among children’s hospitals with integrated high-risk prenatal care and birthing facilities, the reasons for postpartum care-seeking, and descriptions of those encounters, are not well established. At the study setting, postpartum care is frequently provided ad hoc and detailed in unstructured clinician progress notes. Purpose: This descriptive cohort study identifies the reasons for postpartum care-seeking and encounter characteristics among a cohort of women who gave birth on a specialized maternity unit (within a children’s hospital), while their infants remained in intensive care. Methods: Participants were included if they gave birth at the setting and had a documented postpartum maternal encounter that occurred after the time of their maternity discharge but prior to their infants’ discharge from intensive care (or death). Data were abstracted with structured query language, and relevant variables were identified and characterized using natural language processing, a subset of artificial intelligence, by expert annotation. Results: The cohort included 122 mother–infant dyads. On average, the women had 2.41 unanticipated postpartum encounters. The majority of women (n = 97/122, 79.5%) were seen in the first 2 weeks postpartum with a mean of 11.9 days. The highest frequency categories of postpartum healthcare needs were related to cesarean birth (n = 35/122, 28.7%) and lactation (n = 17/122, 13.9%). Implications for Practice and Research: Our findings reveal the need for integrated postpartum maternity care for women with a hospitalized neonate. These findings support the need for clinician or clinician scientist annotators to effectively label unstructured data to generate meaningful findings.