Dr. Kovacheva’s project is entitled: “Development of Novel Machine Learning Tool to Predict Risk for Severe Maternal Morbidity and Optimize Anesthesiology Resources.”
Background: The United States is the only developed country in which the rates of severe maternal morbidity have been steadily increasing over the past decade—this is an important patient safety priority. Every year in the United States, more than 50,000 women experience severe maternal morbidity, and 700 women die from pregnancy-related conditions1. Severe maternal morbidity is highly preventable and considered a “near miss”, since without timely treatment or resources it may lead to maternal death2. There are significant racial disparities in outcomes, and Black women are up to four times more likely to suffer severe maternal morbidity compared to White women3. The risk-adjusted severe maternal morbidity rates can vary up to six times among hospitals, suggesting a large contribution of the quality of care to observed racial disparities in pregnancy-related outcomes3. Up to 46% of Black and 33% of White maternal deaths could be prevented by improving the quality of hospital care4. However, there is currently no universally utilized or validated severe maternal morbidity prediction tool in clinical obstetric practice. Machine learning tools to combine various clinical risk factors have recently become available. In addition, novel approaches, like explainable artificial intelligence are being developed to aid performance evaluation, un-biasing and transparency of the decision-making process.
Aims: In line with the APSF goals to improve patient safety, we propose to leverage our rich patient database and computational tools to improve maternal outcomes during delivery. We will design machine learning models using approaches like regression, decision tree models and neural networks. We will select the best performing model in all racial groups and determine the optimal conditions when anesthesiology resources should be mobilized. We will prospectively evaluate the model accuracy and determine blood product crossmatch, utilization, and staffing escalation. Our long-term goal is to develop a high-fidelity, personalized, and fair algorithm to predict the risk of severe maternal morbidity in pregnant women and support the anesthesiology provider in preparing for and managing the highest risk patients.
Implications: United States has one of the most advanced healthcare systems in the world, yet maternal morbidity and mortality are significantly higher than in similarly developed countries. There are significant practice variations across different states and hospital systems. Encouraging evidence-based stratification of high-risk pregnant patients is one of the two most important objectives launched by the Department of Health and Human Services to achieve the goal of 50% reduction in maternal mortality over the next 5 years5. Our proposed novel tool will aid identification of parturients at risk for adverse outcomes with the long-term goal of increasing maternal safety during delivery.
- Hoyert DL, Minino AM. Maternal Mortality in the United States: Changes in Coding, Publication, and Data Release, 2018. Natl Vital Stat Rep. 2020;69(2):1-18.
- American College of Obstetricians and Gynecologists, Society for Maternal-Fetal M. Obstetric Care Consensus No. 5: Severe Maternal Morbidity: Screening and Review. Obstet Gynecol. 2016;128(3):e54-60.
- Howell EA. Reducing Disparities in Severe Maternal Morbidity and Mortality. Clin Obstet Gynecol. 2018;61(2):387-399.
- Berg CJ, Harper MA, Atkinson SM, et al. Preventability of pregnancy-related deaths: results of a state-wide review. Obstet Gynecol. 2005;106(6):1228-1234.
- Healthy Women, Healthy Pregnancies, Healthy Futures: Action Plan to Improve Maternal Health in America: U.S. Department of Health and Human Services.;2020.
Funding: $149,998 (January 1, 2022-December 31, 2023). This grant was designated as the APSF/Medtronic Research Award, and was also designated as the APSF Ellison C. Pierce, Jr. MD Merit Award with $5000 unrestricted research support.