Researchers at Mass General Brigham have introduced a series of artificial intelligence (AI) tools aimed at identifying individuals at risk for intimate partner violence (IPV) by analyzing data from their electronic medical records (EMRs). A study published in npj Women’s Health reveals that these tools could detect IPV up to four years before a patient seeks help from a domestic violence treatment center. This development underscores the potential for proactive screening and aids healthcare providers in initiating crucial conversations about IPV with their patients.
“Our research offers proof of concept that AI can support clinicians in flagging possible abuse earlier. Earlier identification of intimate partner violence and future risk may enable clinicians to intervene sooner and help prevent significant mental and physical health consequences,”
Bharti Khurana, MD, MBA, principal investigator, corresponding and senior author, founding director of the Trauma Imaging Research and Innovation Center and emergency radiologist, Mass General Brigham Department of Radiology
Statistics reveal that more than one-third of women and one in ten men will experience IPV during their lifetimes. Despite this alarming prevalence, many individuals fail to disclose their experiences to healthcare providers due to fear, stigma, or financial and psychosocial dependence on their abuser. Previous studies have indicated that victims of IPV are more likely to reveal their situations when asked in a private and trauma-informed manner by a trusted health provider.
To facilitate earlier identification and intervention by health providers, Khurana’s team collaborated with experts from the Massachusetts Institute of Technology (MIT), led by Dimitris Bertsimas, PhD. Together, they utilized machine learning to train three distinct models using EMR data from 673 women who sought help at a domestic abuse intervention center in the United States between 2017 and 2022. The team also included 4,169 demographically matched controls who did not report IPV.
The research examined three AI models: a tabular model that utilized structured EMR data such as diagnoses and medications; a notes model that leveraged unstructured clinical notes, as well as radiology and emergency department reports; and a fusion model called Holistic AI in Medicine (HAIM), which combined both data types. When tested on a cohort of 168 patients who visited the IPV intervention center and 1,043 controls, all three models demonstrated high accuracy, with the fusion model achieving an impressive accuracy rate of 88%. Furthermore, when evaluated against time-stamped, archived medical records, the fusion model was able to predict 80.5% of IPV cases on average 3.7 years before patients sought care.
The models were subsequently validated using data from two additional patient groups not included in the initial training or testing datasets. This validation also resulted in similarly high accuracy rates, reinforcing the reliability of the AI tools.
Further investigations led by Khurana revealed that women undergoing frequent imaging studies in emergency departments and presenting specific types of injuries are more likely to report IPV later on. This latest AI research identified additional risk factors, indicating that individuals with mental health disorders, chronic pain, and frequent emergency department visits face an elevated risk for IPV. In contrast, patients who regularly accessed preventive services, such as mammograms and immunizations, were found to have a lower risk.
However, the authors caution that the AI tools were developed and validated using data from individuals who had already sought care for IPV. This focus may limit the models’ accuracy in predicting IPV among individuals less likely to disclose their experiences to healthcare providers. Furthermore, the control group may have included false negatives—patients experiencing IPV but not reporting it, which could potentially diminish the models’ overall accuracy. Khurana emphasizes the need for future training on larger and more diverse patient datasets over extended periods to enhance the models’ predictive capabilities.
As healthcare systems continue to grapple with the implications of IPV, the introduction of these AI tools represents a significant step towards earlier intervention and support for victims, ultimately aiming to alleviate the long-term mental and physical health consequences associated with intimate partner violence.
See also
AI Transforms Health Care Workflows, Elevating Patient Care and Outcomes
Tamil Nadu’s Anbil Mahesh Seeks Exemption for In-Service Teachers from TET Requirements
Top AI Note-Taking Apps of 2026: Boost Productivity with 95% Accurate Transcriptions
Machine Learning Tools Transform Materials Discovery, Enhancing Semiconductor Efficiency
