Algorithms identifying low-acuity emergency department visits: A review and validation study.

Area of Science:

• Health Services Research
• Emergency Medicine
• Health Informatics

Background:

• Identifying low-acuity emergency department (ED) visits is crucial for optimizing resource allocation and improving patient flow.
• Algorithms using International Classification of Disease (ICD) codes are frequently employed to categorize ED visit acuity.
• The variability and accuracy of these ICD-code-based algorithms require thorough investigation.

Purpose of the Study:

• To systematically characterize and validate the performance of existing algorithms that utilize ICD codes for identifying low-acuity ED visits.
• To assess the congruence and predictive accuracy of these algorithms against a defined low-acuity composite (LAC) reference standard.

Main Methods:

• A systematic literature search identified 15 papers detailing ICD-code-based algorithms for low-acuity ED visits.
• Algorithms were applied to National Hospital Ambulatory Medical Care Survey (NHAMCS) data (2011-2017).
• Algorithm agreement was measured using the Jaccard index, and performance was evaluated using positive predictive value (PPV) and sensitivity against LAC criteria.

Main Results:

• Fifteen algorithms were identified, with significant variation in ICD code inclusion (968 codes total) and pairwise Jaccard similarity (mean 0.20).
• When validated against the LAC criteria, algorithms demonstrated a mean PPV of 0.308 and sensitivity of 0.183.
• A small proportion (2.1%) of visits identified as low-acuity by these algorithms experienced severe outcomes (e.g., death, critical care needs).

Conclusions:

• Current algorithms for identifying low-acuity ED visits using ICD codes exhibit poor congruence and are unreliable predictors of actual visit acuity.
• The findings highlight the need for developing more robust and validated methods for acuity assessment in emergency departments.