Mental Health Support Chatbot Effectiveness During Pregnancy and Postpartum
According to recent research, providing mental health support through chatbots during pregnancy and postpartum periods, even with clinical monitoring, is considered "insufficient" for enhancing outcomes.
A trial involving 400 maternal patients revealed minimal improvements in anxiety and depression levels. Patients were divided into two groups: a control group that received a children's book monthly for a year and an experimental group that had access to a generative AI chatbot named Rosie. The experimental group could earn rewards for asking questions, including a set of children's books for the most active users.
The study found that patients using Rosie frequently reported higher anxiety levels, but also better sleep. It's unclear if these users were inherently more anxious or if anxiety was exacerbated by chatbot use.
Researchers concluded that:
"A generative AI maternal health chatbot did not improve the primary outcomes of anxiety or depression at 12 months compared with active control... [chatbots] may be better positioned as adjunctive tools rather than stand-alone interventions for improving maternal health."
The trial focused on pregnant and postpartum patients from racial minority backgrounds, who often face healthcare disparities and worse health outcomes, including higher maternal mortality rates among Black and Hispanic patients. A separate study indicated that non-white postpartum patients are less likely to receive mental health care.
While the research didn't show significant large-scale benefits, the authors suggested potential smaller advantages from the information-oriented format. Rosie was trained on various methods, including community engagement strategies, diverse user feedback, and evidence-based mental health knowledge, and could detect signs of self-harm or distress, triggering a review by a clinical social worker.
Findings varied based on marital status and caregiving responsibilities. Never-married or separated participants were more likely to have emergency room visits and report pregnancy complications. Similarly, those caring for children under 7 months old experienced higher anxiety levels.
The researchers noted that subgroup findings should be further studied and interpreted cautiously:
"The null mental health findings suggest that conversational, information-oriented AI support alone may be insufficient to improve perinatal mental health, which is influenced by structural stressors, caregiving burden, financial strain, social support, and access to clinical care."