Cardiogenic shock, resulting from cardiac dysfunction, poses a dire threat during cardiac emergencies, necessitating prompt inpatient transfers to intensive care units and aggressive interventions for blood pressure management and adjunctive therapies. Hence, developing an optimal non-invasive decision support system for clinicians is paramount for prognostication and efficient patient transfers to specialized care units. This study aims to enhance the medical referral process for cardiogenic shock patients through Machine Learning (ML) algorithms. Analyzing data from 201 heart patients admitted to emergency wards in 2020, the study employs an Artificial Intelligence (AI)-based model with feature selection and decision phases. The feature selection phase entails analyzing 34 parameters related to the patient's health status, while the decision phase determines treatment outcomes using ensemble-based ML algorithms. Results reveal a mean patient age of 69.44 years, with 57.2% being male, and a concerning 47.7% succumbing within 30 days. Notably, the model's decision phase demonstrates an impressive predictive accuracy of 86% in determining treatment efficacy. Thus, the imperative for an optimal non-invasive decision support system for clinicians is emphasized, enabling proactive prognostication and informed patient transfers to specialized care facilities.