Predicting the Aftershock

Can an algorithm help identify those most at risk of PTSD?

By Dalmeet Singh Chawla, published January 2, 2018 - last reviewed on April 17, 2018


Large-scale disasters produce ripple effects well beyond the physical damage they cause. Around a quarter of those who endure earthquakes, for instance, later develop post-traumatic stress disorder (PTSD). "The question of who needs help early on is highly important for providing effective care in the aftermath," says Yuval Neria, a medical psychologist at Columbia University. But clinicians could use help in making such assessments. Can a digital tool predict how likely a person is to end up with clinical symptoms?

That's the possibility explored by a report in the Journal of Psychiatric Research. Using surveys of Chileans before and after a major earthquake in 2010, Boston University psychologist Anthony Rosellini and colleagues developed an algorithm for gauging PTSD risk based on 67 factors, ranging from place of residence to pre-disaster health. This "super learner" algorithm, which integrates multiple statistical approaches to prediction, significantly outperformed existing risk-assessment tools. 

More work needs to be done to validate the method before it can be deployed widely. In the future, Rosellini says, it could be built into a smartphone app: Relief workers might collect information from victims, feed it into the app, and receive a score estimating a person's risk level, as well as a sense of what degree of intervention might be called for.