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

BACKGROUND

The increasing number of acute poisoning (AP) presentations to emergency departments poses a challenge to clinicians. To improve patient care and outcomes, the rapid identification of determinants for poor prognosis that trigger early intervention is key.

OBJECTIVE

This study aims to identify predictive factors for poor outcomes in cases of acute poisoning treated in the emergency department of a tertiary hospital.

METHODS

A retrospective observational study where predictive factors associated with poor outcomes, hospital admission or mortality, were analyzed through medical record reviews was conducted. A multivariable logistic regression model and Chi-squared Automatic Interaction Detection predictive analysis were applied.

RESULTS

Several epidemiological, clinical, and analytical factors were significantly associated with poor outcomes. Female sex, younger age, intentional poisoning and toxic substance responsible for the poisoning being caustics, fumes, or medication were shown to be associated with worse outcomes. Altered mental status, decreased blood pressure, cardiac rhythm and increased body temperature were also predictors of poor outcome. Among the analytical factors studied, altered blood pH, glucose levels and acute-phase reactants were markers for poor outcomes. The proposed parameters for a new early warning system for acute poisoning are R-Respiratory distress, I-Cardiovascular instability, T-Temperature abnormalities, A-Neurological impairment, A-Metabolic acidosis and E-Elevated acute-phase reactants (RITA-AE).

CONCLUSIONS

The findings suggest key parameters for an early warning system for acute poisoning. These factors could aid in early risk stratification and management of AP cases in the ED.