A study funded by the National Institutes of Health has developed an AI tool offering clinical decision support to clinicians by predicting patients at risk of intimate partner violence (IPV) from data collected during medical visits.
Led by researchers from Harvard Medical School, the study trained a machine learning model using several years of hospital data from around 850 female patients affected by IPV, along with 5,200 control patients matched on age and demographics.
Due to differences in the way clinical data is collected across healthcare settings, two separate AI models were designed. One was trained using structured patient data, whilst another was trained using unstructured patient data from medical notes including radiology reports. A further multimodal model was then developed fusing both structured and unstructured data.
Clinical features such as mental health, chest pain, and painkiller use, were correlated with greater likelihood of IPV, researchers highlight, along with social factors including high levels of social deprivation. “Aside from capturing social factors and trauma-related health outcomes, we note that the model captures injury and healthcare utilisation patterns related to IPV,” they observe. “Notably, we observe a correlation between the high utilisation of radiology tests for the upper extremity and unclassified locations (often seen in ED setting) and a higher probability of IPV presence, consistent with previous research findings.”
All three models were reported as achieving “high performance”, but the multimodal model was the most successful, performing accurately in 88 percent of cases, and detecting IPV on average more than three years before patients seek support.
“The fusion model achieved more stable performance than relying on either modality alone,” National Institutes of Health explains. “The scientists explained that the different modalities are processed separately and only merged at the prediction stage. They found that the tabular framework is particularly relevant in healthcare, where there are variations across different hospitals in data availability and in the recording of unstructured data.”
Whilst not intended for making “definitive diagnoses”, researchers suggest the tool could be used in a proactive approach to IPV intervention, improving long-term health outcomes for at-risk patients. Guidance for clinicians on approaching conversations with patients on IPV is included on the project’s website.
Next steps include the development of a decision support tool to be embedded within EHRs offering real-time IPV risk evaluations in clinical settings, National Institutes of Health shares.
Gu, J., Carballo, K.V., Ma, Y. et al. Leveraging multimodal machine learning for accurate risk identification of intimate partner violence. npj Womens Health 4, 15 (2026). https://doi.org/10.1038/s44294-025-00126-3
Wider trend: AI in health and care
HTN was joined for a deep dive into AI strategy, implementation, adoption, and opportunities by Neill Crump, group associate director of innovation & partnerships at The Dudley Group and Sandwell and West Birmingham, and Pip Hodgson, group digital transformation specialist at University Hospitals of Leicester (UHL) and Northamptonshire (UHN). Our panel discussed their organisation’s approaches to AI and AI strategy, best practices in AI strategy development, Ambient Voice Technology and successful implementation, and the opportunities likely to be ahead with the next wave of AI.
Ireland’s first national AI strategy has been published, outlining opportunities for AI deployment over the next five years, and setting out four strategic pillars covering AI for clinical care, operations, research and innovation, and public health. The government commits to using AI to enhance quality of care, empower clinicians, improve system efficiencies, shorten waiting times, streamline patient pathways, develop targeted insights, and deliver more targeted prevention. The needs of patients and the public will be a guiding principle, along with transparency and trustworthiness, and keeping a human-in-the-loop.
Moorfields spin-out Cascader has announced its partnership with Specsavers focused on harnessing the potential of AI innovation to improve patient care in optometry. Cascader, a spin-out from Moorfields, UCL and Topcon Health, is focused on building clinical-grade AI for ophthalmology. A mission statement from its website outlines its work to use AI “to enable safe, evidence-based decisions in high-volume, high-risk eye conditions” and to use oculomics for early detection of systemic disease.
