Use of Machine Learning to Differentiate Children With Kawasaki Disease From Other Febrile Children in a Pediatric Emergency Department
Authors' conclusion
Furthermore, these findings suggested that machine learning with XGBoost can help physicians differentiate children with KD from other febrile children in pediatric emergency departments with excellent sensitivity, specificity, and accuracy.
Study context
- Population
- Febrile children younger than 5 years, 4 hospitals (2 medical centers, 2 regional hospitals), 2010–2019
- Participants
- 74,641 participants4 hospitals
- Comparison
- XGBoost prediction model to differentiate Kawasaki disease (KD) from other febrile children
- KDn = 1,142
- Controln = 73,499reference
- Outcome
- Whether the febrile children fulfilled the diagnostic criteria of KD
Methodological detailsmulticentre (4 hospitals), machine learning (XGBoost)
- Multicentre (4 hospitals)
- Machine learning (XGBoost)
Performance
As reported in the abstract
Prediction model performance for fulfilling KD diagnostic criteria
| Test or model | Sensitivity | Specificity | PPV | NPV | Positive likelihood ratio | AUC |
|---|---|---|---|---|---|---|
| Prediction model | 92.5% | 97.3% | 34.5% | 99.9% | 34.0 | 0.980 (95% CI, 0.974-0.987) |
Prediction model
Sensitivity92.5%
Specificity97.3%
PPV34.5%
NPV99.9%
Positive likelihood ratio34.0
AUC0.980 (95% CI, 0.974-0.987)
Further findings
| Outcome | KD | Control | OR (95% CI) | |
|---|---|---|---|---|
| Male sex, KDvs control | 60.2% | 56.4% | 1.79 (1.55–2.06) | |
| 0.512Odds ratio (log scale) |
Mean, years
| Outcome | KD | Control | Mean difference (years) (95% CI) | |
|---|---|---|---|---|
| Age, KDvs control | 1.1 (SD 0.8) | 1.6 (SD 1.4) | −0.6 (−0.6 to −0.5) | |
| −1−0.500.5Difference, years (linear scale) |
Original publication
Use of Machine Learning to Differentiate Children With Kawasaki Disease From Other Febrile Children in a Pediatric Emergency Department. JAMA Netw Open. Published 3 April 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →