Home > Health > Machine Learning Flags Key Risk Factors for Suicide Attempts

Machine Learning Flags Key Risk Factors for Suicide Attempts

23 Views
Machine Learning Flags Key Risk Factors for Suicide Attempts 2

A history of suicidal behaviors or ideation, functional impairment related to mental health disorders, and socioeconomic disadvantage are the three most important risk factors predicting subsequent suicide attempts, new research suggests.

Investigators applied a machine-learning model to data on over 34,500 adults drawn from a large national survey database. After analyzing more than 2500 survey questions, key areas were identified that yielded the most accurate predictions of who might be at risk for later suicide attempt.

These predictors included experiencing previous suicidal behaviors and ideation or functional impairment because of emotional problems, being at a younger age, having a lower educational achievement, and experiencing a recent financial crisis.

“Our machine learning model confirmed well-known risk factors of suicide attempt, including previous suicidal behavior and depression; and we also identified functional impairment, such as doing activities less carefully or accomplishing less because of emotional problems, as a new important risk,” lead author Angel Garcia de la Garza, PhD candidate in the Department of Biostatistics, Columbia University, New York City, told Medscape Medical News.

“We hope our results provide a novel avenue for future suicide risk assessment,” Garcia de la Garza said.

The findings were published online January 6 in JAMA Psychiatry.

“Rich” Dataset

Previous research using machine learning approaches to study nonfatal suicide attempt prediction has focused on high-risk patients in clinical treatment. However, more than one-third of individuals making nonfatal suicide attempts do not receive mental health treatment, Garcia de la Garza noted.

To gain further insight into predictors of suicide risk in nonclinical populations, the researchers turned to the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC), a longitudinal survey of noninstitutionalized US adults.

“We wanted to extend our understanding of suicide attempt risk factors beyond high-risk clinical populations to the general adult population; and the richness of the NESARC dataset provides a unique opportunity to do so,” Garcia de la Garza said.

The NESARC surveys were conducted in two waves: Wave 1 (2001-2002) and Wave 2 (2004-2005), in which participants self-reported nonfatal suicide attempts in the preceding 3 years since Wave 1.

Assessment of Wave 1 participants was based on the Alcohol Use Disorder and Associated Disabilities Interview Schedule DSM-IV.

“This survey’s extensive assessment instrument contained a detailed evaluation of substance use, psychiatric disorders, and symptoms not routinely available in electronic health records,” Garcia de la Garza noted.

The Wave 1 survey contained 2805 separate questions. From participants’ responses, the investigators derived 180 variables for three categories: past-year, prior-to-past-year, and lifetime mental disorders.

They then identified 2978 factors associated with suicide attempts and used a statistical method called balanced random forest to classify suicide attempts at Wave 2. Each variable was accorded an “importance score” using identified Wave 1 features.

The outcome variable of attempted suicide at any point during the 3 years prior to the Wave 2 interview was defined by combining responses to three Wave 2 questions:

  • In your entire life, did you ever attempt suicide?

  • If yes, how old were you the first time?

  • If the most recent event occurred within the last 3 years, how old were you during the most recent time?

Suicide risk severity was classified into four groups (low, medium, high, and very high) on the basis of the top-performing risk factors.

A statistical model combining survey design and nonresponse weights enabled estimates to be representative of the US population, based on the 2000 census.

Out-of-fold model prediction assessed performance of the model, using area under receiver operator curve (AUC), sensitivity, and specificity.

Daily Functioning

Of all participants, 70.2% (n = 34,653; almost 60% women) completed Wave 2 interviews. The weighted mean ages at Waves 1 and 2 were 45.1 and 48.2 years, respectively.

Of Wave 2 respondents, 0.6% (n = 222) attempted suicide during the preceding 3 years.

Half of those who attempted suicide within the first year were classified as “very high risk,” while 33.2% of those who attempted suicide between the first and second year and 33.3% of those who attempted suicide between the second and third year were classified as “very high risk.”

Among participants who attempted suicide between the third year and follow-up, 16.48% were classified as “very high risk.”

The model accurately captured classification of participants, even across demographic characteristics, such as age, sex, race, and income.

Younger individuals (aged 18 to 36 years) were at higher risk, compared with older individuals. In addition, women were at higher risk than men, white participants were at higher risk than nonwhite participants, and individuals with lower income were at greater risk than those with higher income.

The model found that 1.8% of the US population had a 10% or greater risk of a suicide attempt.

The most important risk factors identified were the three questions about previous suicidal ideation or behavior; three items from the 12-Item Short Form Health Survey (feeling downhearted, doing activities less carefully, or accomplishing less because of emotional problems); younger age; lower educational achievement; and recent financial crisis.

“The clinical assessment of suicide risk typically focuses on acute suicidal symptoms, together with depression, anxiety, substance misuse, and recent stressful events,” co-investigator Mark Olfson, MD, PhD, professor of epidemiology, Columbia University Irving Medical Center, New York City, told Medscape Medical News.

“The new findings suggest that these assessments should also consider emotional problems that interfere with daily functioning,” Olfson said.

Extra Vigilance

Commenting on the study for Medscape Medical News, April C. Foreman, PhD, an executive board member of the American Association of Suicidology, noted that some of the findings were not surprising.

“When discharging a patient from inpatient care, or seeing them in primary care, bring up mental health concerns proactively and ask whether they have ever attempted suicide or harmed themselves — even a long time ago — just as you ask about a family history of heart disease or cancer, or other health issues,” said Foreman, who is also the chief medical officer of the Kevin and Margaret Hines Foundation.

She noted that half of people who die by suicide have a primary care visit within the preceding month.

“Primary care is a great place to get a suicide history and follow the patient with extra vigilance, just as you would with any other risk factors,” Foreman said.

The study was funded by the National Institute on Alcohol Abuse and Alcoholism and its Intramural Program. The study authors and Foreman have reported no relevant financial relationships.

JAMA Psychiatry. Published online January 6, 2021. Abstract

For more Medscape Psychiatry news, join us on Facebook and Twitter.

Source link

TAGS , , , , ,
Hi guys, this is Kimmy, I started LicensetoBlog to help you with the latest updated news about the world with daily updates from all leading news sources. Beside, I love to write about several niches like health, business, finance, travel, automation, parenting and about other useful topics to keep you find the the original information on any particular topic. Hope you will find LicensetoBlog helpful in various ways. Keep blogging and help us grow as a community for internet lovers.