Framework for automating the categorization of hospital corrective building maintenance work orders based on end-user requests

Abstract

Hospitals, characterized by their complexity, high costs, and unique challenges compared to other facilities types, impose significant demands on operation and maintenance (O&M) practices. O&M professionals consistently strive to enhance and standardize the quality of available information. However, they face substantial challenges in the collection, retrieval, and sharing of data due to the multidisciplinary nature of hospital activities, which require extensive information management. Access to up-to-date and easily comprehensible information is essential to support O&M practitioners in their decision-making processes. While the integration of information technologies (ITs), such as computerized maintenance management systems (CMMSs), may seem like a straightforward solution, they present one of the primary challenges in managing hospital facilities. Although these systems capture and store data, they often lack the capabilities to efficiently sort and exchange information. The transformation of raw data, particularly from end-user requests, into actionable insights holds significant potential for improving maintenance processes. However, this transformation is often hindered by data overload and the complexity of extraction techniques. Recent research on information management (IM) in hospital building maintenance remains limited. Successful IM implementation depends on two key factors: strategically identifying and prioritizing operational information and integrating appropriate tools. In this context, the primary objective of this study is to develop a framework for automating the categorization of hospital corrective building maintenance work orders (WOs) based on information derived from end-users requests. This research adopts a data-driven research (DDR) approach, supported by a design science (DS) methodology, and focuses on a federal public hospital in Maceió, State of Alagoas, Brazil. Building upon the objective, the study utilizes CMMS data spanning 2017– 2022 to identify critical performance metrics, including lead time, response patterns, and keyword associations. These insights informed the development of domain-specific term dictionaries through semantic analysis, which serve as the foundation for an automated classification framework. By integrating association rule mining (ARM) via the Apriori algorithm, the framework achieved strong validation metrics, with 89.68% accuracy on historical data and 93.17% accuracy on new requests from 2023. This automated method transforms unstructured textual data into actionable information, helping to address inefficiencies in hospital O&M practices and better align maintenance workflows with operational demands. Key contributions of this study include a replicable methodology for CMMS data optimization, tools for automating request categorization, and a scalable framework adaptable across diverse institutional contexts. By bridging the gap between raw data and structured insights, the framework empowers maintenance managers to address operational inefficiencies. These findings underscore the transformative potential of integrating data analytics and semantic tools into hospital facility management, fostering efficiency, scalability, and a foundation for future innovations in data-driven operations.