Fuzzy Logic AGgregation of Crisp Data Partitions as Learning Analytics in Triage Decisions

This paper provides an analytical learning system based on Fuzzy Logic AGgregation (FLAG) of crisp data partitions to improve the Triage process in a hospitality emergency department. The method compares patient rankings made by nurses with those made by an Expert to detect points for improvement. Specifically, a normalized concordance index per nurse and the average of them allow for the evaluation of the ability of the nurses of well aggregating the cases in Triage process. The proposed FLAG system is tested through an empirical case study by simulating the patients arriving at two Emergency Departments Triage. The main contribution is the definition of the global performance index combining both the nurse’s partitioning concordance with respect to the Expert’s one and the accuracy in class assignment. The empirical distribution function of the global concordance index is derived through permutation method. In this way, Kolmogorov-Smirnov testing provides the comparison of the performances of the two healthcare units. The pay-off table concordance-accuracy allows to address improvement actions. Another tool of the system is the correspondence analysis to visualize the accuracy of decisions on the class priority as well as the sharing behaviours that influence the nurse’s judgements. All this becomes part of the FLAG learning analytics system which is able to outline critical points by red flags. to improve further assignments and the overall management organization. FLAG system can be adapted in all other situations of risk where cognitive heuristics face an accuracy-effort trade-off such that their simplified decision process leads to reduced accuracy.