A person commits suicide every 40 seconds somewhere on the planet. Almost a million die each year for this reason. Suicide is one of the three main causes of death in the 15-44 years old population and the second one in the 10 to 24 years old group. In Chile, the trend is dramatic. Six people kill themselves a day. And it is the second OECD country where the suicide rate has grown the most, surpassed only by South Korea.
“Determining the highest risk groups is key. The studies have focused on analyzing factors such as clinical history and demographic as well as genetic and metabolic influences. Yet, they have had mixed results that do not explain suicidal behavior at particular moments in people’s lives,” explains Jaime Ortega, a researcher at Universidad of Chile Center for Mathematical Modeling (CMM).
“The suicide risk detection was, is and will be a difficult problem to solve. Because the nature of this subject is multi-determinate, it is not possible to associate it with a single fact or cause. And when it happens, the association of variables that explain it is particular to each subject and not linear,” says Jorge Barros, a psychiatrist at Universidad Católica. “You cannot said, as in the case of a person who suffers obesity, is a smoker, hypertensive and sedentary, who could suffer a heart attack. Here the variables are more mobile.”
To reduce this uncertainty, a group led by the CMM mathematician and the UC physician created a predictive model (PDF) that identifies if a patient belongs to the risk group. This makes possible to take measures in the treatment of patients with suicidal risk.
“The therapist’s expertise is vital, but this is a powerful tool for his or her work in order to design treatments and make decisions,” says Ortega.
Comprehensive inquiries
In the study, the team used data mining and machine learning techniques to analyze five questionnaires. These were applied to 707 patients with mood disorders of 14-83 years old, including 349 who attempted suicide before.
The use of these mathematical tools allowed them to analyze 343 clinical and demographic variables. For example, anxiety, depression, interpersonal relations, rabies, reasons for not suicide, family satisfaction, and others.
After testing the model several times, they were able to discover patterns hidden in data that seemed often meaningless. This allowed them to determine the 22 most relevant variables to classify the at-risk and non-risk patients. Two are the most important: levels of personal satisfaction –understood as the perception that oneself is happy, feels satisfied and has achieved personal achievements– and reasons for living.
“This allowed us to generate a quick and easy-to-use tool to understand with 77.9% accuracy if a patient is in the risk group,” adds Ortega, who worked with the student Arnol García (PDF).
Barros already makes some projections: “We are going to test this instrument during the next year. It is an assessment that we will track patients’ distance and online weekly.”
A challenge for the research is to develop more reliable methods to detect and measure the likelihood of imminent suicide. In fact, patients with alcohol, drug, or eating disorders, or psychotic or cognitive problems cannot use this model. It is also likely only a part of those in the risk zone will actually attempt suicide.
However, it is a powerful tool, since it recognizes a particular patient belongs to this area and allows taking timely measures.
“We want to recognize risk from understanding how it has been built. Each person has his or her own constellation of variables and one could move forward with a treatment or psychotherapy. Because this indicator not only says if a patient is at risk, but which variables are more influential in order to see what we can do to promote protective factors and mitigate risk factors,” says Barros. “We would like to go towards personalized medicine. And this is the prolog.”
The research
The study is part of an ongoing research by the Research Group on Depression and Suicidality at UC Department of Psychiatry. The responsible researcher is the psychologist Susana Morales, who was funded by the Fondecyt Initiation Research 11121390. She receives the support of the Innovation Fund for Competitiveness (FIC) of the Ministry of Economy, Development, and Tourism, through the Millennium Scientific Initiative.
