| dc.description.abstract | The survival rate of out-of-hospital cardiac arrest
is still low despite improvements in treatment strategies.
The recent decades have brought improved technologies for
data collection, and automatic annotation of treatment events and
cardiac arrest rhythm transitions. This opens the possibility
for studying the interaction between treatment and patient response by
searching for patterns in state sequence representations of
rhythm and treatment labels.
A method using clustering by partitioning around medoids on optimal matching
distances to identify groups of similar treatment sequences was proposed.
The method was tested in the form of a case study of 250 cardiac arrest
episodes. Eight clusters of cardiac compression depth, and cardiac compression rate sequences were identified. The outcome measure used to evaluate the efficacy of the clusters
were based on the proportion of favorable local rhythm transitions.
Overall, better outcomes were seen for treatment clusters with low
hands off fractions and cardiac compression rates lower than 125/min, however
these differences between the clusters could not be concluded to
be due to the variation in treatment. | |