Emergency Departments (ED) across United States are distraught with issues like overcrowding, ambulance diversion, medical errors, patient left without being seen etc. The primary cause for all these interrelated problems is artificial variability that results mainly because of inaccurate severity estimation leading to inappropriate bed allocation and final disposition. To this effect, we propose Bayesian decision support tools that accurately classify new incoming patients into different severity types based on their chief complaints and at the same time assist doctors in subsequent diagnosis and disposition of patients. These tools are developed based on the decision making principles of attending physicians and information fed into them is obtained at the triage station. Further, to increase the reliability, we use kappa statistics to continuously improve the agreement between the doctors and tools developed in this research. We demonstrate applications of our models via case studies on chest pain and dyspnea, two of the most common chief complaints in hospital emergency departments.