Predictive Model for Poor Outcomes in Acute Poisonings Using Chi-Squared Automatic Interaction Detection Analysis

Area of Science:

• Clinical Toxicology
• Emergency Medicine
• Medical Informatics

Background:

• Acute poisoning (AP) presents a significant challenge to emergency departments (EDs).
• Early identification of prognostic determinants is vital for timely intervention and improved patient outcomes.
• Effective risk stratification for AP cases remains an ongoing clinical need.

Purpose of the Study:

• To identify predictive factors associated with poor outcomes in acute poisoning cases managed in a tertiary hospital's emergency department.
• To establish key parameters for a novel early warning system for acute poisoning.

Main Methods:

• Retrospective observational study analyzing medical records of acute poisoning patients.
• Multivariable logistic regression and Chi-squared Automatic Interaction Detection (CHAID) were employed for predictive analysis.
• Factors analyzed included epidemiological, clinical, and laboratory data.

Main Results:

• Female sex, younger age, intentional poisoning, and exposure to caustics, fumes, or medications were linked to adverse outcomes.
• Clinical predictors of poor outcomes included altered mental status, hypotension, cardiac arrhythmias, and hyperthermia.
• Laboratory markers such as altered blood pH, glucose levels, and elevated acute-phase reactants indicated poor prognosis.
• The RITA-AE (Respiratory distress, Cardiovascular instability, Temperature abnormalities, Neurological impairment, Metabolic acidosis, Elevated acute-phase reactants) score was proposed as an early warning system.

Conclusions:

• Key parameters for an early warning system for acute poisoning have been identified.
• These findings can facilitate early risk stratification and enhance the management of acute poisoning cases in the ED.
• The proposed RITA-AE system offers a potential tool for improving clinical decision-making in AP management.