Arturo LoAlza-Bonilla, Co-Founder of Massive Bio, shared on LinkedIn:
“Clinical AI needs a clear plan for when to call the clinician.
Expert attention is one of healthcare’s scarcest resources. How we allocate it deserves the same rigor we apply to model performance.
I’m pleased to share my new paper in the Journal of Clinical Informatics: ‘Calibrated refresh systems for clinical artificial intelligence: A sociotechnical framework for expert escalation.’
Calibrated Refresh Systems (CRS) proposes a framework for deciding when an AI-supported action can proceed and when uncertainty requires deeper expert review. It connects:
- Escalation when uncertainty could change the clinical action;
- Mandatory review for high-risk situations;
- Auditing of accepted decisions, subgroup safeguards, and accountable governance.
The paper explores potential applications in sepsis, clinical deterioration, and oncology trial matching. It presents a conceptual framework and a roadmap for prospective clinical validation.
Looking forward to discussing these ideas with colleagues at St. Luke’s University Health Network and across the clinical AI community, including at:
- AI at Meta,
- American Society of Clinical Oncology (ASCO) AI CoP,
- Massive Bio.
Where could clearer escalation rules make the biggest difference in your practice?”
Title: Calibrated refresh systems for clinical artificial intelligence: A sociotechnical framework for expert escalation
Authors: Arturo Loaiza-Bonilla.
You can also read: Iyad Sultan at COGC 2026: The Hidden Risk – Why a Good AI Model Can Still Fail
