ORBI Score Validation as Predictor of Cardiogenic Shock in Patients With St Elevation Myocardial Infarction in Two Medical Centers in Argentina.

Cardiogenic shock(CS) after ST-segment elevation myocardial infarction(STEMI) has an in-hospital mortality of 50%. The ORBI score identifies patients at risk of CS after primary angioplasty. We aim to validate the score in an Argentinian cohort. A retrospective validation analysis was carried out from a cohort of patients with STEMI in 2 centers in Buenos Aires Metropolitan Area. The predictive value of the score were estimated through its discrimination power by AUC-ROC and calibration with the Hosmer Lemeshow (HL) goodness of fit test. Four hundred and twenty-four patients were analyzed. The incidence of CS was 8.5%. The median ORBI score was 10 (IQR 7-13) vs 5 in those without CS (IQR 3-7) (P < 0.0001). The performance of the test showed an AUC-ROC of 0.80 (95%CI 0.73-0.87; P < 0.0001); and a HL X² of 4.26 (P = 0.74). The ORBI score presented an adequate predictive capacity and calibration, suggesting its possible application in this population.

Predicción de riesgo de sufrir un síndrome coronario agudo mediante un algoritmo de Machine Learning (ANGINA) / A Machine Learning Algorithm to Predict Risk for Acute Coronary Syndrome

RESUMEN Introducción: Las consultas por dolor torácico son frecuentes en los servicios de emergencias médicas (SEM). Aún no se ha identificado una estrategia diagnóstica que utilice tanto los datos objetivos como los subjetivos del dolor. Objetivos: Evaluar un clasificador de machine learning para predecir el riesgo de presentar un síndrome coronario agudo (SCA) sin elevación del segmento ST, en pacientes que consultan a un SEM con dolor torácico. Material y métodos: Se analizaron 161 pacientes que consultaron al SEM con dolor torácico. Se registró mediante un clasificador de machine learning las variables objetivas y subjetivas de caracterización del dolor. Resultados: La edad promedio fue de 57 mas/menos 12, 72,7% masculinos eran de sexo masculino y 17,4 % presentaban evento coronario previo. El 57,8% presentaba un síndrome coronario agudo con una incidencia de IAM de 29,8%, de los cuales requirieron revascularización por ATC el 35%, y CRM el 9,9% en el período de seguimiento a 30 días. Como modelo de clasificación se utilizó un Random Forest Classifier que presentó un área bajo la curva ROC de 0,8991, sensibilidad de 0,8552, especificidad de 0,8588 y una precisión de 0,8441. Las variables predictoras más influyentes fueron peso (p = 0,002), edad (p = 5,011e-07), intensidad del dolor (p = 3,0679e-05), tensión arterial sistólica (p = 0,6068) y características subjetivas del dolor (p = 1,590e-04). Conclusiones: Los clasificadores de machine learning son una herramienta útil a fin de predecir el riesgo de sufrir un síndrome coronario agudo a 30 días de seguimiento.

ABSTRACT Background: Consultations for chest pain are common in emergency medical services (EMS). A diagnostic strategy using both objective and subjective pain has not been identified yet. Objective: To evaluate a machine learning classifier as a tool for prediction of the risk of presenting a non-ST segment elevation acute coronary syndrome (ACS) in patients consulting an SEM with chest pain. Methods: 161 patients consulting SEM with chest pain were analyzed. Objective variables of the patient and subjective variables of pain characterization were recorded during the triage stage by means of a machine learning classifier. Results: The mean age was 57.43±12 years, 75% male and 16% had prior cardiovascular disease. 57.8% presented an ACS with an incidence of 29.8%, which 35% required PCI and 9.9% CRM in a 30-day follow-up period. A Random Forest Classifier was used as a classification model. The Random Forest Classifier presented an area under the ROC curve of 0.8991, sensitivity of 0.8552, specificity of 0.8588 and precision of 0.8441. The most strongest predictor variables were weight (p=0.002), age (p=5.011e-07), pain intensity (p=3.0679e-05), systolic blood pressure (p = 0.6068) and subjective pain characteristics (p=1.590e-04). Conclusions: Machine learning classifiers are a useful tool for predicting the risk of acute coronary syndrome at 30 days follow-up period.