This AI tool can detect domestic abuse risk years before victims seek help

Key summary
Study was published in the journal Nature, study published in Nature.
Research dataset used records from nearly 850 women who had experienced intimate partner violence and more than 5,200 control patients, dataset included nearly 850 women with IPV and over 5,200 control patients.
Researchers trained a machine learning model using data collected during regular hospital visits, researchers trained machine learning models using hospital visit data.
AI systems consisted of three models: one using structured hospital data, one using written medical notes, and one combining both, three separate systems used structured data, written notes, and a combined approach.
Combined model correctly identified risk in 88 percent of cases, combined model accuracy reported.
