Development and internal validation of machine learning models for gouty arthritis classification using routine...

Area of Science:

• Rheumatology
• Medical Informatics
• Machine Learning

Background:

• Gouty arthritis (GA) is a prevalent crystal-induced inflammatory arthropathy.
• GA is influenced by multiple factors including inflammatory, metabolic, renal, hematologic, demographic, and lifestyle elements.
• Serum uric acid levels alone may not fully characterize GA, necessitating integrative models for improved classification.

Purpose of the Study:

• To develop and evaluate machine learning models for classifying gouty arthritis (GA) using routinely available hospital variables.
• To assess the performance of different supervised learning algorithms in distinguishing GA from other conditions.

Main Methods:

• A retrospective study of 7,383 adult participants from an electronic medical record database.
• GA cases were identified using the 2015 ACR/EULAR classification framework.
• Four supervised learning algorithms (random forest, SVM, k-NN, logistic regression) were developed and evaluated using a hold-out design and cross-validation.

Main Results:

• The random forest model exhibited the best performance, achieving an Area Under the ROC Curve (AUC) of 0.8995 in the training set and 0.8686 in the test set.
• Random forest demonstrated high sensitivity and negative predictive value, outperforming logistic regression.
• Support vector machine ranked second, while k-nearest neighbors and logistic regression showed weaker discrimination.

Conclusions:

• Machine learning models, particularly random forest, show promise for preliminary classification of gouty arthritis using routine hospital data.
• These models can serve as a screening aid but are not a substitute for definitive diagnosis.
• External validation is crucial before widespread clinical implementation of these models.