Related Experiment Video
Updated: Jul 6, 2025
07:31
Implementation of a Real-Time Psychosis Risk Detection and Alerting System Based on Electronic Health Records using CogStack
Published on: May 15, 2020
7.1K
Predicting quetiapine dose in patients with depression using machine learning techniques based on real-world evidence
Yupei Hao1,2, Jinyuan Zhang3, Jing Yu1,2
1Department of Clinical Pharmacy, the First Hospital of Hebei Medical University, Shijiazhuang, China.
Annals of General Psychiatry
|January 6, 2024
Summary
Machine learning accurately predicts quetiapine doses for depression treatment. This approach identifies key patient factors, aiding clinicians in personalized medication regimens and improving patient outcomes.
Area of Science:
- Pharmacogenomics and Computational Psychiatry
- Application of Machine Learning in Clinical Decision Support
Background:
- Depression is a widespread illness causing significant dysfunction, with many patients not fully responding to standard antidepressant treatments.
- Quetiapine is a commonly prescribed antipsychotic used as an augmentation strategy for depression, but optimal dosing remains challenging.
- Personalized quetiapine treatment regimens are needed to improve efficacy and patient outcomes.
Purpose of the Study:
- To identify key variables influencing quetiapine dosage in depressed patients using real-world data.
- To develop and evaluate machine learning models for predicting optimal quetiapine doses.
- To support clinical decision-making for personalized depression treatment regimens.
Main Methods:
- A univariate analysis was performed on data from 308 hospitalized depressed patients treated with quetiapine.
- Nine machine learning models (XGBoost, LightGBM, RF, GBDT, SVM, LR, ANN, DT) were compared for prediction accuracy.
- The best-performing algorithm was selected to build the quetiapine dose prediction model.
Main Results:
- Four significant predictors of quetiapine dose were identified: quetiapine TDM value, age, mean corpuscular hemoglobin concentration, and total bile acid.
- The XGBoost algorithm achieved the highest predictive performance (accuracy = 0.69) among the nine models evaluated.
- The model demonstrated strong predictive ability in a testing cohort and high AUROC values across different daily quetiapine doses (100-400 mg).
Conclusions:
- Machine learning techniques were successfully applied for the first time to estimate quetiapine dosage in depression.
- The developed predictive model offers valuable support for clinical drug recommendations and personalized treatment strategies.
- This approach has the potential to optimize quetiapine use in managing depression.
