Monitoring Temporal Patterns in Guideline-based Care

Clinical guidelines and protocols have become increasingly important in clin-
ical practice. Computer-based application of guidelines is one of the keys to
improved patient care. Therefore, integration of guideline execution into
the clinical data ow becomes more and more important. Temporal data
abstraction is required to apply high-level medical knowledge to low-level
measurements and data.
The guideline modelling language Asbru provides strong temporal ab-
straction capabilities integrated with guideline execution. Because of the
complexity of the language, writing an interpreter for Asbru is non-trivial,
and an execution engine that would live up to the potentials of the language
was not available before.
In this thesis, I describe the design and implementation of a framework to
support the execution of Asbru guidelines, building on existing work. Asbru
guidelines are compiled into a network of abstraction and plan modules. This
network performs the content of the plans synchronised with the patient
state.
Monitoring patient data requires algorithms to match the real world data
with the temporal patterns dened in the guideline. In order to provide de-
cision support in high-frequency domains such as intensive care, these algo-
rithms must comply with dened runtime constraints. This thesis describes
suitable algorithms for the most important temporal abstraction features of
Asbru.
The described framework and algorithms were integrated in the Asbru
Interpreter and successfully evaluated in a European project.