Related Experiment Videos
Classification trees and logistic regression applied to prognostic studies: a comparison using meningococcal disease
G L Werneck1, D M de Carvalho, D E Barroso
1Department of Epidemiology, Harvard School of Public Health, Boston, MA 02115, USA.
Journal of Tropical Pediatrics
|September 1, 1999
Summary
Predicting fatal outcomes in pediatric meningococcal disease is possible using statistical models. Classification trees offer a user-friendly approach for clinical settings, comparable to logistic regression in accuracy.
Area of Science:
- Pediatric Infectious Diseases
- Clinical Epidemiology
- Biostatistics
Background:
- Meningococcal disease poses a significant threat to children's health.
- Accurate prediction of fatal outcomes is crucial for timely intervention.
- Existing predictive models may lack ease of application in emergency settings.
Purpose of the Study:
- To develop and compare prediction models for fatal outcomes in pediatric meningococcal disease.
- To evaluate the clinical utility of logistic regression versus classification trees.
Main Methods:
- A cohort of 829 children hospitalized with meningococcal disease was analyzed.
- Logistic regression and classification trees were employed to build predictive models.
- Model performance was assessed using the area under the receiver operator characteristic (ROC) curve.
Main Results:
- Both logistic regression (92%) and classification trees (88%) demonstrated strong predictive accuracy.
- Logistic regression provides explicit statistical inference measures.
- Classification trees offer a more intuitive graphical display and simpler application in clinical practice.
Conclusions:
- Classification trees provide a practical and understandable tool for predicting fatal outcomes in pediatric meningococcal disease.
- The choice of model may depend on the primary objective: statistical inference (logistic regression) or clinical application (classification trees).