Friday, March 13, 2026
NIH-funded automated clinical decision support may enable timely interventions for at-risk patients long before they seek help.
Researchers funded by the National Institutes of Health (NIH) have created an innovative artificial intelligence (AI) tool designed to assist clinicians in identifying patients at risk of intimate partner violence (IPV). This tool utilizes data routinely gathered during medical visits, which allowed the researchers to train a highly accurate machine-learning model capable of detecting IPV among study participants.
Intimate partner violence encompasses abuses from current or former partners that can lead to severe consequences such as potentially life-threatening injuries, chronic pain, and mental health disorders. This issue affects millions across the United States, impacting both men and women throughout their lives. Unfortunately, many instances remain unreported, as victims may hesitate to disclose abuse due to safety concerns, fear of repercussions, and societal stigma.
In their study, led by researchers from Harvard Medical School, the team introduced three different AI models aimed at identifying IPV in healthcare settings, assessing each model’s effectiveness in predicting risk.
“This clinical decision support tool could significantly enhance the prediction and prevention of intimate partner violence,” stated Dr. Qi Duan, Ph.D., Director of the Division of Health Informatics Technologies at NIH’s National Institute of Biomedical Imaging and Bioengineering (NIBIB). “Given the prevalence of such cases, the tool could be revolutionary for public health.”
Many instances of IPV go unrecognized, resulting in missed opportunities for early intervention, according to the study authors. They highlighted that existing screening tools only detect a small percentage of cases, while clinical and imaging records can hold valuable information for assessing IPV risk. Notably, radiologists are uniquely positioned to recognize indicators of IPV, including recurring patterns of physical trauma.
The researchers analyzed several years of hospital data from nearly 850 affected female patients alongside 5,200 unaffected patients matched by age and demographics. Due to variations in how clinical data is collected across healthcare settings, they designed two distinct AI models: one based on structured patient data, organized in tabular format, and another using unstructured patient data sourced from medical notes, including radiology reports. Additionally, they developed a multimodal model that synthesizes both structured and unstructured data.
All models demonstrated high performance in the study; however, the multimodal fusion model surpassed those using either structured or unstructured data alone, achieving accurate predictions 88% of the time. Remarkably, both the tabular and fusion models are capable of identifying IPV risk more than three years prior to patients enrolling in hospital-based domestic abuse intervention programs. While the tabular model slightly outperformed in early recognition, the fusion model was better at identifying a greater number of IPV cases in advance.
The fusion model proved to have more consistent performance compared to relying solely on one data type. The researchers clarified that the different data modalities are processed separately and combined only during the prediction phase. They noted that the tabular framework is particularly relevant in healthcare, where data availability and the recording of unstructured information can differ significantly among hospitals.
The team underscored that AI tools such as their machine-learning models can support healthcare providers in having proactive discussions with patients about IPV and guiding them to appropriate resources. Importantly, these tools are not intended for making definitive diagnoses.
“For decades, our healthcare system has relied primarily on patient self-disclosure to identify instances of intimate partner violence, resulting in many cases going unrecognized and unsupported,” commented Dr. Bhati Khurana, senior author of the study and emergency radiologist at Mass General Brigham, as well as an associate professor of radiology at Harvard Medical School. “Our research represents a significant shift from relying on reactive disclosures to proactive risk identification within standard clinical care. By examining patterns already present in healthcare data, this approach empowers clinicians to initiate timely, safer, and more informed conversations with patients.”
The researchers contend that when utilized in a patient-centered manner, this tool could play a crucial role in a proactive strategy for IPV intervention, facilitating timely and effective support, ultimately enhancing long-term health outcomes for at-risk individuals. They have developed guidance available on the project website to aid clinicians in thoughtfully engaging with patients.
“The objective is never to coerce disclosure; rather, we aim to assist clinicians in communicating supportively with patients and linking them to necessary resources,” Dr. Khurana reiterated.
The research team plans to use these AI models to develop a decision-support tool that will be integrated into electronic medical record systems, offering real-time IPV risk assessments in clinical environments.
For additional information on IPV, visit: About Intimate Partner Violence | Intimate Partner Violence Prevention | CDC
For more on 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 enhance health through the advancement and application of biomedical technologies. The Institute is dedicated to integrating physical and engineering sciences with the life sciences to improve both fundamental research and medical care. NIBIB supports emerging technology research and development through its internal laboratories as well as through grants, collaborations, and training opportunities. More information can be found on the NIBIB website.
About the National Institutes of Health (NIH): NIH is the nation’s medical research agency, comprising 27 Institutes and Centers, and operates as part of the U.S. Department of Health and Human Services. NIH is the primary federal entity conducting and supporting research that seeks to understand the causes, treatments, and cures for both common and rare diseases. For further details on NIH and its initiatives, visit www.nih.gov.
NIH…Turning Discovery Into Health®
Reference
Gu J, Villalobos Carballo K, Ma Y, Bertsimas D, and Khurana B. Leveraging multimodal machine learning for accurate risk identification of intimate partner violence. Nature Portfolio Journal: Women’s Health. 2026. DOI: 10.1038/s44294-025-00126-3
