Dealing with uncertainty in healthcare performance assessment: a fuzzy network‐DEA approach with undesirable outputs

Abstract Data Envelopment Analysis (DEA) is currently the most widely used nonparametric method for assessing system performance. However, the DEA standard approach ignores the unit's structure and assumes that the data are exact and reliable. In healthcare, these assumptions may not always hold true. To address these issues, a new approach was developed, which transformed the data into fuzzy trapezoidal numbers and used a network framework. The study was conducted using data from Portuguese public hospitals, including 18 variables related to efficiency, quality, and access. The data were then applied using a slack‐based fuzzy network‐DEA model that could handle undesirable outputs. Due to significant operational and environmental differences between hospitals in Portugal, a subsampling frontier approach based on exogenous variables was used. The results suggest that there is potential to improve hospital efficiency in Portugal by around 20%, particularly in light of the COVID‐19 pandemic. Additionally, variations in performance were observed depending on the size of the hospital.