Friday, March 13, 2026
NIH-funded automated clinical decision support could facilitate timely intervention years before at-risk patients seek help.
A team of researchers funded by the National Institutes of Health (NIH) has developed an artificial intelligence (AI) tool to assist clinicians in decision-making by predicting whether a patient is at risk for intimate partner violence (IPV). The research team used data collected regularly during medical visits to train a machine learning model, a type of AI, that detected IPV in the patients studied with high accuracy.
IPV refers to abuse by a current or former partner that results in serious consequences, including potentially life-threatening injuries, chronic pain, and mental health disorders. Millions of people in the United States, both men and women, are affected at some point in their lives. However, many cases remain undetected because patients may be reluctant to disclose abusive relationships due to safety concerns, fear, or stigma.
In their study, a research team led by researchers at Harvard Medical School in Boston deployed three AI models for IPV detection in healthcare settings and compared their predictive performance.
“This clinical decision support tool could have a major impact on the prediction and prevention of intimate partner violence,” said Dr. Qi Duan, director of the Division of Health Information Technology at NIH’s National Institute of Biomedical Imaging and Bioengineering (NIBIB). “Given the prevalence of cases, this tool could be a game-changing asset for public health.”
According to the study authors, many cases of IPV go unrecognized, resulting in missed opportunities for timely intervention. They report that while current screening tools only capture a fraction of cases, clinical and imaging records provide valuable information in detecting IPV risk. In particular, radiologists have an advantage in recognizing signs of IPV, such as the frequency of certain patterns of physical trauma.
The researchers used several years of hospital data from approximately 850 affected female patients and 5,200 unaffected age- and demographic-matched control patients. Because the collection of relevant clinical data varies by clinical setting, the team designed two different AI models. One was trained on structured patient data in tabular format, and the other was trained on unstructured patient data in medical notes, including radiology reports. Additionally, we developed a multimodal model that blends both structured and unstructured data.
In this study, all models achieved high performance. However, the multimodal fusion model showed better performance than models using only structured or unstructured data. It was accurate 88% of the time. Both the tabular and fused models can detect IPV risk on average more than 3 years before a patient enrolls in a hospital-based domestic violence intervention center. The tabular model was able to recognize IPV risk slightly earlier, while the fused model was able to detect more IPV cases in advance.
The fusion model achieved more stable performance than relying on either modality alone. The scientists explained that the different modalities are processed separately and are only integrated at the prediction stage. They found that tabular frameworks are particularly relevant to the medical field. In the medical field, this is due to variations in data availability and recording of unstructured data from hospital to hospital.
The researchers emphasized that using AI tools, such as machine learning models, could help health care providers have timely conversations with patients about IPV and connect patients with appropriate support resources. These AI tools are not intended to make a definitive diagnosis.
“For decades, our health care system has relied heavily on patient self-disclosure to identify intimate partner violence, leaving many cases unrecognized and unsupported,” said study lead author Bhati Khurana, MD, an emergency radiologist at Massachusetts General Brigham and associate professor of radiology at Harvard Medical School. “Our work represents a fundamental shift from reactive disclosure to proactive risk recognition in routine clinical care. By analyzing patterns that already exist in medical data, this approach helps healthcare clinicians start earlier, safer, and more informed conversations with patients.”
According to the researchers, when used in a patient-centered manner, this tool can serve as a key component of a proactive approach to IPV intervention, allowing for timely and effective support, ultimately leading to improved long-term health outcomes for at-risk patients. The team has created guidance on the project website to help clinicians approach conversations with patients thoughtfully.
“The goal is never to force disclosure, but to help clinicians communicate with patients in a collaborative way and connect them to resources and support,” Khurana said.
The research team plans to use the AI model to develop a decision support tool embedded in electronic health record systems to provide real-time IPV risk assessment in clinical settings.
For more information about IPV, see About Intimate Partner Violence | Intimate Partner Violence Prevention | CDC
Learn more about automated IPV risk support: https://bhartikhurana.bwh.harvard.edu/airs
This research was co-funded by NIBIB grant R01EB032384 and the NIH Office of the Director.
About the National Institute of Biomedical Imaging and Bioengineering (NIBIB): NIBIB’s mission is to improve health by leading the development and accelerating the application of biomedical technologies. The Institute is committed to integrating the physical and engineering sciences with the life sciences to advance basic research and medicine. NIBIB supports research and development of emerging technologies within its internal laboratories and through grants, collaborations, and training. For more information, please visit the NIBIB website.
About the National Institutes of Health (NIH): The nation’s medical research agency, NIH, has 27 institutes and centers and is part of the U.S. Department of Health and Human Services. NIH is the primary federal agency that conducts and supports basic, clinical, and translational medical research, investigating the causes, treatments, and cures for both common and rare diseases. For more information about NIH and its programs, please visit www.nih.gov.
NIH…Turning discovery into health®
reference
Gu J, Villalobos Carballo K, Ma Y, Bertsimas D, Khurana B. Accurately identify intimate partner violence risk using multimodal machine learning. Nature Portfolio Journal: Women’s Health. 2026.DOI: 10.1038/s44294-025-00126-3
