The utility of machine learning-based decision support system in referral of suspected rheumatic disease
Authors' conclusion
This ML-based triage tool demonstrates strong potential for accurately identifying appropriate referrals, reducing unnecessary consultations, and enhancing resource utilisation in rheumatology clinics.
Study context
- Population
- Participants attending a rheumatology outpatient clinic for the first time with suspected rheumatic diseases (RD)574 diagnosed with RD
- Participants
- 843 participants
- Outcome
- Accuracy (sensitivity, specificity, AUC) of the ML model in identifying correct vs unnecessary referral, against rheumatologic disease status at follow-up
- Follow-up
- 6 months
Methodological detailsmachine learning model development and evaluation, fivefold cross-validation, 6-month follow-up
- Machine learning model development and evaluation
- Fivefold cross-validation
- 6-month follow-up
Performance
As reported in the abstract
Identification of appropriate referral, mean across five folds
| Test or model | AUC (mean) | Sensitivity (mean) | Specificity (mean) |
|---|---|---|---|
| Machine learning (ML) model | 77.9% (95% CI: 74.9%-80.9%) | 87.1% (95% CI: 84.4%-89.8%) | 67.8% (95% CI: 62.2%-73.3%) |
Machine learning (ML) model
AUC (mean)77.9% (95% CI: 74.9%-80.9%)
Sensitivity (mean)87.1% (95% CI: 84.4%-89.8%)
Specificity (mean)67.8% (95% CI: 62.2%-73.3%)
- Participants referred unnecessarily
- 31.9%
Further findings
Best-performing fold
- AUC
- 81.34% (95% CI 78.58%-84.22%)
- Sensitivity (78.58%- 4.90%)
- 81.74%
- Specificity (76.26%-85.64%)
- 80.95%
Best fold after addition of four questions (n=245)
- AUC
- 90.77% (95% CI 87.20-94.34)
- Sensitivity
- 89.74% (95% CI 85.14-94.34)
- Specificity
- 92.05% (95% CI 86.05-98.05)
- Performance metrics significantly improved
Original publication
The utility of machine learning-based decision support system in referral of suspected rheumatic disease. Clin Exp Rheumatol. Published 3 July 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →