Computational modeling for decision making in the establishment and implementation of management systems in laboratories

Abstract

Implementing Laboratory Management Systems (LMS) involves an accreditation process whereby the laboratory documents its processes and demonstrates the quality and reliability of those processes. The ISO/IEC 17025 standard specifies the accreditation requirements in an international context. Since accreditation processes are complex, expensive, involve multiple steps and variables and a heavy documentation burden, it would be desirable to have some form of quick a priori assessment of a laboratory before it embarks on the accreditation journey. This paper presents a computational model based on Bayesian Networks (BN) for this purpose. The formalism focused on the use of Bayesian statistics and decision graphs. Identification of the BN variables, conditional dependencies, probabilistic information and domain knowledge were obtained from field investigations. Model evaluation was performed using a set of known scenarios that represent unequivocal prescriptions. The model enables predicting the results of implementing a management system, and simulating the accreditation process results. This development also identified the variables that have a significant influence on the expected results of the LMS accreditation process.