Aim: To compare the prognostic accuracy of six scoring models for up to three-year mortality and rates of hospitalisation due to acute decompensated heart failure (ADHF) in STEMI patients.
Methods and results: A total of 593 patients treated with primary PCI were evaluated. Prospective follow-up of patients was ≥3 years. Thirty-day, one-year, two-year, and three-year mortality rates were 4.0%, 7.3%, 8.9%, and 10.6%, respectively. Six risk scores--the TIMI score and derived dynamic TIMI, CADILLAC, PAMI, Zwolle, and GRACE--showed a high predictive accuracy for six- and 12-month mortality with area under the receiver operating characteristic curve (AUC) values of 0.73-0.85. The best predictive values for long-term mortality were obtained by GRACE. The next best-performing scores were CADILLAC, Zwolle, and Dynamic TIMI. All risk scores had a lower prediction accuracy for repeat hospitalisation due to ADHF, except Zwolle with the discriminatory capacity for hospitalisation up to two years (AUC, 0.80-0.83).
Conclusions: All tested models showed a high predictive value for the estimation of one-year mortality, but GRACE appears to be the most suitable for the prediction for a longer follow-up period. The tested models exhibited an ability to predict the risk of ADHF, especially the Zwolle model.