Related Experiment Videos
The comparison of injury severity instrument performance using likelihood ratio and ROC curve analyses
R R Lett1, J A Hanley, J S Smith
1Department of Surgery, McGill University, Montreal, Canada.
The Journal of Trauma
|January 1, 1995
Summary
Likelihood ratio and ROC curve analyses effectively differentiate trauma injury severity instruments. Non-physiological data in CRAMS, PHI, and RTI improved performance over the physiological-only Revised Trauma Score (RTS).
Area of Science:
- Trauma research
- Medical instrument evaluation
- Statistical analysis in healthcare
Background:
- Assessing injury severity instruments is challenging due to varied research methodologies.
- Statistical techniques and study populations often obscure performance differences between trauma instruments.
Purpose of the Study:
- To highlight the benefits of likelihood ratio and receiver operator characteristic (ROC) curve analyses.
- To demonstrate these analytical methods using data that clearly distinguishes instrument performance.
Main Methods:
- Analysis of previously published data on the Injury Severity Score (ISS) and four triage instruments for mortality prediction.
- Paired comparison of two instruments using a new trauma registry dataset.
- Application of likelihood ratio and ROC curve analyses.
Main Results:
- Triage instruments with non-physiological data (CRAMS, PHI, RTI) outperformed the physiological-only Revised Trauma Score (RTS).
- Performance gains over RTS ranged from 5.9% to 43.5%.
- The ISS demonstrated consistent performance for blunt and penetrating trauma, raising concerns about its overall adequacy.
Conclusions:
- Likelihood ratio and ROC curve analyses reveal performance disparities between injury severity instruments.
- These advanced statistical methods overcome limitations of less rigorous comparison techniques.