Predicting Avoidable Emergency Department Visits Using the NHAMCS Dataset.

Area of Science:

• Health Services Research
• Predictive Modeling
• Emergency Medicine

Background:

• Avoidable emergency department (ED) visits significantly impact healthcare costs and quality of care.
• Limited research exists on predictive models for identifying patients at high risk for avoidable ED visits.

Purpose of the Study:

• To develop and evaluate predictive models for identifying patients likely to present with avoidable ED visits.
• To define 'avoidable' ED visits conservatively as those not requiring diagnostic services, procedures, or medications, with patients discharged home.

Main Methods:

• Utilized a large dataset from emergency departments across the United States.
• Trained and tested predictive models using a conservative definition of avoidable ED visits.
• Evaluated model performance across various demographic groups, including race, gender, and insurance status.

Main Results:

• Models achieved a training AUC of 0.723 and a testing AUC of 0.703.
• Performance was comparable between white/black patients and male/female groups.
• Reduced performance was observed in Hispanic populations and patients with Medicaid.
• Predictors for non-avoidable visits included increased age, chronic diseases, and digestive symptoms.
• Injuries and psychiatric symptoms were associated with avoidable ED visits.

Conclusions:

• Predictive models can identify avoidable ED visits with moderate accuracy.
• Demographic factors influence model performance, highlighting potential disparities.
• Understanding key predictors can inform targeted interventions to reduce unnecessary ED utilization.