Abstract:
Case-Based Reasoning is a general artificial intelligence standard for reasoning from experience and its methodology has been investigated in improving human decision-making and has received much attention in developing knowledge-based systems in medicine. Emergency Department crowding is a major problem in Ethiopia different hospitals. This problem is negatively affecting the safety of the patients who rely on receiving a timely treatment in triage department. As a part to solving this problem, a triage process is utilized. Triage is a pre-hospital process by which patients are sorted according to the severity of their illnesses or injuries. Developing CBRSPTT will improve the triage problem process would affect the patient flow positively and in turn would enhance patient satisfaction and quality of care, save time and money. The main goals of this study is to design or develop CBR system for pre-medical triage treatment to improve the quality of decision made by general practitioners or nurses, to provide effective and efficient services to the patients, and to improve shortage of human expert in triage service at University of Gondar Hospital. To achieve the goal of these studies different literatures were reviewed such as related to artificial intelligence in the field of case based reasoning, other important articles related to the study disease types and finally related to triage articles were reviewed. To implement successfully the prototype previously solved patient cases were collected from University of Gondar Hospital statics and information office. Important attributes were identified by domain experts that are used to categorize patients according to their illness or injury. To implement the prototype we use JCOLIBRI case-based reasoning framework is realized. Finally, the performance of case-based reasoning system testing is done using different evaluation methods.
The first testing is in terms of recall and precision the system achieved 85% and 61% respectively. Secondly the performance of the system is evaluated by using users acceptance testing they achieved 86.4% rates of performance. Finally the systems were evaluated by comparing with domain experts by giving sample of cases to assign their recommended disease based on qualifications and genders. Doctors correctly and incorrectly identify 75.7% and 24.3% respectively and nurses also identify the recommended diseases correctly and incorrectly 61.2% and 38.8% respectively but the systems were correctly identified 85% and incorrectly 15%.
