Secondary triage classification using an ensemble random forest technique

Area of Science:

• Emergency Medicine
• Artificial Intelligence
• Data Science

Background:

• Emergency department triage is complex, involving uncertainty and ambiguity.
• Rapid triage (2-5 minutes) is crucial to prevent fatalities and reduce wait times.
• Human error is a significant concern in emergency triage processes.

Purpose of the Study:

• To develop an intelligent triage system to minimize human error in emergency departments.
• To create a secondary triage model using advanced machine learning techniques.

Main Methods:

• The study employed an ensemble random forest technique for secondary triage modeling.
• A randomized resampling method was utilized to address data imbalance before model development.
• The system was based on the objective primary triage scale (OPTS).

Main Results:

• The random forest model with 300% resampling achieved a low out-of-bag error of 0.02, compared to 0.37 without pre-processing.
• The developed model demonstrated high performance with a sensitivity of 0.98 and specificity of 0.89 on unseen data.
• Resampling effectively balanced the dataset, improving model robustness.

Conclusions:

• The combination of random forest and randomized resampling effectively reduces variance and bias, respectively.
• This intelligent system shows promise for enhancing the accuracy and efficiency of emergency department triage.
• The findings suggest a significant improvement in handling complex triage scenarios.