Suicide risk prediction tools fail racialized patients

0
382


Current models used to predict suicide the risk falls short for racialized populations, including blacks, indigenous peoples, and people of color (COPD), according to new research.

The researchers developed two models of suicide prediction to examine whether such tools are accurate in their predictive skills or whether they have deficiencies.

They found that both prediction models could not identify high-risk COPD individuals. In the first model, nearly half of the outpatient visits followed by suicide were identified in white patients versus only 7% of the visits followed by suicide in patients with COPD. The second model had a sensitivity of 41% for white patients, but only 3% for black patients and 7% for Native American / Alaska Indian patients.

“You don’t know if a prediction model will be useful or harmful until it’s evaluated. The message we bring home from our study is this: you have to look,” said lead author Yates Coley, PhD, associate researcher of Kaiser Permanente Washington Health Said the Seattle Research Institute Medscape Medical News.

The study was published online April 28 a Psychiatry JAMA.

Racial inequalities

Suicide risk prediction models have been “developed and validated in a variety of settings” and are now regularly used in the Veterans Health Administration, HealthPartners and Kaiser Permanente, the authors write.

But the performance of suicide risk prediction models, while accurate in the general population, “remains unexamined” in particular underpopulations, they point out.

“Health record data reflect existing ethnic and racial inequalities in health access, quality, and outcomes; and prediction models that use health record data can perpetuate these disparities by assuming that past health patterns accurately reflect real needs, ”Coley said.

Coley and his team “wanted to make sure that any suicide prediction model we implemented in clinical care reduced health disparities rather than exacerbated them.”

To investigate, the researchers examined all outpatient mental health visits to seven major integrated health systems for patients 13 years of age and older (n = 13,980,570 visits from 1,422,534 patients; 64% women, mean) [SD] 42 years [18] years). The study covered from January 1, 2009 to September 30, 2017 with a follow-up to December 31, 2017.

In particular, the researchers examined suicides that took place within 90 days of the outpatient visit.

The researchers used two prediction models: logistic regression with the selection of LASSO variables (Least Absrute Shrinkage and Operator Operator) and the random forestry technique, a “tree-based method that explores interactions between predictors (including those with race and ethnicity) to estimate the probability of an outcome. “

Models considered predetermined interactions between predictors, including previous diagnoses, suicide attempts, and PHQ-9. [Patient Health Questionnaire-9] responses and data of race and ethnicity.

The researchers evaluated the performance of the prediction models in the overall validation set and within subgroups defined by race / ethnicity.

The area under the curve (AUC) measured model discrimination and sensitivity was estimated for global and race / ethnic specific thresholds.

Unacceptable Stage

Within the total population, there were 768 deaths from suicide in the 90 days following 3143 visits. Suicide rates were higher on visits from patients without a registered race / ethnicity, followed by visits from Asian, white, American Indian / Alaska Native, Hispanic, and black patients (Table 1).

Table 1. Suicide rates by ethnicity

Ethnicity External visits were followed
for suicide within 90 days, n
Suicide rate
for every 10,000 visits
Not registered 313 5.71
Asian 187 2.99
White 2134 2.65
American Indians / Alaska Natives 21 2.18
Hispanic 392 1.18
Black 65 .56

Both models showed “high” AUC sensitivity for white, Hispanic, and Asian patients, but “low” AUC sensitivity for COPD and patients without registered race / ethnicity, according to the authors (Table 2).

Table 2. Sensitivity of the area under the curve (AUC) for preaching models

Ethnicity, model AUC ≥ 95% sensitivity
Complete validation set

Logistic regression
Random forest

.822
.816

41.1%
38.0%
White

Logistic regression
Random forest

.828
.812

46.8%
40.6%
Hispanic

Logistic regression
Random forest

.855
.831

36.8%
38.2%
Black

Logistic regression
Random forest

.775
.786

6.7%
3.3%
Asian

Logistic regression
Random forest

.834
.882

31.8%
60.0%
American Indians / Alaska Natives

Logistic regression
Random forest

.599
.642

6.7%
6.7%
Not registered

Logistic regression
Random forest

.640
.676

23.4%
20.4%


“The implementation of prediction models must be considered in a broader context of unmet health needs,” Coley said.

“In our specific example of suicide prediction, BIPOC populations already face substantial barriers to accessing quality mental assistance and, as a result, have poorer outcomes and the use of any of the prediction models. of suicide examined in our study will provide fewer benefits to populations that are no longer served. and widen existing care gaps “: a scenario that Coley said is” unacceptable. “

“We must insist that new technologies and methods be used to reduce racial and ethnic inequalities in care, not to exacerbate them,” he added.

Biased algorithms

Commenting on the study for Medscape Medical News, Jonathan Singer, PhD, LCSW, Associate Professor, School of Social Work, Loyola University, Chicago, Illinois, described it as an “important contribution because it points to a systemic problem and also to the fact that the algorithms we create are biased, created by humans and humans are biased. “

Although the study focused on the health care system, Singer believes the findings have implications for individual physicians.

“If doctors may be biased against identifying suicide risk in black and Native American patients, they may attribute suicide risk to something else. For example, we know that in black Americans expressions of intense emotions are often interpreted. how aggression or be threatening, as opposed to indicators of sadness or fear, “said Singer, who is also president of the American Academy of Suicide and did not participate in the study.

“Clinicians who misinterpret these intense emotions are less likely to identify a black client or suicidal patient,” Singer said.

The research was supported by the Mental Health Research Network of the National Institute of Mental Health. Coley reported that he has received support through a grant from the Agency for Health Research and Quality. The article details the disclosures for the other authors.
Singer has not reported any relevant financial relationships.

Psychiatry JAMA. Published online April 28, 2021. Summary

For more Medscape Psychiatry news, join us Facebook i Twitter.





Source link