Acute non-traumatic chest pain is one of the common causes of presentation in emergency patients, but the causes of acute non-traumatic chest pain are complex, the severity of the condition varies greatly, and the specificity of symptoms is not high. Machine learning and intelligent auxiliary models can greatly shorten the time of clinical decision-making, and improve the accuracy of etiological diagnosis in patients with chest pain, reduce the rate of misdiagnosis and missed diagnosis, and provide a clear direction for further treatment.

Prospective observational studies used outpatient and follow-up information to construct an auxiliary early warning model of acute non-traumatic chest pain based on federated learning, and optimized the accuracy of early warning models through retrospective and prospective studies of large cohort data, and established an efficient and stable early warning and classification model for acute non-traumatic chest pain.

Inclusion Criteria:

1. Age ≥ 18 years
2. Symptom onset or worsening within 24 hours before presentation, with a chief complaint of acute chest pain meeting the broad definition of chest pain (2021 AHA)
3. Presentation to the emergency department, with a clinical diagnosis consistent with non-traumatic chest pain
4. Signed informed consent

Exclusion Criteria:

1. traumatic chest pain
2. systemic pain caused by malignant tumors or rheumatic diseases involving the chest
3. Patients were lost to follow-up

The study population included patients admitted to the chest pain center of the pilot hospital from August 2022 to December 2027 with chest pain (including tingling, burning pain, pressure, tightness, heartburn and similar discomfort) as the main manifestations. Screening of patients with chest pain is intended from the following sources:

1. outpatients with chest pain;
2. outpatients with a history of cardiovascular disease;
3. patients from other departments of the hospital referred to the cardiology outpatient clinic due to acute chest pain.