Subpopulation process mining in healthcare

In clinical pathways, there are thought to be differences between the treatment of different patient subpopulations. This paper provides a method for comparing clinical pathways of different patient subpopulations. To perform and validate the method three diseases were chosen, diabetes type II, chronic kidney disease and urinary tract infection. Within these diseases a number of different subpopulations were chosen from the MIMIC-III v1.4 data set to be compared against each other. Analysis of data shows statistically significant differences in clinical pathways in the form of graph per subpopulation. The results indicate that it is possible to apply process comparison within process mining on medical data and that the resulting models are sound within the medical domain.