LimitationsThe MUSIC cardiac mortality risk score was derived in a cohort ofCHF patients with predominant systolic CHF (75% of cases);however, the proportion of patients with preserved LVEF in ourstudy is the same as in the CHARM-11 The SHFM was developedand validated mainly in patients with systolic HF, but its accuracywas similar in the IN-CHF population, in whom one-third ofpatients had an EF of 40%10 Finally, we demonstrated that theperformance of the cardiac mortality model was equally good inthe preserved and depressed LVEF populations, with c-indices 80 and 77, Our model was not validated in a separatedata set, because there are no other similar contemporarycohorts of CHF patients in which the 10 multivariable predictorssummarized in Figure 1 have been On the otherhand, the relatively small number of SDs (n ¼ 90) precluded splittingour cohort into training and validation However, thebootstrap re-sampling technique ruled out any ‘over-optimism’ inthe predictive discrimination with similar c-indices, using the originaland re-fitted 11,21,22 Furthermore, using the re-fittedmodel on the OOB observations after bootstrapping resulted inalmost equally good c-indices, which is nowadays considered as areliable estimation of external validation and a good argumentagainst the existence of significant overfitting of the 23,24 Aspreviously stated, PFD and SD models are based on fewerevents than the cardiac mortality risk score; therefore, theformer are more prone to overfitting than the Howeverthe PFD and SD models are only secondary endpoints of theMUSIC study, while the number of events of the primary endpoint(cardiac mortality, n ¼ 213) is clearly enough to avoid significantoverfitting in an 8-predictor model, built from 20 candidate Nevertheless, further studies are needed with validation ofthe MUSIC cardiac mortality risk score in other CHF cohorts toconfirm its value as a generalizable clinical prediction In conclusion, this independent, non-interventional study in anoutpatient CHF population showed the ability to predict mortalityin CHF patients using a simple score including a limited number ofpredictors that can be easily applied in clinical None ofthese predictors require invasive or expensive procedures, andall are unequivocally objective The use of this modelidentifies a subgroup of high-risk patients who should be