Machine learning predicts emergency physician specialties from treatment strategies for patients suspected of myocardial infarction

Abstract

BACKGROUND: Our investigation aimed to determine how the diverse backgrounds and medical specialties of emergency physicians (Eps) influence the accuracy of diagnoses and the subsequent treatment pathways for patients presenting preclinically with MI symptoms. By scrutinizing the relationships between EPs' specialties and their approaches to patient care, we aimed to unveil potential variances in diagnostic accuracy and treatment choices. METHODS: In this retrospective, monocenter cohort study, we leveraged machine learning techniques to analyze a comprehensive dataset of 2328 patients with suspected MI, encompassing preclinical diagnoses, electrocardiogram (ECG) interpretations, and subsequent treatment strategies by attending EPs. RESULTS: We demonstrated that diagnosis and treatment patterns of different specialties were distinct enough, that machine learning (ML) was able to differentiate between specialties (maximum area under the receiver operating characteristic = 0.80 for general medicine and 0.80 for surgery). In our study, internist demonstrated the highest accuracy for preclinical identification of STEMI (0.96) whereas surgeons showed the highest accuracy for identifying NSTEMI. Our findings highlight significant correlations between EP specialties and the accuracy of both preclinical diagnoses and subsequent treatment pathways for patients with suspected MI. CONCLUSIONS: Our results offer valuable insights into how the diverse backgrounds and specialties of EPs can influence the optimization of patient care in emergency settings. Understanding these patterns can help in the development of tailored training programs and protocols to enhance diagnostic accuracy and treatment efficacy in emergency cardiac care, ultimately optimizing patient treatment and improving outcomes.

Ähnliche Arbeiten