Assessing human reliability and estimating evacuation time in hospital emergencies: A Bayesian network and Petri net approach

Abstract

The complex operational environment of hospitals, characterized by varying patient mobility levels, inconsistent staffing patterns, and resource distribution across departments, creates unique challenges for evacuation planning. This study employs Bayesian networks (BN) to enhance the Success Likelihood Index Method (SLIM), specifically addressing uncertainties related to expert judgment and data scarcity. The proposed BN-SLIM technique integrates both methods to calculate more robust Human Error Probabilities (HEPs). Furthermore, a Petri net model is developed to simulate hospital emergency evacuations, estimating evacuation time both with and without human error. The HEPs calculated by BN-SLIM were used to construct error networks, which were then integrated into the base Petri net model to quantify the probability of error along token movement paths. When modeling task errors as reattempts, the error-integrated scenario incorporated probabilistic delays at transitions associated with tasks susceptible to human error. Best-Worst Method (BWM) was used to weigh five Performance Shaping Factors (PSFs) based on expert elicitation. BN-SLIM analysis, incorporating PSF weights (e.g., competency = 0.262), generated task-specific HEPs ranging from 0.02 (equipment disconnection) to 0.15 (wheelchair positioning). Petri net simulations demonstrated that human errors increase total evacuation time by 23.37%, with wheelchair transfers showing the highest error susceptibility. The integrated methodology provides hospitals with actionable insights for optimizing evacuation plans, particularly for mobility-impaired patients.