From Raw to Reliable Event Logs in Healthcare: Data Governance, Quality, and Standardization for Predictive Process Mining

Abstract

The principle of “Garbage In, Garbage Out” (GIGO) states that the quality of analytical outputs is fundamentally determined by the quality of their inputs. In Process Mining, this implies that predictive performance is highly dependent on the quality of event logs used as input data. This study investigates how Data Governance, Data Quality, and Data Standardization influence Predictive Process Monitoring outcomes in healthcare by improving event log quality. In healthcare settings, these logs are generated through the actions of clinicians and administrative staff, whose recording practices directly affect data consistency and reliability. Data Governance and Data Quality provide mechanisms to ensure that such recordings are consistent, reliable, and suitable for predictive analysis. We focus on Failure To Rescue (FTR), defined as a preventable clinical deterioration resulting from delayed recognition and intervention. Using a real hospital dataset, we conduct a comparative analysis between raw and standardized event logs to evaluate their impact on FTR prediction performance.