Chemotherapy ordering and delivery: Rigorously defining and analyzing a complex process employing software engineering techniques

Abstract

17514 Background: Software engineering tools offer a novel approach to the definition and analysis of complex and defect-prone healthcare processes such as chemotherapy ordering and delivery. Methods: Software engineering researchers and oncologists, nurses, and pharmacists collaboratively defined the processes necessary to order and deliver a first cycle of adjuvant chemotherapy. Discrete procedure steps were precisely defined employing Little-JIL, a language originally created to define software development processes. Process defects were then sought through finite-state verification (using the PROPEL tool to encode constraints on event sequences for specific properties, followed by analysis employing the FLAVERS finite-state verifier), algorithmically evaluating pathways through the process model to determine if any execution of the process violates a pre-specified system property. Results: The chemotherapy order and delivery process is large–more than 250 steps described in Little-JIL notation. Medical processes were often described by terms (“verify”, “confirm”, “check”, etc.) that were used loosely and with different meanings in different contexts, resulting in inconsistency. Consequently, a precise naming glossary was created to reduce ambiguity. Little-JIL uncovered process errors and weaknesses, particularly inconsistencies between articulated and actual processes, as well as medical process guidelines that were insufficiently detailed to handle exception and atypical scenarios. Finite-state verification helped to identify missing/misplaced steps, and helped healthcare professionals think of high-level safety and regulatory goals rather than failure anecdotes and simplistic descriptions. Errors arising from unconventional treatment plans, stale patient parameters, and deadlock scenarios were revealed. Conclusions: Rigorously defining and analyzing healthcare processes led to improved insights into healthcare processes, with errors, inconsistencies, and failure to anticipate and handle exceptions identified. Such robustly-described processes may then be evaluated by other testing approaches, such as fault-tree analysis and discrete event simulation. No significant financial relationships to disclose.