A Virtual Agent to Screen for Mental Health in Pregnant and Postpartum Women

Andrea Wittenborn, PhDPrincipal Investigator: Andrea Wittenborn, Associate Professor, Human Development and Family Studies (HDFS), (MSU)

Investigators. Michigan State University: Heather McCauley, Assistant Professor, HDFS; Mi Zhang, Assistant Professor, Electrical and Computer Engineering; and Amy Bonomi, Professor and Chair, HDFS. Sparrow Health System: Kathy Marble, Director, Women and Children’s Services.

Aims. Depression impacts 9-13% of pregnant and postpartum women, and half of women seeking care in reproductive health clinics experience intimate partner violence (IPV) in their lifetimes. The US Preventive Services Task Force (USPSTS) recommends universal screening for depression and IPV among pregnant and postpartum women given the substantial risk in women and infants. Yet, many women don’t receive depression or IPV screening (25% and 42%, respectively) during prenatal visits, and one-third to one-half of depression and IPV cases aren’t identified. Three major challenges limit the application of screening in prenatal and postnatal care: (1) depression and IPV carry a social stigma which reduces the likelihood of disclosure to a clinician, (2) clinicians have limited time to complete in-depth screenings given their clinical responsibilities, and (3) clinicians report feeling ill-prepared to respond to a disclosure, which reduces their likelihood of initiating screening. Artificial intelligence powered technologies are emerging as tools to improve health care delivery and quality, and these technologies could benefit screening for depression and IPV. Prior research suggests the use of a virtual agent increases the level of disclosure and expressions of emotion among participants. Virtual agents can also provide patients with feedback or resources based on their level of risk in real time. Thus, this project proposes the novel use of a virtual agent to screen for depression and IPV among pregnant and postpartum women. Specifically, we will (1) develop a protocol and script for the virtual agent to screen for depression and IPV, (2) develop a novel virtual agent powered by artificial intelligence by leveraging two existing technologies and developing one additional technology, and (3) assess the reliability of the virtual agent to screen depression and IPV.

Significance. This project has the potential to transform care for pregnant and postpartum women. Despite advances in screening, depression and IPV remain under-detected. When faced with intractable problems,
Clayton Christiansen noted the need for “disruptive technologies” that move beyond incremental improvements to a new vision of care. Using a virtual agent powered by artificial intelligence to overcome challenges associated with depression and IPV screening by human clinicians could provide the technology breakthrough needed to improve the detection of two common and costly health conditions.

Approach. Forty pregnant and postpartum women will be recruited from Sparrow OB/GYN clinics and asked to complete an assessment at MSU, including a virtual agent-based screening for depression and IPV, and the USPSTF recommended screening measures. Analyses will investigate the reliability of the virtual agent screening, to prepare us for an R01 to test the broader impact of the tool in transforming care for two highly prevalent and costly conditions.

Future Funding. Following completion, we will submit an R01 application to NIH PAR 14-180.