Thanks to a five-year, $2.7 million grant from the National Institute of Mental Health, researchers at Vanderbilt University Medical Center (VUMC) will plan to use artificial intelligence (AI) to predict suicide thoughts and attempts in patients.
Specifically, the researchers will use computational methods to shed light on suicidal ideation and its relationship to attempted suicide, predict suicidal ideation and suicide attempt using routine electronic health records (EHRs) and explore the genetic underpinnings of both, officials at VUMC noted.
They pointed out that from 1999 to 2017, the all-ages suicide rate in the U.S. increased 33 percent, from 10.5 to 14.0 per 100,000 population. In 2017 there were 47,173 recorded suicides, making it the nation’s 10th leading cause of death.
The principal investigators for this study were internist and clinical informatician Colin Walsh, M.D., assistant professor of Biomedical Informatics, Medicine, and Psychiatry and Behavioral Sciences, and geneticist and computational biologist Douglas Ruderfer, Ph.D., assistant professor of Medicine, Psychiatry and Behavioral Sciences, and Biomedical Informatics.
In previous work Walsh and colleagues used EHR data and machine learning techniques to develop predictive algorithms for attempted suicide. Walsh and Ruderfer have used this algorithm to assign suicide risk scores to thousands of genotyped patients, thus gaining statistical power to explore the genetic underpinnings of attempted suicide, according to researchers.
What’s more, their previous work relied strictly on structured data such as healthcare billing codes. But this time they’ll work with Cosmin Bejan, Ph.D., assistant professor of Biomedical Informatics, to include EHR text in their analysis, which might help both to identify and to predict suicidal ideation and suicide attempt.
The study will also include data from two major biobanks: Vanderbilt’s BioVU, the world’s largest collection of human DNA stored at a single site, and UK Biobank. De-identified data from BioVU are linked to de-identified health records. Using data from UK Biobank, colleagues at Stanford University will run genetic analyses of suicidal ideation risk in a second population.
“In this study, we tackle two key challenges. First, we know that many cases of suicidal thoughts and behaviors are missed if we rely on structured data alone. In one study, suicidal thoughts were only coded 3 percent of the time even when documented in text in primary care. Second, text features extracted through natural language processing of physician notes, patient messages and more should allow us to improve our predictive algorithms by capturing more complete and nuanced risk factors,” Walsh said.
Ruderfer added, “Expanding our capture of suicidal phenotypes will enable improved understanding of the genetic architecture of suicidal thoughts and behaviors and the contributing genetic and clinical risk factors. While most individuals who consider suicide do not attempt it, we know ideation is an important risk factor and if identified could provide a flag for patients whom we might be able to help.”
