Abstract:
Healthcare decision-making is a fundamental and sophisticated field that generally con- sists of a series of actions taken with the aim of attaining a healthcare service require- ment. The decision-making process in this domain can be incredibly challenging due to a variety of aspects such as the diverse branches of the health industry, the presence of multiple stakeholders, the uncertainty of patients’ lives, the management of large amounts of health data with complex clinical guidelines, and so on. In such instances, the conventional and mostly manual decision-making process is usually inefficient with slow response in achieving the desired outcome with a proper management of health- care delivery system. Artificial intelligence (AI) in healthcare decision-making based on clinical knowledge and data are gaining traction as a way to enhance healthcare de- livery by making smart decisions. Therefore, the objective of this research is to propose an appropriate and sustainable framework of an intelligent healthcare decision support system (IHDSS) by combining AI-assisted decision-making methodologies with a fo- cus on the most critical aspects of the healthcare sector; which are disease diagnosis & prediction, resource management and treatment management. The study includes three types of AI-based decision-making approaches proposed for the three core stages of the integrated framework where diverse fields of AI have been employed. However, as a test case scenario, the framework has been designed focusing on decision making in healthcare support for burn patients as burns being one of the most prevalent injuries worldwide and leading causes of clinically significant morbidity which can lead to a dramatic physiological reaction with prolonged repercussions, catastrophic organ fail- ure, and death if not properly handled. Thus, for disease diagnosis and prediction phase, the study has proposed a deep convolutional neural network (DCNN) based approach for detecting the severity of burn injury utilizing real-time images of skin burns from victims. At the second phase, the study has proposed a machine learning regression approach to predict the length of stay for patients based on their clinical records with an aim to decision-making in hospital resource management. And, lastly, in the third phase of decision making, the study has proposed a fuzzy logic based model to predict the adequate intravenous fluid resuscitation rate for a burn patient’s critical treatment management. Finally, to evaluate the long term sustainability of the proposed system, this research explores the key sustainability indicators for incorporating AI in healthcare decision-making and conducts a systematic assessment to prioritize the indicators based on the perspectives of relevant experts in context of the Bangladeshi health industry.
