UCLA Health study suggests hospital emergency departments miss signs of suicidal behavior in many boys, Black and Hispanic children. The study is published in the Journal JAMA Open Network, involving electronic health records of about 3000 children and teens.
The researchers used machine learning algorithms to find that standard medical record surveillance methods fail to identify youth with suicide related emergencies. Despite this, these methods missed suicide related visits of all suicide related visits among Black, Hispanic, male and preteen youths more than they did other races and ethnic groups, female youths, and adolescents.
Existing methods miss kids, and miss them at random, said study Corresponding author Dr. Juliet Edgcomb, assistant professor in residence in UCLA Health Department of Psychiatry and associate director of the UCLA Health Semel Institute for Mental Health Informatics and Data Science Hub.
“It’s difficult for suicide prevention strategies to help those whom they are trying to serve without accurate, and equitable detection of suicide-related emergencies.”
Two factors are used by the standard approach of emergency departments to detect suicidality among youth. The chief concern is the one that the patient or caregiver tells the triage nurse why they came. The second are diagnostic codes used by clinicians to describe and note symptoms and conditions that they see.
Diagnostic codes may include suicidality, but clinicians may code for underlying disorders or symptoms, rather than suicidality itself. ‘It’s transdiagnostic,’ Edgcomb said. Often clinicians code for depression, trauma and use a symptom code like sadness rather than a diagnostic code of suicidality,” O’Leary said.
Edgcomb then tested the effectiveness of this standard approach using electronic health record phenotyping, a method that uses features of someone’s medical record to determine what disease or condition the person may have.
Researchers collected electronic health data from 2,700 children and teenagers 6 to 17 years of age who saw emergency departments at two academic health system campuses between 2017 and 2019. The findings came from an analysis of data from a single health care system, and Edgcomb said more than one emergency department would be needed to validate the study.
The findings suggest the need to improve algorithms for detecting suicide, Edgcomb says, and natural language processing improvements could help by incorporating doctor notes, which artificial intelligence can now process.
Without a diagnostic code or chief concern of suicide, the next clinician is reading medical record notes trying to determine if the child had suicidal thoughts or made an attempt, Edgcomb said. “We can do a lot to improve care by bringing suicidality out into the clinical arena if we can automate that and make it more meaningful, more relevant to the clinician.”
Reference: Edgcomb JB, Loes Olde Loohuis, Tseng C, Klomhaus AM, Choi KR, Ponce CG, et al. Electronic Health Record Phenotyping of Pediatric Suicide-Related Emergency Department Visits.Â
