StoriesAI Infrastructure

St. Michael’s team secures $200k for early ICU deterioration AI; deployment is pending

The tool targets imperceptible patterns in minute-by-minute ICU data, aiming to give physicians more lead time than current alarms allow.

AI Compass Team (St. Michael's Hospital)Toronto, CanadaSources & technical notes
St. Michael’s team secures $200k for early ICU deterioration AI; deployment is pending
What matters

Key Takeaway

The tool targets imperceptible patterns in minute-by-minute ICU data, aiming to give physicians more lead time than current alarms allow.

Impacted Sectors

Primary sector: AI Infrastructure

Next Steps / Actionable Advice

Whether the team successfully completes the retrospective study using historical patient data, demonstrating that AI Compass would have identified critical deterioration earlier than current systems.

Dr. Bijan Teja and Dr. Michael Sklar of St. Michael’s Hospital are developing AI Compass, an early-stage clinical tool designed to predict critical patient deterioration inside the Intensive Care Unit (ICU). The project stems from a fundamental gap in care: when existing hospital alarms only ring after a patient's condition has become severe, as seen in the case of a recovered surgery patient who suddenly declined.

AI Compass is engineered to analyze vast amounts of granular data collected every minute, including ventilator settings and vital signs, from tens of thousands of ICU patients. This level of continuous monitoring allows the AI to identify subtle patterns that human observation or current warning systems might miss, potentially alerting doctors before a patient takes a turn for the worse.

While St. Michael’s currently uses ChartWatch, a similar early-warning system in general wards, AI Compass is distinct: it specifically predicts deterioration within the ICU environment, where data volume and complexity are significantly higher than in standard medical wards. The project has secured $200,000 via the Mitchell Award through the St. Michael’s Hospital Foundation’s Angel's Den competition.

This initial funding is intended to buy the multidisciplinary team time to train its AI model and move toward a proof-of-concept stage. Teja stressed that this award is only a starting point, as securing further support will be necessary for widespread adoption across Canadian ICUs. The ultimate goal remains retrospective validation: determining if AI Compass could have caught patients like Bob earlier.

The team's focus is currently on validating the model against St. Michael’s historical patient data, ensuring that the predicted patterns hold up when tested retrospectively before any real-time deployment can occur.

Sources & technical notes

Linked sources

Check the cited material to distinguish announcements from demonstrated results.

Evidence limit: The tool has secured only initial seed funding ($200k) and remains in a pre-deployment, proof-of-concept stage; performance claims are attributed to the team's hypothesis.

What to evaluate next

This box highlights the systems, workflows, and decisions the article helps you assess.

  • The project's immediate success depends on completing a retrospective validation study at St. Michael’s to prove the AI can identify critical events missed by human observation or existing monitoring tools.
  • AI Compass is engineered to analyze vast amounts of granular data collected every minute, including ventilator settings and vital signs, from tens of thousands of ICU patients.
  • Operational lens: ICU Patient Deterioration Prediction AI
AI