AI Tool Flags Suicide Risk in ER Using Routine Data

Imagine being the only nurse in an emergency room on a Saturday night with 30 patients who need help. Now imagine having a quiet assistant that flags which patients might be silently struggling with thoughts of suicide.

Researchers at the University of Johannesburg just published that assistant. Their AI tool scans routine patient intake information and alerts ER staff when someone might be at risk of suicide-related behavior.

The system doesn't replace nurses or doctors. Instead, it gives overworked medical teams a heads-up so they can have critical conversations with the right patients at the right time.

Lead researcher Tsholofelo Mokheleli saw the problem firsthand in South African hospitals. Late shifts often leave one staff member responsible for triaging dozens of patients. The current mental health screening takes at least 15 minutes per patient, and a full evaluation can take more than a day.

"Not everyone who goes into a medical institution has mental health issues," Mokheleli says. "What we are trying to do is to find those people who do not even say it out loud."

The AI model looks at information already collected during admission: age, vital signs, pain levels, and previous visits to the same hospital. No extra forms. No additional questions. Just data that's already in the system.

When the algorithm spots concerning patterns, it flags the patient and explains why. Maybe they're showing signs of a prior mood disorder, or their vital signs match risk profiles. A healthcare professional still makes every final decision about care.

The Ripple Effect

The approach solves a painful reality: many people at risk of suicide never mention it. Stigma keeps them silent. They might not even recognize what they're feeling. Creating a safe space to talk requires knowing who needs that conversation.

Co-author Nompumelelo Ndlovu, a nursing professor at UJ, points out that some patients don't realize they're struggling until someone asks. "It is not an easy topic to speak about," she says. "There is so much stigma. So how safe do you feel to tell the next person about what is happening with you?"

The system also tackles data privacy concerns that plague many AI health tools. Because it uses records already stored in hospital systems, patient data never leaves the building. No cloud servers. No external vendors. Just local records protecting local patients.

The researchers built the model using historical patient data from the United States, which mirrors the admission information collected in South Africa. They published their findings in Scientific Reports, a peer-reviewed journal.

Right now, the tool works best in hospitals with electronic records. Paper-based systems would need digitization first, which could be a barrier in some facilities.

A small addition to existing hospital systems could help catch thousands of people who might otherwise suffer in silence.