Development and validation of a clinical risk score to predict mortality after percutaneous coronary intervention

Objective To develop and validate a contemporary clinical risk score to predict mortality after percutaneous coronary intervention (PCI). Methods Using data collected from patients undergoing PCI at the South Yorkshire Cardiothoracic Centre, Sheffield, UK, between January 2007 and September 2013, a risk score was developed to predict mortality. Logistic regression was used to evaluate the effect of each variable upon 30-day mortality. A backwards stepwise logistic regression model was then used to build a predictive model. The results were validated both internally and externally with data from Manchester Royal Infirmary, UK. 30-Day mortality status was determined from the UK Office of National Statistics. Results The development data set comprised 6522 patients from Sheffield. Five risk factors, including cardiogenic shock, procedural urgency, history of renal disease, diabetes mellitus and age, were statistically significant to predict 30-day mortality. The risk score was validated internally on a further 3290 patients from Sheffield and externally on 3230 patients from Manchester. The discrimination of the model was high in the development (C-statistic=0.82, 95% CI 0.79 to 0.85), internal (C-statistic=0.81, 95% CI 0.76 to 0.86) and external (C statistics=0.90, 95% CI 0.87 to 0.93) cohorts. There was no significant difference between observed and predicted mortality in any group. Conclusion This contemporary risk score reliably predicts 30-day mortality after PCI using a small number of clinical variables obtainable prior to the procedure, without knowledge of the coronary anatomy.