GUIDA della Società

L’intelligenza artificiale nella previsione del rischio di suicidio

AI suicide risk prediction uses statistical and machine learning models built on health records to estimate which patients have a higher chance of a suicide attempt or death in the coming months, so health systems can prioritize outreach and assessment.

Ascolta questa guidaNarrato dall'IA · ~3 min

In questa pagina4 minuti di lettura
  1. Panoramica
  2. Immersione profonda
  3. Implementazione nel mondo reale
  4. The Future of AI in Suicide Risk Prediction
  5. Metterlo in pratica
  6. Continua a esplorare
  7. Domande frequenti

Panoramica

It matters because many people who die by suicide saw a health care provider in the year before, and clinicians' unaided predictions have proven only slightly better than chance. Even strong models flag far more people who will not attempt suicide than people who will.

Immersione profonda

Most health system models draw on data already in the electronic health record: prior suicide attempts and self-harm diagnoses, mental health and substance use diagnoses, prescriptions, emergency visits and inpatient stays, and screening answers such as item 9 of the PHQ-9, which asks about thoughts of death or self-harm. Methods range from penalized logistic regression to random forests and gradient-boosted trees. Well-known efforts include models from the Mental Health Research Network, which includes Kaiser Permanente sites, work at Vanderbilt University Medical Center, and the VA's REACH VET, launched in 2017, which flags roughly the top 0.1 percent of risk scores for review. Several large studies report good discrimination, meaning the models rank higher-risk patients above lower-risk ones reasonably well. The main misconception is that these models predict who will die. They do not. Suicide is rare in any given month, so even an accurate model produces mostly false positives; most flagged patients will never attempt. The models are best understood as tools for prioritizing limited clinical attention. That makes the response to a flag the real ethical question. Good practice treats a flag as a reason for caring, voluntary outreach: a check-in, a review of the care plan, safety planning, and discussing limits on access to lethal means such as safe firearm storage. Poor practice uses flags punitively or coercively, shares them outside clinical care, or leaves them unexplained to patients. Health systems also need to test performance across racial, ethnic, age and sex groups, because a model trained on past records inherits gaps in who was diagnosed and treated. Outside health care, platforms such as Facebook have used automated detection of suicidal posts since 2017, with much less public validation.

Implementazione nel mondo reale

The US Department of Veterans Affairs' REACH VET program scores veterans in VA care each month, and a local coordinator reviews the highest-risk group with their providers to check care plans and arrange contact.

A health system shows a risk score at outpatient mental health visits, prompting the clinician to run a structured assessment such as the Columbia Suicide Severity Rating Scale instead of relying on the score alone.

An emergency department uses a model flag to trigger a safety planning intervention before discharge, followed by caring contacts such as brief follow-up messages.

A health system audits its model, finds it performs worse for one demographic group, and recalibrates it before expanding its use.

The Future of AI in Suicide Risk Prediction

Researchers are exploring richer inputs, such as clinical notes processed with language models, alongside traditional structured data, though more inputs raise privacy concerns and do not guarantee better outcomes. The central open question is less about accuracy than about impact: whether flag-driven programs reduce suicide attempts and deaths compared with good usual care. Expect more emphasis on fairness audits, transparency with patients and pairing models with proven interventions such as safety planning and follow-up contacts. Claims of dramatic accuracy gains should be judged by prospective results, not retrospective ones.

Metterlo in pratica

Perché è importante

  • Rischio e sicurezza. I danni catastrofici e quotidiani dell’IA dipendono entrambi da chi comprende i rischi e da chi può agire.
  • Decisioni più chiare. L’alfabetizzazione pubblica e professionale determina la possibilità politica di una forte politica di sicurezza.
  • Tagliare il clamore. Spiegazioni chiare riducono la cattura da parte di montature pubblicitarie, PR di laboratorio e vaghi teatrini etici.

Attenzione

  • Trattare il rischio esistenziale come fantascienza mentre le capacità si aggravano.
  • Confondere la sicurezza del prodotto superficiale con l'allineamento in condizioni di elevata autonomia.
  • Lasciando il pubblico non inglese e non esperto solo con fonti di bassa qualità.

Prossimi passi

  1. Separare i rischi di danni al prodotto, uso improprio e perdita di controllo/disallineamento.
  2. Chiedi quali prove cambierebbero la tua opinione sulle tempistiche e sulla gravità.
  3. Preferire fonti primarie e valutazioni concrete alle affermazioni di marketing.
  4. Identifica un percorso d’azione: carriera, politica, finanziamenti o competenze, non solo consapevolezza.

Continua a esplorare

Newsletter gratuita

Ricevi il briefing quotidiano sull'IA

Tre storie verificate sull'IA ogni mattina nei giorni feriali, scritte in un inglese semplice. Gratuito per sempre, senza pubblicità.

Un'email ogni giorno feriale. Disiscriviti con un clic. Non vendiamo né condividiamo mai il tuo indirizzo.

Mettiti alla prova

Take the AI in Suicide Risk Prediction quiz

Feedback immediato su ogni risposta e un certificato condivisibile con un ID verificabile una volta superato un corso.

Inizia il quiz

Sostieni l'istruzione gratuita sull'IA. AI Understanding è un'organizzazione non profit 501(c)(3) — nessun annuncio, nessun paywall, mai. Fai una donazione

Domande frequenti

What is AI in Suicide Risk Prediction?

AI suicide risk prediction uses statistical and machine learning models built on health records to estimate which patients have a higher chance of a suicide attempt or death in the coming months, so health systems can prioritize outreach and assessment. It matters because many people who die by suicide saw a health care provider in the year before, and clinicians' unaided predictions have proven only slightly better than chance. Even strong models flag far more people who will not attempt suicide than people who will.

A model with 90 percent sensitivity and specificity is applied where 100 of 100,000 patients attempt suicide. Roughly what share of flagged patients will actually attempt?

It catches about 90 true cases but also flags about 9,990 people who will not attempt, so positive predictive value falls below 1 percent.

According to the guide, what are suicide risk models best understood as?

Because suicide is rare, even accurate models mostly flag people who will not attempt. They help decide where to focus outreach, not who will die.

Roughly what share of risk scores does the VA's REACH VET program flag for review?

REACH VET, launched in 2017, flags roughly the top 0.1 percent of scores for review by a local coordinator and providers.

Why do health systems often set flagging thresholds by capacity?

A flag is only useful if someone can act on it, so thresholds are matched to the outreach staff available.

Which response to a risk flag fits the ethical practice described in the guide?

Good practice treats a flag as a reason for caring, voluntary contact, including means safety, rather than punitive or coercive action.