|
Domestic abuse often goes undiagnosed until too late, yet medical records may contain telltale clues. Now, researchers from the Children’s Hospital Informatics Program (CHIP) and the Division of Emergency Medicine demonstrate that electronic health records could help doctors spot abuse early, when patients don’t reveal it on their own.
The project, led by Ben Reis, PhD, and CHIP colleagues Kenneth Mandl, MD, MPH, and Isaac Kohane, MD, PhD, built a predictive model based on six years of anonymized insurance claims for hospitalizations and Emergency Department visits, involving more than half a million adults. Among the variables associated with domestic abuse were a higher number of annual visits, certain mental health diagnoses and visits for injury.
The sample display above, designed for physicians, which is based on this model, was published in the September 29 issue of the British Medical Journal. Each colored bar represents a diagnosis in the patient’s chart, with the most recent diagnoses at the bottom. Green bars indicate a low degree of abuse risk; yellow, medium risk; red, high risk. As indicated by the blue “detect” arrow, this model would have alerted the doctor to a pattern suggestive of domestic abuse nearly three years before the diagnosis was made. “By providing this safety net, we’re trying to minimize the chances that a high-risk patient might fall through the cracks,” Reis says.
In the future, Reis plans to adapt his model to include child abuse, and to predict abuse before it even occurs. He also wants to expand it to include other conditions that are often missed, like depression and early-stage diabetes.
Read more at intelligenthistories.org.
|
