CHARLESTON, WV | JUNE 13, 2025
Quality Insights is proud to announce that two of its team members, Abubakar Sadiq Bouda Abdulai and Dr. Jean Storm, have co-authored a newly published study in the International Journal of Medical Informatics titled: “I don’t know”: An uncertainty-aware machine learning model for predicting patient disposition at emergency department triage.
This innovative research addresses a critical gap in the application of artificial intelligence (AI) in healthcare, particularly in emergency department (ED) settings. Traditional machine learning (ML) models for predicting patient disposition often return definitive results, even when the model is uncertain. This can lead to overconfident predictions and unintended risks in clinical decision-making.
The study introduces a conformal prediction model designed to indicate when the model is unsure, returning an “I don’t know” output for cases with high uncertainty. This uncertainty-aware approach allows clinicians to understand the confidence level of AI-generated predictions better, improving transparency and promoting safer patient care.
Key highlights from the study include:
- A retrospective analysis of over 560,000 ED visits from three hospital systems.
- Development of a conformalized XGBoost model capable of balancing accuracy with uncertainty.
- Evidence that acknowledging uncertainty leads to fewer misclassifications and more informed clinical decisions.
- At 90% confidence, the model returned “Don’t know” in 34.5% of cases—prioritizing accuracy over forced predictions.
“This study reinforces the idea that transparency in AI systems isn’t just a technical enhancement—it’s a patient safety imperative,” said Dr. Jean Storm, co-author and clinical contributor.
Abubakar Sadiq Bouda Abdulai, lead author and data science specialist at Quality Insights added, “Our goal was to develop a model that reflects the real-world complexity of emergency care. Being able to say ‘I don’t know’ isn’t a weakness—it’s a powerful way to promote clinical judgment and trust in AI tools.”
This publication represents Quality Insights’ continued commitment to driving healthcare improvement through innovation, data analytics, and real-world application of research.
The full article is available online via ScienceDirect.
###
About Quality Insights
Quality Insights is dedicated to improving health and healthcare quality through dynamic partnerships, evidence-based practices, and a patient-first philosophy. With a track record of driving significant improvements in healthcare outcomes, Quality Insights remains at the forefront of quality innovation and healthcare transformation. Learn more at www.qualityinsights.org.
