An online friend in Richmond, Virginia with a mutual interest in suicide prevention emailed asking a question about my perspective on using assessment questionnaires to conduct suicide risk assessments.
She shared a post from a listserv called PAUSE (People Addressing and Understanding Suicide Experiences). The PAUSE post used Gemini to explore the question: “How effective is the C-SSRS (aka “the Columbia”) in predicting and eliminating false positives in individual risk assessment situations?”
Gemini’s AI discussion was fascinating—in all the ways you might imagine.
Gemini led with the basic truth that suicide prediction suffers from the statistical “Low base rate problem.” The low frequency (or low base rate) of death by suicide in the general population makes it nearly impossible to predict without having too many false positives. A false positive in suicide assessment is when someone is identified as “high risk” for suicide, but the true risk is not high. Because the U.S. has about 14 suicide deaths per 100,000 people, prediction efforts tend to produce a plethora of false positives, making accurate predictions difficult.
The bigger problem in suicide assessment and screening is the problem of the false negatives. Gemini wasn’t asked about false negatives and didn’t go there. A false negative occurs when the instrument or clinician concludes “low risk” when, in fact, risk is high. This is a bigger problem because false negatives can lead to death whereas false positives lead to overidentification or unnecessary intervention. This bigger problem may be why Craig Bryan advises against using instruments to categorize risk. Because death by suicide is often driven by acute variables (e.g., a relationship break-up, intoxication, job loss, etc.), categorizing anyone as “low risk” based on a screening or assessment will result in too many cases where suicidality is tragically underestimated.
I’d love to dive deeper into questionnaires, actuarial vs. clinical assessment, and the science of predicting human behavior (which was the topic of my Master’s thesis and Doctoral dissertation). However, I’m stopping here because I don’t want to go down the same road as Gemini. . .which is a road with lots of information and statistics and that almost COMPLETELY MISSES THE POINT. What is the point? Suicide assessment questionnaires never prevent suicide.
I should put that statement in ALL BOLD CAPS. To repeat:
SUICIDE ASSESSMENT QUESTIONNAIRES NEVER PREVENT SUICIDE.
Suicide assessment questionnaires are tools that help, but they cannot save lives. Suicide assessments must be used by clinicians; clinicians are the driving force in prevention and saving lives. Here’s another point that deserves bolding and all caps.
ONLY CLINICIANS, FRIENDS, COWORKERS, FAMILY, OR OTHER HUMANS PREVENT SUICIDE AND SAVE LIVES.
As much as I love exploring statistics, science, and prediction all these fascinating concepts cannot replace human connection, trusting relationships, and skills-based interventions. Suicide assessment questionnaires, checklists, and screenings are tools. That’s all.
Since way back in the 1970s, the general rule in the world of psychological assessment has been that actuarial based assessment is a better prediction tool than clinician assessment. This is probably still true, but when we’re talking about human behavior and suicide, the answer is sometimes “it depends.”
So. . . “Which suicide questionnaire is the best for predicting and preventing suicide?” Although I think questionnaire validity and reliability are important, in the case of suicide assessment, the underlying statistics are nearly irrelevant. What’s relevant is whether a specific clinician can use a specific questionnaire to connect with, care for, and effectively a specific person who is suicidal.
Assessment instruments and efforts to accurately predict suicide risk are so intellectually fascinating that we often risk missing the point. Suicide prevention is extremely difficult, suicide prediction is extremely difficult, and the success of prediction and prevention efforts depend almost completely on human interaction.
Thanks for reading.
In case you’re interested—and especially for my Virginia and Alabama friends—this week I’ll be in Virginia doing three presentations at the Virginia Counseling Association conference in Norfolk, VA on Thursday, November 12.
Next week I’ll be at Virginia Commonwealth University on Monday November 16.
Then I’ll be at the Bon Air Library in Richmond doing a suicide-related Lunch N’ Learn on Tuesday, November 17 (see above).
On Thursday, November 19, I’ll be in Mobile, Alabama doing two presentations at the Alabama Counseling Association.
https://www.alabamacounseling.org/uploads/1/2/3/9/123960750/2026_alca_new_brochureflyer.pdf
Big thanks go to Robert Froelich of VCA, Waganesh Zeleke of VCU, Erin Gallagher of Jay Walkers, and Daniel Birdsong of ALCA for organizing all these events.
