A Probabilistic Framework for Medical Equipment Reliability: Integrating Survival Analysis and Bayesian Learning

The increasing complexity of medical equipment, combined with operational constraints and safety requirements, poses significant challenges for healthcare technology management (HTM). Traditional failure prediction methods, including machine learning and rule-based models, often lack interpretability and fail to account for uncertainty and interdependencies among variables. To address these limitations, this study proposes a unified analytical framework that integrates survival analysis, Bayesian Networks, and Bayesian regression. Survival analysis is used to model time-to-failure and assess risk based on factors such as device age and brand. Bayesian Networks uncover conditional dependencies between lifecycle variables and support probabilistic inference under partial information. Bayesian regression quantifies the strength and uncertainty of key predictors, refining insights obtained from the network structure. Applied to maintenance data from a national public hospital network, the framework enables both descriptive and inferential analysis of equipment reliability. This study contributes a replicable and interpretable methodology to support clinical engineering decisions in complex hospital environments.