News

AI can help ICU doctors identify the sickest patients faster

Artificial intelligence could help intensive care teams identify which patients are most at risk of deterioration, allowing hospitals to prioritise care when every minute counts, according to new research led by Australian Catholic University.

Key points:

• Using AI to predict mortality is helpful, if not vital for tailoring treatments, improving care and reducing costs.
• The findings highlight machine learning’s potential to optimise ICU decision-making and support clinicians.
• Professor Shafiabady said the technology was intended to assist clinicians, rather than replacing their expertise.

---

Researchers have developed an explainable artificial intelligence (AI) model that accurately predicts which intensive care patients are at greatest risk of dying in hospital.

Published in BMJ Health & Care Informatics, the study addresses one of the biggest barriers to using AI in healthcare: trust by doctors.

While many AI systems operate as "black boxes", providing predictions without revealing how they were made, the new approach allows doctors to see which clinical factors contributed most to an individual patient's risk score.

Lead researcher Professor Niusha Shafiabady from ACU’s Peter Faber Business School, said the technology is designed to support, not replace, clinical decision-making.

“Intensive care units are often under enormous pressure, and medical staff need every available tool to help identify patients who might need urgent intervention - including AI.

“AI has enormous potential, but clinicians must be able to understand and trust its recommendations."

Using one of the world's largest publicly available intensive care databases, the research compared multiple machine learning models for predicting in-hospital mortality.

The highest-performing model achieved 96.7 per cent accuracy, while explainable AI techniques revealed the key factors influencing each prediction, giving doctors greater confidence in how the system reached its conclusions.

The researchers say explainable AI could eventually become an important support tool by helping ICU teams:

• identify patients at greatest risk earlier
• prioritise monitoring and treatment
• support decision-making during periods of high demand
• improve transparency and trust in AI-assisted healthcare

As Australian hospitals continue exploring the use of AI, the researchers say explainability is essential before AI can be safely integrated into everyday clinical care.

Media Contact: Sally Young, 0467 609 302, [email protected]