Related Experiment Videos
Predicting ankylosing spondylitis disease activity via patient-reported outcome measures: Building prediction models
Yifan Gong1, Aomei Liu2, Li Zhuo2
1Guang'anmen Hospital, China Academy of Chinese Medical Sciences, Beijing, China.
Plos One
|July 15, 2026
Summary
Machine learning models effectively predict ankylosing spondylitis disease activity using patient-reported outcomes. This approach shows strong agreement with clinical measures, improving disease monitoring and management.
Area of Science:
- Rheumatology and Artificial Intelligence
- Clinical Informatics and Predictive Modeling
Background:
- Ankylosing spondylitis (AS) disease activity monitoring is crucial for patient management and treatment decisions.
- Patient-reported outcome measures (PROMs) are increasingly used in AS care, but their predictive power for clinical scores is underexplored.
Purpose of the Study:
- To develop and evaluate machine learning (ML) models for predicting Ankylosing Spondylitis Disease Activity Score-C-reactive protein (ASDAS-CRP) using PROMs data.
- To assess the efficacy of ML techniques in estimating disease activity in AS patients based on patient-reported information.
Main Methods:
- Utilized data from 389 AS patients, splitting into training (80%) and testing (20%) sets.
- Employed 34 variables, including PROMs (e.g., BASDAI, BASFI) and clinician data, for feature selection.
- Constructed ten models using Support Vector Machine (SVM) and K-Nearest Neighbour (KNN) with five feature selection methods (FSOR, TRC, RFS, PCC, ReliefF).
Main Results:
- The FSOR+SVM model demonstrated superior performance with an AUROC of 0.930 in the validation set.
- This model achieved high accuracy (79.35%), specificity (90.50%), and sensitivity (83.78%).
Conclusions:
- ML models derived from PROMs data can effectively predict AS disease activity.
- The developed models show strong concordance with clinical ASDAS-CRP measures, offering a valuable tool for AS management.
