Automatic Process Comparison for Subpopulations: Application in Cancer Care

Abstract

Processes in organisations, such as hospitals, may deviate from the intended standard processes, due to unforeseeable events and the complexity of the organisation. For hospitals, the knowledge of actual patient streams for patient populations (e.g., severe or non-severe cases) is important for quality control and improvement. Process discovery from event data in electronic health records can shed light on the patient flows, but their comparison for different populations is cumbersome and time-consuming. In this paper, we present an approach for the automatic comparison of process models that were extracted from events in electronic health records. Concretely, we propose comparing processes for different patient populations by cross-log conformance checking, and standard graph similarity measures obtained from the directed graph underlying the process model. We perform a user study with 20 participants in order to obtain a ground truth for similarity of process models. We evaluate our approach on two data sets, the publicly available MIMIC database with the focus on different cancer patients in intensive care, and a database on breast cancer patients from a Dutch hospital. In our experiments, we found average fitness to be a good indicator for visual similarity in the ZGT use case, while the average precision and graph edit distance are strongly correlated with visual impression for cancer process models on MIMIC. These results are a call for further research and evaluation for determining which similarity or combination of similarities is needed in which type of process model comparison.

Keywords: MIMIC database; breast cancer care; cancer types; process comparison; process mining; quality control.

Figure A1

Discovered process models for different cancer types in MIMIC.

Figure A2

Relation of non-aggregated measures for MIMIC. Based on measures reported in Table 3. Diagonal: normalized histogram and kernel density estimation of the distribution. Lower triangle: scatterplot with estimated linear regression line. Upper triangle: pairwise Spearman’s rank correlations with p-values.

Figure A3

Discovered process models on ZGT for sub-populations with events on level 1.

Figure A4

Relation of non-aggregated measures for ZGT. Based on measures reported in Table 7. Diagonal: normalized histogram and kernel density estimation of the distribution. Lower triangle: scatterplot with estimated linear regression line. Upper triangle: pairwise Spearman’s rank correlations with p-values.

Figure 1

Overview of the approach.

Figure 2

Overview of methods for pairwise comparison of process models.

Figure 3

Intuition for graph-based similarity. Left: for graphs constructed from IvM models; right: for graphs constructed from Petri nets. Event nodes match only if they represent the same event. Other nodes match if they have the same type, e.g., represent the start of an XOR branch (XOR-S), independent on the ID of the graph node.

Figure 4

Discovered process models for cancer types 2, 7, and 8 in MIMIC.

Figure 5

Process model (left) and constructed directed graph (right) for cancer type 10. Graph on the right was drawn using the networkx implementation of Kamada-Kawai path-length cost-function. Note that there is node and edge overlap for all event nodes, making the position nodes hardly visible.

Figure 6

Relation of aggregated measures for MIMIC. Based on measures reported in Table 3. Diagonal: normalized histogram and kernel density estimation of the distribution. Lower triangle: scatterplot with estimated linear regression line. Upper triangle: pairwise Spearman’s rank correlations (${\rho }_s$) with p-values.

Figure 7

Event types on different levels of granularity (level 1 on top, level 2 at the bottom)

Figure 8

Discovered process models on ZGT for sub-populations with events on level 1.

Figure 9

Process model for NoSVOB population (top) and derived directed graph (bottom). Edge weights and loops omitted in the graph for readability.

Figure 10

Discovered process models on ZGT for sub-populations with second-level events.

Figure 11

Relation of aggregated measures for ZGT. Based on measures reported in Table 7. Diagonal: normalized histogram and kernel density estimation of the distribution. Lower triangle: scatterplot with estimated linear regression line. Upper triangle: pairwise Spearman’s rank correlations (${\rho }_s$) with p-values.