Related Experiment Videos
Improving risk classification of critical illness with biomarkers: a simulation study
Christopher W Seymour1, Colin R Cooke, Zheyu Wang
1Departments of Critical Care and Emergency Medicine, University of Pittsburgh School of Medicine, Pittsburgh, PA, USA; The Clinical Research, Investigation, and Systems Modeling of Acute Illness (CRISMA) Center, Department of Critical Care, Medicine, University of Pittsburgh School of Medicine, Pittsburgh, PA, USA.
Journal of Critical Care
|April 10, 2013
Summary
Accurate risk prediction for critical illness requires strong biomarkers and large sample sizes to improve patient triage. Even moderate biomarkers significantly enhance clinical models, aiding in better classification of high-risk patients.
Area of Science:
- Emergency Medicine
- Biomarker Discovery
- Clinical Risk Prediction
Background:
- Effective triage for critical illness relies on accurate risk prediction.
- Limited data exists on the performance criteria for clinically useful biomarkers.
Purpose of the Study:
- To determine the biomarker strength and sample size needed to improve risk classification for critical illness.
- To assess biomarker utility beyond existing clinical models.
Main Methods:
- Analysis of a large adult cohort (57,647 encounters) from emergency medical services (EMS).
- Simulation of hypothetical biomarkers with varying associations with critical illness.
- Evaluation of biomarker impact on risk classification and reclassification.
Main Results:
- A moderate-strength biomarker (OR=3.0) improved model discrimination (c-statistic 0.85 vs 0.8) and reclassification (NRI=0.15).
- Significant improvements in net reclassification required at least 1000 subjects.
- Biomarker correlation with physiological variables did not alter results.
Conclusions:
- Biomarkers can substantially improve clinical models for critical illness triage.
- Substantial sample sizes and significant biomarker strength are necessary for clinical utility.
