Machine learning assisted differentiation of low acuity patients at dispatch: The MADLAD randomized controlled trial

Area of Science:

• Emergency Medicine
• Artificial Intelligence
• Clinical Decision Support

Background:

• Resource Constrained Situations (RCS) in emergency medical dispatch are common, with more patients needing ambulances than available resources.
• Machine Learning (ML) offers a potential solution for risk assessment among patients in RCS, but requires further validation.

Purpose of the Study:

• To evaluate if ML-based risk scores improve dispatcher accuracy in identifying high-risk patients during RCS.
• To determine if ML tools enhance the dispatch of the first available ambulance to patients most in need.

Main Methods:

• A randomized trial involving 1,245 patients requiring low-priority ambulance response in two Swedish regions.
• Patients were assigned to either an ML-based risk assessment tool or current clinical practice.
• The primary outcome measured was the correct dispatch of the first available ambulance to the patient with the highest National Early Warning Score (NEWS 2).

Main Results:

• The ML intervention arm showed a 68.3% correct assessment rate compared to 62.5% in the control group.
• This resulted in an odds ratio of 1.28 (95% CI [1.00, 1.63], p=0.047) favoring the ML tool.
• The study was limited to low-priority patients and was underpowered due to a smaller than anticipated sample size.

Conclusions:

• Clinical ML-based decision support tools show potential in influencing care provider decisions for rapid risk differentiation.
• Further research is needed to validate these tools in larger, diverse patient populations and settings.
• The trial was registered on ClinicalTrials.gov (NCT04757194).