An AI-Directed CDS Tool to Reduce Iron Deficiency Anemia in Pregnancy: A Randomized Controlled Trial (AID-IDA Trial)
Integrating a predictive model into the electronic health record (EHR) via a clinical decision support (CDS) tool provides a scalable, resource-conscious solution for improving iron deficiency screening and management, potentially reducing maternal morbidity and mortality from postpartum hemorrhage in the United States and advancing the use of AI-aided clinical tools in obstetrics at the point of care.
Project Details -
Ongoing
-
Grant NumberR21 HS030148
-
AHRQ Funded Amount$274,984
-
Principal Investigator(s)
-
Organization
-
LocationBostonMassachusetts
-
Project Dates08/01/2024 - 07/31/2026
-
Care Setting
-
Medical Condition
-
Population
-
Type of Care
-
Health Care Theme
Each year, nearly 4 million people give birth in the United States, with an increasing number experiencing complications like postpartum hemorrhage (PPH), which affects 10 percent of deliveries and significantly contributes to maternal morbidity. While effective PPH management is crucial, individuals without anemia are better equipped to handle blood loss during delivery. Studies show that anemia increases the risk of maternal complications and severe morbidity, with more severe anemia linked to higher rates of PPH, maternal shock, and intensive care admissions, particularly impacting non-Hispanic Black patients who experience higher rates of anemia. Despite routine screening during pregnancy, 35 percent of individuals still present with anemia at delivery. The current two-step screening method recommended by the American College of Obstetricians and Gynecologists (ACOG) tests for iron deficiency only if anemia is already present. This often leads to missed or delayed diagnoses, resulting in inadequate treatment, as oral iron therapy—the first-line treatment for iron deficiency anemia—takes weeks to replenish iron stores. Intravenous (IV) iron is a quicker, more effective alternative, though it requires clinic resources for administration. At the study institution, over 33 percent of patients who are anemic in the third trimester are not screened for iron deficiency despite following ACOG guidelines, highlighting a need for other solutions such as in digital health. This research aims to develop and test a CDS tool to enhance anemia screening and management, particularly for those at high risk of PPH, to reduce maternal morbidity and address this public health concern.
The specific aims of the research are as follows:
- Develop and validate a PPH prediction model using structured data available before the 3rd trimester.
- Design a CDS tool to improve iron deficiency management among patients at high risk for PPH.
- Determine the efficacy of the CDS tool to reduce anemia at delivery among those at high risk for PPH.
- Quantify the acceptability of the CDS tool among providers within the intervention arm of the randomized controlled trial (RCT).
Researchers will develop a machine learning-based predictive model using structured data from the end of the second trimester to identify individuals at high risk for PPH. They will then create a CDS tool integrated into the EHR to prompt providers to proactively screen and treat iron deficiency in these high-risk patients. Researchers will then conduct a RCT to evaluate the CDS tool’s effectiveness in reducing anemia prevalence before delivery and assess its acceptability among obstetric providers.
By enhancing anemia and iron deficiency screening protocols, researchers aim to improve pregnancy outcomes and reduce preventable complications, significantly lowering maternal morbidity and mortality in the United States, with a particular focus on addressing racial and ethnic disparities. This study seeks to inform and potentially transform current practices and policies for high-risk individuals, bridging a critical gap where traditional methods often fail to timely detect iron deficiency. Additionally, the research aims to advance the use of AI-aided clinical tools in obstetrics, promoting the widespread implementation of embedded machine-learning models to improve patient care at the point of delivery.