The Need for Interactive Data-Driven Process Simulation in Healthcare: A Case Study

Abstract

In healthcare, more and more process execution information is stored in Hospital Information Systems. This data, in conjunction with data-driven process simulation, can be used, e.g. to support hospital management with Capacity Management decisions. However, real-life event logs in healthcare often suffer from data quality issues, affecting the reliability of simulation results. In this work, we illustrate the effects of disregarding data quality issues on simulation outcomes and the importance of domain knowledge using a case study at the radiology department of a hospital. Current literature on data-driven process simulation acknowledges the need for domain expertise but does not provide a framework for conceptualising the involvement of domain experts. Therefore, we propose a novel conceptual framework which interactively involves experts during data-driven simulation model development.

Supporting capacity management decisions in healthcare using data-driven process simulation

“…Some scans took precisely zero seconds, while others took several years, and a few observations were even negative, indicating that the end timestamp occurred before the start timestamp. In our previous work [61],…”

“…Even though more and more attention is devoted to the assessment and correction of data quality issues in the field of Process Mining, within the context of (automated) DDPS model discovery little explicit attention has been paid to these issues. Nevertheless, the quality of the input data has a significant impact on the reliability of the simulation results [40,61]. Therefore, it is imperative to deal with these issues appropriately.…”

“…One of the biggest challenges of this project was dealing with data quality issues. Because the quality of the input data has a significant impact on the reliability of the simulation results [40,61], it is imperative to deal with these issues appropriately. The two most prominent data quality issues were: (i)…”

“…Completely neglecting these issues, i.e. directly using the data without any pre-processing or filtering, would result in significantly higher throughput and waiting times due to congestion [61]. Therefore, it is necessary to exclude anomalies from the data.…”

“…However, this technique is not guaranteed to mark anomalies correctly, as some of these observations might be exceptional, i.e. unlikely, but not impossible [61]. In this study, we validated the scanning durations with the domain experts by verifying the durations for each examination type.…”