Webb Fellow wins award for her work in leading-edge research
In the week following discharge from medical hospitalization, patients are 10 times more likely to die by suicide compared with matched controls, and half of individuals who complete suicide within three days of hospital discharge were hospitalized for non-psychiatric reasons.
As a result, national suicide prevention initiatives have called for the development of precision algorithms to identify individuals at high risk of suicide.
Working at this leading-edge research is 2018-19 Webb Fellow Juliet Edgcomb, MD, PhD, Department of Psychiatry and Biobehavioral Sciences, UCLA, who has been selected for this year’s ACLP Dlin/Fischer Award, presented for significant achievement in clinical research and the best paper submitted for presentation at the annual meeting.
At CLP 2019 in San Diego this month, Dr. Edgcomb will present her paper, High Risk Phenotypes of Suicidality following Medical Hospitalization.
“After medical hospitalization, people with serious mental illness are at high risk of psychiatric destabilization and emergence of suicidal ideation and behaviour,” she says.
Physical disorders associated with hospitalization, including HIV/AIDS, malignancy, epilepsy, and pain, compound the risk of suicidality.
But precision approaches, utilizing information available in electronic health records to model profiles of patients with rare outcomes (i.e., high risk phenotypes), are resulting in algorithms that demonstrate efficacy in identifying patients at high risk of suicide, and hold promise for enhancing early intervention and prevention.
“My work, conducted as part of the ACLP Webb Fellowship, has involved using electronic health record data to advance prediction of suicide following medical hospitalization,” says Dr. Edgcomb. “Electronic health record data science is an emerging translational field with capacity to identify phenotypes of rare outcomes in mental health and medical care.
“Individuals with co-existing severe mental and medical illness are at high risk of both psychiatric and medical decompensation, resulting in reduction in quality of life for the individual and cost for the health system.
“Statistical learning based-models provide an opportunity to process thousands of heterogeneous variables, and offer many advantages over traditional statistical models for precise diagnostic classification and personalized prediction of risk of negative outcomes, such as readmission and suicidality.
“These learning-based approaches are particularly important for disentangling the complex network of risk factors experienced by individuals with multimorbidity.”
Previously, models of acute psychiatric care utilization had not been adapted to specifically address the needs of individuals who are both medically and psychiatrically ill.
To address this gap, Dr. Edgcomb conducted a series of longitudinal analyses of codified electronic health record data derived from the University of California Health Care System to identify individuals with multimorbid illness and high risk of psychiatric hospital readmission and suicidality following medical hospitalization. The data comprised 16,552 medical hospitalizations of patients with serious mental illness (major depressive disorder, bipolar disorder, or primary psychotic disorder) from 2006-2016. Of these, 287 patients were rehospitalized (5.5% of all patients) for suicide attempt and/or suicidal ideation.
“These analyses utilized classification trees to sort hundreds of potential predictors into a clinically-relevant subset readily interpretable by hospitalist physicians,” says Dr. Edgcomb. “Medical comorbidity and prior care utilization modulated suicidality risk in multimorbid populations.
“These findings enhance prediction of psychiatric decompensation in a seriously medically-ill population, newly bringing predictive analytics to the interface of Psychiatry and hospital care.”
Dr. Edgcomb will present her paper at the annual meeting on Friday, November 15, from 4:00 PM.