To make one of the most sensitive diagnoses, hospitals might do well to call in a computer program. Researchers created a program that can sift through medical records and flag those patients who show signs of domestic violence. While emergency room doctors are always on the lookout for telltale signs, it may be hard to make the diagnosis from one or two isolated incidents. With this program, says lead researcher Ben Reis,
"You are potentially able to detect high abuse risk years ahead of time: you don't wait for a very bad thing to happen" [New Scientist].
Researchers first let the program scan through more than 350,000 medical records which included a substantial number of people who were eventually diagnosed as victims of domestic abuse, and asked the program to look for differences in their medical histories. The computer came up with a collection of risk factors that were highly associated with a future diagnosis. The study, published in the British Medical Journal, found that
women who showed up in emergency rooms with injuries, poisoning, and alcoholism were ... later likely to report domestic abuse. For men, those treated for mental health conditions such as depression and psychosis were the most likely to later also be diagnosed as domestic abuse victims [Reuters].
The program then applied those guidelines to another 180,000 records, and usually noted signs of abuse at an earlier stage in a patient's record than the person's doctor had. Reis says tweaking the program may make it more effective--although it could also lead to more unnecessary doctor-patient confrontations.
Setting the program to tolerate higher rates of "false positives" – instances where the program flags up abuse, even though none was ever detected by a doctor – meant it also caught more cases of real abuse earlier. Under one scenario, a 10 per cent false-positive rate turned up two-thirds of abuses an average of two years before being reported [New Scientist].
Related Content: 80beats: Child Abuse May Leave a Lasting Mark on Victims’ DNAImage: iStockphoto
