Details

Autor(en) / Beteiligte

• Fernandez Cisnal, A
• Lopez-Ayala, P
• Minana, G
• Valero, E
• Koechlin, L
• Catarrala, A
• Boeddinghaus, J
• Noceda, J
• Nestelberger, T
• Miro Andreu, O
• Nunez, J
• Mueller, C
• Sanchis, J

Titel

Derivation and external validation of machine-learning- models for risk stratification in chest pain with normal troponin

Ist Teil von

• European heart journal, 2023-11, Vol.44 (Supplement_2)

Erscheinungsjahr

2023

Quelle

Oxford Journals 2020 Medicine

Beschreibungen/Notizen

• Abstract Importance Risk stratification of patients with chest pain and a high-sensitivity cardiac troponin T (hs-cTnT) concentration 180 minutes). Probability thresholds for safe discharge were derived in the derivation cohort. Methods This prognosis study used a single centre cohort (July 2016-Decembre 2019) to develop and internally validate the models and a international prospective APACE cohort to perform the external validation. All Patients presenting to the ED with a complaint of chest pain and a first hs-cTnT concentration <99th percentile (14 ng/L).Candidate predictors included demographics, tradicional cardiovascular risks factors, electrocardiogram and blood tests (haemoglobin, creatinine and hs-cTnT). The endpoint was90-days death or myocardial infarction Four machine learning-based models and one logistic regression (LR) model were trained on 4075 patients and externally validated on 3609 patients. Models were compared with GRACE and HEART scores and a single undetectable hs-cTnT-based strategy (u-cTn; hs-cTnT<5ng/L and time from symptoms onset>180 minutes). Probability thresholds for safe discharge were derived in the derivation cohort. Results A 99.1% (4074) of the patients completed the follow-up in the derivation cohort and a 99.4% (3609) in the validation cohort.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). Conclusions and relevance 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.