Background
Risk stratification of patients with chest pain and a high-sensitivity cardiac troponin T (hs-cTnT) concentration <upper reference limit (URL) is challenging.
Objectives
To develop and externally validate clinical models for risk prediction of 90-days death or myocardial infarction in patients presenting to the emergency department with chest pain and an initial hs-cTnT concentration <URL.
Methods
Four machine learning-based models and one logistic regression (LR) model were trained on 4075 patients (single-center Spanish cohort) and externally validated on 3609 patients (international prospective APACE cohort). Models were compared with GRACE and HEART scores and a single undetectable hs-cTnT-based strategy (u-cTn; hs-cTnT < 5 ng/L and time from symptoms onset > 180 minutes). Probability thresholds for safe discharge were derived in the derivation cohort.
Results
The endpoint occurred in 105 (2.6%) patients in the training set and 98 (2.7%) in the external validation set. Gradient boosting full (GBf) showed the best discrimination (AUC = 0.808). Calibration was good for the reduced NNET and LR models. GBf identified the highest proportion of patients for safe discharge (36.7% vs 23.4% vs 27.2%; GBf vs LR vs u-cTn, respectively) with similar safety (missed endpoint per 1000 patients: 2.2 vs. 3.5 vs. 3.1, respectively). All derived models were superior to the HEART and GRACE scores (p < 0.001).
Conclusion
Machine learning and logistic regression prediction models were superior to the HEART, GRACE and u-cTn for risk stratification of patients with chest pain and a baseline hs-cTnT < URL. GBf models best balanced discrimination, calibration and efficacy reducing the need for serial hs-cTn determination by more than one-third.
Clinical Trial Registration
ClinicalTrials.gov number, NCT00470587, https://clinicaltrials.gov/ct2/show/NCT00470587