- MEDIUM
- Research
- Tier 1
- Peer-Reviewed Research
Development and external validation of an interpretable machine learning model
- Source
- Journal of Clinical Periodontology
- Author
- Hanbo Shi, Jingjing Yao, Zhongliang Jiang, Mengting Chen, Qian Zhuge, Chenchen Jiang, Yan Chen, Yun Qian
- Originally at
- onlinelibrary.wiley.com(opens in a new tab)
Summary and scoring by The Bell Brief (Dr. Jennifer Bell) using the Drill-Down Protocol. Not the original publisher.
Why it matters for dental
Researchers at Zhejiang University School of Medicine developed a blood-biomarker machine learning model to classify Stage III–IV versus Stage I–II periodontitis; the model is not ready for clinical use and does not replace periodontal examination.
Key points
- 01
Zhejiang University School of Medicine developed a stacking machine learning model using age, sex, smoking status and 13 routine blood markers to separate Stage III–IV from Stage I–II periodontitis.
- 02
The model achieved an external-validation AUC of 0.774 (95 % CI 0.730–0.818) and showed less favourable calibration (intercept 0.785, slope 0.822).
- 03
Blood-derived predictors added only modest discrimination beyond age, sex and smoking status alone.
- 04
Researchers concluded that further multi-centre validation and calibration are required before clinical deployment.
Who should care
- Academia / research
- Specialist
Read the original
Journal of Clinical Periodontology
Full reporting and any paywalled content live on onlinelibrary.wiley.com. We summarize and score; we do not republish.