Prospective Validation of the STOPSHOCK Score - Artificial Intelligence Based Predictive Scoring System to Identify the Risk of Developing Cardiogenic Shock (CS) in Patients Suffering From Acute Coronary Syndrome (ACS)

Cardiogenic shock is a serious life-threatening condition affecting almost 10% of patients suffering from acute coronary syndrome (ACS). When untreated, it can rapidly progress to collapse of circulation and sudden death. Despite recent improvements in diagnostic and treatment options, mortality remains incredibly high, reaching nearly 50%. Currently available mechanical circulatory support devices can replace the function of the heart and/or lungs, thereby essentially eliminating the primary cause. However, cardiogenic shock is not only an isolated decrease in cardiac function but a rapidly progressing multiorgan dysfunction accompanied by severe cellular and metabolic abnormalities. The window for successful treatment is relatively narrow, and when missed, even the elimination of the underlying primary cause is not enough to reverse this vicious circle. The ability to identify high-risk patients prior to the development of shock would allow to take pre-emptive measures, such as the implantation of mechanical circulatory support, and thus prevent the development of shock leading to improved survival. For this purpose, Premedix Academy has developed and validated a predictive scoring system STOP SHOCK (Score TO Predict SHOCK). This scoring system showed better prediction compared to standard models and was accepted to the Late- Breaking Science section at the European Society of Cardiology (ESC) Congress 2024. STOP SHOCK was validated on an external cohort of 5123 ACS patients with area under the receiver operating characteristic curve (ROC AUC) of 0.844 (95% confidence interval: 0.841-0.8470) surpassing other externally validated cardiogenic shock (CS) models (e.g. ORBI score). Furthermore, our model is based on variables that are readily available at the first contact with patients and thus STOPSHOCK can be utilized in emergency room (ER) or ambulance even before catheterization. Novelty of our project is also in the concept of continuous training, improvement, and validation to ensure validity and clinical applicability in the future as well. Current medical models are developed, verified, and published. Once the model enters medical practice, research teams will either validate it or replace it with their own model based on a new cohort of patients. However, experience from other fields shows that as soon as machine learning models are deployed, their performance degrades. In order to preserve and even further improve the model, continuous performance monitoring and training/retraining are vital. A small prospective validation study on a cohort of 103 consecutive higher-risk ACS patients, enrolled in intensive cardiac care units in 8 centers from USA, Europe, and Asia demonstrated very good performance with ROC AUC of 0.97 and was presented at the 2023 American Heart Association Annual Meeting. The STOPSHOCK score is currently available as a smartphone application and as an online calculator: https://stopshock.org.

The primary objective of this study is to prospectively validate the STOPSHOCK score on a large cohort of ACS patients. The methods and results of this project follow the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) statement.

• Sensitivity (Recall) of the STOPSHOCK Score [Time frame: Up to hospital discharge (average of 14 days)]
• Specificity of the STOPSHOCK Score [Time frame: Up to hospital discharge (average of 14 days)]
• Positive Predictive Value (Precision) [Time frame: Up to hospital discharge (average of 14 days)]
• Negative Predictive Value [Time frame: Up to hospital discharge (average of 14 days)]
• F1 Score of the STOPSHOCK Score [Time frame: Up to hospital discharge (average of 14 days)]
• Accuracy of the STOPSHOCK Score [Time frame: Up to hospital discharge (average of 14 days)]
• Area Under the Precision-Recall Curve (PR AUC) [Time frame: Up to hospital discharge (average of 14 days)]
• Matthews Correlation Coefficient [Time frame: Up to hospital discharge (average of 14 days)]
• Youden's J Statistic [Time frame: Up to hospital discharge (average of 14 days)]
• Brier Score [Time frame: Up to hospital discharge (average of 14 days)]
• Calibration Slope and Intercept [Time frame: Up to hospital discharge (average of 14 days)]