Evaluation of risk scores for risk stratification of acute coronary syndromes in the Myocardial Infarction National Audit Project (MINAP) database
- 1Division of Cardiovascular and Diabetes Research, Leeds Institute of Genetics Health and Therapeutics, University of Leeds, Leeds, UK
- 2Biostatistics Unit, Centre for Epidemiology and Biostatistics, University of Leeds, Leeds, UK
- 3Swansea University, Singleton Park, Swansea, UK
- 4National Audit of Myocardial Infarction Project, National Institute for Clinical Outcomes Research, The Heart Hospital, London, UK
- 5Pinderfields General Hospital, Aberford Road, Wakefield, UK
- 6Academic Unit of Cardiovascular Medicine, The Yorkshire Heart Centre, The General Infirmary at Leeds, Leeds, UK
- Dr Christopher P Gale, Division of Cardiovascular and Diabetes Research, Leeds Institute of Genetics Health and Therapeutics, University of Leeds, Clarendon Way, West Yorkshire, Leeds LS2 9JT, UK;
- Accepted 15 April 2008
- Published Online First 8 May 2008
To compare the discriminative performance of the PURSUIT, GUSTO-1, GRACE, SRI and EMMACE risk models, assess their performance among risk supergroups and evaluate the EMMACE risk model over the wider spectrum of acute coronary syndrome (ACS).
Design: Observational study of a national registry.
Setting: All acute hospitals in England and Wales.
Patients: 100 686 cases of ACS between 2003 and 2005.
Main outcome measures: Model performance (C-index) in predicting the likelihood of death over the time period for which they were designed. The C-index, or area under the receiver-operating curve, range 0–1, is a measure of the discriminative performance of a model.
Results: The C-indexes were: PURSUIT C-index 0.79 (95% confidence interval 0.78 to 0.80); GUSTO-1 0.80 (0.79 to 0.81); GRACE in-hospital 0.80 (0.80 to 0.81); GRACE 6-month 0.80 (0.79 to 0.80); SRI 0.79 (0.78 to 0.80); and EMMACE 0.78 (0.77 to 0.78). EMMACE maintained its ability to discriminate 30-day mortality throughout different ACS diagnoses. Recalibration of the model offered no notable improvement in performance over the original risk equation. For all models the discriminative performance was reduced in patients with diabetes, chronic renal failure or angina.
Conclusion: The five ACS risk models maintained their discriminative performance in a large unselected English and Welsh ACS population, but performed less well in higher-risk supergroups. Simpler risk models had comparable performance to more complex risk models. The EMMACE risk score performed well across the wider spectrum of ACS diagnoses.
CG researched and wrote the manuscript. SM performed the statistics and wrote the article. CW, PB, JB and AH reviewed, wrote and critically appraised the article. All authors read and approved the final manuscript. The extract from the MINAP database was provided by JB.
Funding: MINAP is funded by the Healthcare Commission.
Competing interests: None.
Ethics approval: Ethics approval was not required.