Prediction of QTc Prolongation in Acute Poisoning with Atypical Antipsychotics Using Machine Learning Techniques: A...

Area of Science:

• Pharmacology
• Toxicology
• Data Science

Background:

• Atypical antipsychotic use has increased, leading to more acute intoxication cases.
• Predicting adverse outcomes like prolonged QTc interval and need for mechanical ventilation (MV) is crucial.

Purpose of the Study:

• To develop and validate machine learning models for predicting prolonged QTc interval and MV in acute atypical antipsychotic poisoning.
• To assess the performance of various machine learning classifiers, particularly tree-based models.

Main Methods:

• Retrospective study of 355 patients with acute atypical antipsychotic poisoning.
• Development and comparison of eight machine learning classifiers, including Logistic Regression, SVM, KNN, and tree-based models (Random Forest, XGBoost, LightGBM, CatBoost, Gradient Boosting).
• Model validation using external testing datasets and internal five-fold cross-validation with hyperparameter optimization.

Main Results:

• Tree-based models achieved 100% specificity, recall, precision, accuracy, and AUC on the training data for predicting prolonged QTc interval.
• Similar high performance was observed for models predicting the need for MV.
• Upon validation, tree-based models maintained strong AUCs (0.927-0.958) and accuracy (>0.901).

Conclusions:

• Machine learning, especially tree-based models, demonstrates high predictive power for adverse outcomes in acute atypical antipsychotic poisoning.
• The models show promise for clinical decision support, but further validation with larger cohorts is needed for broader generalization.