A multi-tiered medical data-mining and visualization framework for rule investigations

Hospital ED crowding has led to an increase in patients’ waiting times; thus, solving this problem requires a better understanding of the hospital’s patient flow and the behaviors of patients. Existing research on ED crowding is sparse and has tended to focus on the present crowding state. Recent researches have addressed the importance of analyzing the length of stay (LOS) to understand the behaviors of patients in the ED, and these provide a good departure point for understanding patients’ behaviors based on the LOS factor. In our ongoing work, we proposed a domain-driven, ED-crowding data-mining approach to investigate the relationship between various types of patient behaviors and their LOS and to build a model to predict patients’ LOS. The objective of this study is to build an interactive decision support system (DSS) for Mackay Memorial Hospital, which has the second-largest ED in Taiwan and is a representative institute. Accordingly, the contributions of this study are (1) building the DSS based on the proposed domain-driven medical data-mining process in the ED and (2) visualizing the extracting rules and the statistical data in the proposed rule-based medical decision support (R-MDS) visualization portal. We introduce the system framework with associated modules in this study. We aim to integrate domain knowledge of the hospital ED with the data-mining technique to develop the system and provide interactive DSS using modern visualization techniques. We also believed that the qualified rules can be validated effectively and efficiently by experts with the aid of the proposed framework.