A guideline representation language for pervasive healthcare

This research aims to enable the use of clinical guidelines by pervasive healthcare systems to provide the best possible care to patients during their daily lives. On the one hand, pervasive healthcare systems have much potential to address the challenges arising from factors such as an increasing prevalence of chronic diseases and an ageing population. On the other hand, clinical guidelines have been used in the traditional hospital setting to assure the quality of patient care. Indeed, computer-interpretable languages have been developed such that these guidelines can be formalised and automatically executed by computer systems.

Traditionally, guideline representation languages capture the control flow between different tasks, thereby assuming a centralised system architecture. This is inappropriate for pervasive healthcare as system components may require dynamic reconfiguration in response to, for example, changing patient requirements and unreliable communications environments. Therefore, a new guideline language was developed that focuses on the data flow instead, with tasks modelled as processes executing in parallel. By dynamically distributing these processes, each device that constitutes the system can be adapted in real-time and operate independently of each other, thereby avoiding a single point of failure.

Under this new language, clinical guidelines are formalised into a network of four types processes: Monitoring, Analysis, Decision and Effectuation. To model the data flow, an accompanying archetype language was also developed in which six types of data are distinguished: Measurement, Observation, Abstraction, Action Plan, Action Instruction and Control Instruction. Each type of process and data is given a precise mathematical interpretation using axiomatic set theory, while the syntax of the languages themselves is specified formally using the EBNF notation.

Based on its formal specification, a reference implementation of the languages was developed. The result is a set of libraries implemented on top of Rosette, which provides facilities to not only execute the formalised guidelines but also formally verify the reference implementation. The languages were validated by formalising, verifying and testing a clinical guideline for diabetes management. Future work includes performing clinical studies to evaluate the languages and developing tool suites and systems to fully utilise the languages.