Prediction of patient choice tendency in medical decision-making based on machine learning algorithm

Area of Science:

• Artificial Intelligence
• Medical Informatics
• Health Services Research

Background:

• Machine learning (ML) algorithms, a subset of artificial intelligence, offer capabilities to simulate human behavior through data training.
• Predicting patient choice tendencies in medical decision-making is crucial for personalized healthcare.
• Understanding patient preferences aids physicians in developing effective clinical treatment strategies.

Purpose of the Study:

• To predict patient choice tendencies in medical decision-making using ML algorithms.
• To compare the performance of Decision Tree (DT), K-Nearest Neighbor (KNN), and Support Vector Machine (SVM) algorithms in this prediction task.
• To provide a resource for developing patient-centered decision-making schemes in clinical treatment.

Main Methods:

• Collected primary survey data from 248 participants in Chinese hospitals.
• Defined 12 predictor variables and 4 outcome variables (treatment effect, cost, side effect, experience).
• Applied and compared three ML classification algorithms: DT, KNN, and SVM.

Main Results:

• SVM achieved the highest prediction accuracy across all outcome variables, with specific accuracies of 82% (effect), 76% (cost), 80% (side effect), and 94% (experience).
• SVM also demonstrated superior performance based on F1-scores (0.81, 0.74, 0.73, 0.94).
• DT and KNN algorithms showed varying levels of accuracy and F1-scores for different decision-making aspects.

Conclusions:

• The Support Vector Machine (SVM) algorithm exhibits the highest accuracy and best overall performance for predicting patient medical decision-making tendencies.
• These findings offer valuable guidance for physicians in formulating clinical treatment plans.
• The study supports the development of patient-centered medical decision assistance systems to enhance physician-patient communication and facilitate scientific decision-making.