Related Experiment Videos
Models for diagnosing chest pain: is CART helpful?
N J Crichton1, J P Hinde, J Marchini
1Royal College of Nursing Institute, London, U.K.
Statistics in Medicine
|April 15, 1997
Summary
Classification and Regression Trees (CART) show potential for diagnosing anterior chest pain by identifying key indicators. However, current performance is disappointing, requiring methodological improvements for clinical use.
Area of Science:
- Medical informatics
- Clinical decision support systems
- Biostatistics
Background:
- Anterior chest pain diagnosis presents challenges in identifying causal factors.
- Existing diagnostic methodologies may have limitations in handling complex patient data.
Purpose of the Study:
- To evaluate the Classification and Regression Tree (CART) methodology for diagnosing anterior chest pain.
- To compare CART's performance against correspondence analysis and independent Bayes classification.
Main Methods:
- Application of Classification and Regression Tree (CART) algorithms.
- Comparative analysis with correspondence analysis and independent Bayes classification.
Main Results:
- CART demonstrated potential in identifying significant indicators and optimal cutpoints for continuous variables.
- The overall diagnostic classification performance of CART was found to be unsatisfactory.
- Comparison revealed limitations in CART's current efficacy for this specific clinical application.
Conclusions:
- CART methodology offers promise for identifying critical diagnostic factors in anterior chest pain.
- Further methodological enhancements are necessary to improve CART's accuracy and suitability for clinical practice.
- Future research should focus on refining CART for better diagnostic performance in complex medical scenarios.