Many times, depression is difficult to pinpoint, even when it comes to a person you care about, so it should come as no surprise that suicidal thoughts are even harder to recognize. But if they could be foreseen, somehow, wouldn’t that change and save so many lives? We certainly think so. That’s why we were excited to hear about an algorithm that can tell if someone wants to commit suicide #machinemagic
Researchers from Vanderbilt University Medical Center developed a machine learning program that is able to tell if a person is likely to commit suicide. The system was trained on 5,167 patients from the Vanderbilt University Medical Center displaying signs of self-harm and suicidal thoughts. Afterwards, researchers trained and tested it on more than 12,000 randomly selected patients to see if the machine had learned to identify people who are thinking about ending their lives.
During these trials, the algorithm predicted with 80-90% accuracy that a given person will attempt suicide in the next two years. And that’s not all – it was able to tell with an impressive 92% accuracy whether a person would commit suicide in the following week. To get these astonishing results, the machine verifies different data, from age, gender and prior diagnoses to hospital admissions and current or past medication.
These risk factors are supposed to give a glimpse into one person’s probability of giving up on life. It’s not an exact science, as you can imagine – that 8-10% is left to fate, to the unpredictable changes of disposition one might have.
Still, it’s a step in the right direction. If it can alert doctors and relatives in time, many lives could be saved. In fact, the team wants to test their program one more time using patients from another hospital before using their resources for the greater good.