An Intelligent System for Classifying Patient Complaints Using Machine Learning and Natural Language Processing:...

Area of Science:

• Healthcare Informatics
• Artificial Intelligence in Medicine
• Natural Language Processing

Background:

• Accurate classification of patient complaints is vital for healthcare satisfaction management.
• Traditional manual methods are inefficient and imprecise.
• Automated approaches are needed to streamline complaint classification.

Purpose of the Study:

• To develop and validate an intelligent system for automatic patient complaint classification.
• Utilize machine learning (ML) and natural language processing (NLP) techniques.
• Enhance efficiency and precision in healthcare feedback analysis.

Main Methods:

• Developed an ML-based NLP system to extract key dissatisfaction terms.
• Utilized a dataset of 1465 complaint records (2019-2023) and an external set of 376 complaints.
• Employed Synthetic Minority Oversampling Technique (SMOTE) for data balancing.
• Trained and validated Multifactor Logistic Regression, Multinomial Naive Bayes, and Support Vector Machines (SVM) algorithms.
• Performed 5-fold cross-validation on external data.

Main Results:

• The Support Vector Machines (SVM) model achieved the highest accuracy.
• SVM demonstrated a weighted average accuracy of 0.93 in the training set and 0.87 in the internal test set.
• The SVM algorithm achieved an average accuracy of 0.91 on the external test set (95% CI: 0.87-0.97).
• Ngram-level term frequency-inverse document frequency showed minimal impact on classification performance.

Conclusions:

• The NLP-driven SVM algorithm effectively classifies patient complaint texts.
• The system showed superior performance for communication and management problems.
• Caution is advised for classifying sense of responsibility complaints.
• This approach is promising for institutions with high complaint volumes and limited resources.