A Machine Learning-Based Guide for Repeated Laboratory Testing in Pediatric Emergency Departments

Area of Science:

• Pediatric Emergency Medicine
• Clinical Decision Support Systems
• Laboratory Medicine

Background:

• Laboratory tests are sometimes repeated in pediatric emergency departments (PEDs) shortly after community-based testing.
• Unnecessary repeat testing causes child discomfort and increases healthcare costs.
• Minimizing redundant tests is crucial without compromising diagnostic quality.

Purpose of the Study:

• To develop a decision tree (DT) model for minimizing unnecessary repeat blood tests in PEDs.
• To provide physicians with an interpretable tool for guiding repeat testing decisions.
• To optimize healthcare resource utilization in pediatric emergency care.

Main Methods:

• Utilized the minimal decision tree (MDT) algorithm for model development.
• Included children aged 3 months to 18 years with prior community CBC, ELE, and CRP tests.
• Evaluated repeat tests within 12 hours, defining justification by normal-to-abnormal transitions or ≥20% value changes.

Main Results:

• The DT model demonstrated high accuracy in predicting the necessity of repeat CBC and ELE tests.
• The model's performance surpassed that of logistic regression models.
• The DT model was less accurate for predicting repeat C-reactive protein (CRP) tests.

Conclusions:

• A data-driven DT model offers a practical and interpretable guide for clinicians.
• The model aids in reducing unnecessary repeat laboratory testing in PEDs.
• Implementation can enhance patient care and optimize resource allocation.