A groundbreaking artificial intelligence (AI) tool developed by researchers funded by the National Institutes of Health (NIH) is poised to transform the way clinicians assess the risk of intimate partner violence (IPV) among patients. By utilizing data routinely collected during medical visits, the team has trained a machine-learning model capable of accurately detecting IPV risk, offering a potentially life-saving decision support tool for healthcare providers.
Intimate partner violence, a serious issue affecting millions of individuals across the United States, often remains hidden due to victims’ fears of disclosure. The abuse, perpetrated by current or former partners, can lead to severe consequences such as life-threatening injuries, chronic pain, and mental health disorders. Despite its prevalence, many cases go undetected, underscoring the need for innovative solutions like this AI tool.
AI Models in Healthcare Settings
The research team, led by experts from Harvard Medical School in Boston, introduced three distinct AI models designed for IPV detection within healthcare environments. The models’ performances were compared to determine the most effective method for predicting IPV risk.
“This clinical decision support tool could make a significant impact on 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 cases, the tool could be a game-changing asset to public health.”
According to the study authors, traditional screening tools capture only a fraction of IPV cases, missing crucial opportunities for intervention. However, clinical and imaging records hold valuable information that can aid in detecting IPV risk. Radiologists, for instance, have an advantage in recognizing patterns of physical trauma indicative of IPV.
Data-Driven Insights and Model Performance
The researchers leveraged several years of hospital data from nearly 850 affected female patients and 5,200 unaffected control patients, matched by age and demographics. Recognizing the variability in clinical data collection across different healthcare settings, the team developed two AI models: one trained on structured data, such as tables, and another on unstructured data from medical notes, including radiology reports. They also created a multimodal model that integrates both data types.
The multimodal fusion model outperformed its counterparts, achieving an impressive accuracy rate of 88%.
This model was able to predict IPV risk more than three years before patients sought help at hospital-based domestic abuse intervention centers. While the tabular model detected risks slightly earlier, the fusion model identified more cases in advance, demonstrating greater stability and accuracy.
The fusion model’s success lies in its ability to process different data modalities separately before merging them at the prediction stage. This approach is particularly relevant in healthcare, where data availability and recording practices vary significantly across institutions.
Implications for Healthcare and Future Developments
The introduction of AI tools like these machine learning models marks a shift from relying on patient self-disclosure to proactive risk recognition. “For decades, our healthcare system has depended largely on patient self-disclosure to identify intimate partner violence, leaving many cases unrecognized and unsupported,” explained Bhati Khurana, M.D., senior author of the study and an emergency radiologist at Mass General Brigham and associate professor of radiology at Harvard Medical School. “Our work represents a fundamental shift from reactive disclosure to proactive risk recognition within routine clinical care.”
By analyzing existing healthcare data patterns, these AI tools empower clinicians to initiate earlier, safer, and more informed conversations with patients. The researchers emphasize that these tools are not intended for definitive diagnoses but to facilitate supportive communication and connect patients with necessary resources.
The research team envisions embedding these AI models within electronic medical record systems to provide real-time IPV risk evaluations in clinical settings. Such integration could enhance the timeliness and effectiveness of IPV interventions, ultimately improving long-term health outcomes for at-risk patients.
For more information on intimate partner violence and prevention strategies, readers can explore resources provided by the Centers for Disease Control and Prevention (CDC) or visit the project website for detailed guidance on automated IPV risk support.
“The goal is never to force disclosure, but to help clinicians communicate with patients in a supportive way and to connect them with resources and support,” Khurana emphasized.
This research, co-funded by an NIBIB grant and the NIH Office of the Director, represents a promising step forward in the fight against intimate partner violence, offering hope for a future where timely intervention and support are readily accessible.
