Using a decision tree algorithm to distinguish between repeated supra-therapeutic and acute acetaminophen exposures

Area of Science:

• Toxicology
• Machine Learning in Medicine
• Clinical Chemistry

Background:

• Acetaminophen overdose is a common cause of acute liver injury.
• Distinguishing between acute and repeated supra-therapeutic ingestion (RSTI) of acetaminophen is crucial for appropriate management.
• Clinical and laboratory features can help differentiate these exposure types.

Purpose of the Study:

• To compare clinical and laboratory characteristics of RSTI and acute acetaminophen exposures.
• To develop and evaluate a predictive decision tree (DT) algorithm for differentiating acetaminophen exposure types.

Main Methods:

• Retrospective cohort study utilizing the National Poison Data System (NPDS).
• Inclusion of 4,522 RSTI acetaminophen exposure cases and 4,522 randomly selected acute ingestion cases (2012-2017).
• Application of a DT machine learning algorithm to classify exposure types.

Main Results:

• The DT model achieved an accuracy, precision, recall, and F1-score of 0.75.
• Patient age was the most significant predictor for differentiating exposure types.
• Serum aminotransferase concentrations, abdominal pain, drowsiness/lethargy, and nausea/vomiting were also important distinguishing factors.

Conclusions:

• DT models show potential in distinguishing between acute and RSTI acetaminophen exposures.
• Further validation is required to confirm the clinical utility of this predictive model.