Predicting acupuncture efficacy for major depressive disorder using baseline clinical variables: A machine learning...

Area of Science:

• Integrative Medicine
• Computational Psychiatry
• Clinical Neuroscience

Background:

• Acupuncture shows promise for treating major depressive disorder (MDD), but patient response varies.
• Predicting individual treatment efficacy is crucial for optimizing care.

Purpose of the Study:

• To develop and validate a machine learning model for predicting acupuncture treatment efficacy in MDD patients.
• To identify key baseline clinical variables that predict treatment response.

Main Methods:

• Utilized data from 124 MDD patients across five research centers undergoing 6-week acupuncture treatment.
• Employed max-relevance and min-redundancy (mRMR) and Pearson correlation for feature selection from 26 variables.
• Compared five machine learning models (logistic regression, SVM, KNN, random forest, XgBoost) for predictive performance using Hamilton Depression Scale-17 (HAMD-17) scores.

Main Results:

• The XgBoost model achieved the highest performance with an AUC of 0.835, accuracy of 0.730, sensitivity of 0.670, specificity of 0.774, and F1 score of 0.751.
• Key predictors for acupuncture efficacy included anxiety scores (Self-Rating Depression Scale), Traditional Chinese Medicine syndrome of deficiency in heart and spleen, and Body Mass Index (BMI).

Conclusions:

• A machine learning model effectively predicts acupuncture treatment outcomes for MDD.
• This predictive tool can assist clinicians in identifying patients most likely to benefit from acupuncture, enhancing clinical decision-making and treatment effectiveness.