OBJECTIVES: Although young individuals (<55 years) are less frequently seen with acute coronary syndromes (ACS), this clinical event shows high recurrence rates and triggers considerable economic burden. Young individuals with ACS (yACS) are usually underrepresented and show idiosyncratic epidemiologic features compared to older subjects. These differences may justify why available risk prediction models usually penalize yACS with higher false positive rates compared to older subjects. We hypothesized that exploring temporal framing structures such as prediction time, observation windows and subgroup-specific prediction, could improve time-dependent prediction metrics.

METHODS: Consecutive ACS individuals (n_{global_cohort}=6341 and n_yACS=2242) undergoing coronarography up-to-48h after hospital admission were included. The observation window in STW_m and GF_m included the first 48h upon hospital admission, and LTW_m included in-hospital information. yACS cohort and global_cohort were divided into train/validation-set (70%), and both were tested in a hold-out sample of 673 yACS individuals (test-set). STW_m evaluated the occurrence of in-hospital cardiovascular deaths and recurrent ACS (MACE) with C-statistics and LTW_m estimated events occurring post-discharge from index ACS hospitalization considering time-dependent concordance (C^td-index) with competing risks (MACE versus non-cardiovascular deaths). Models were repeated over five-fold cross-validation, then assessed in test-set.

RESULTS: After median follow-up of 6.67 years, 2008 individuals presented MACE. The best strategy was to design models specifically in yACS individuals combining STW_m and LTW_m, where best results were, respectively, C-statistics [0.921(95%CI 0.889-0.953)] and C^td-index [0.722(95%CI 0.678-0.760)], while the best C^td-index in GF_m was 0.681(95%CI 0.654-0.703). There was very low concordance among top predictors of MACE for yACS versus global_cohort, as well as for STW_m versus LTW_m, reinforcing a need for specific prediction rules.

CONCLUSIONS: The predictive accuracy for adverse clinical events was optimized by using specific rules for yACS and splitting short-term and long-term prediction windows, leading to the detection of 80% of events, compared to 69% by using a rule designed for the global_cohort.