Related Experiment Videos
ROC methodology within a monitoring framework
Corette B Parker1, Elizabeth R DeLong
1RTI International, P.O. Box 12194, Research Triangle Park, NC 27709, USA. RETTE@RTI.ORG
Statistics in Medicine
|November 6, 2003
Summary
This study extends receiver operating characteristic (ROC) methodology for diagnostic tests used over time. It accounts for changing event status and varying information per subject, improving test performance assessment.
Area of Science:
- Biostatistics
- Medical Diagnostics
- Health Services Research
Background:
- Receiver operating characteristic (ROC) methodology is standard for evaluating diagnostic tests with static event status.
- Current methods do not adequately address scenarios with repeated testing and dynamic event status over time.
Purpose of the Study:
- To extend ROC methodology for diagnostic tests applied repeatedly over time.
- To accommodate time-varying event status and varying information per individual.
- To provide a framework for assessing test performance in longitudinal monitoring.
Main Methods:
- Extension of standard ROC methodology to a framework with repeated diagnostic testing.
- Modeling data as incomplete discrete failure time data with time-varying covariates.
- Development of computational simplifications for variance and covariance calculations for areas under ROC curves.
Main Results:
- The proposed framework appropriately relates diagnostic markers to the true condition over time.
- It accounts for varying amounts of information per individual.
- The method allows for assessment of test performance in conjunction with the testing schedule.
Conclusions:
- The extended ROC methodology offers a robust approach for evaluating diagnostic tests in dynamic monitoring settings.
- This framework is particularly useful for analyzing periodic monitoring data, such as for reperfusion after myocardial infarction.
- The computational simplification aids in the practical application of these advanced ROC analyses.