Complex data analysis methods require optimisation techniques such as evolutionary algorithms in order to generate reliable results. The objective of this study is to analyse the relationships of particular perioperative care in colorectal surgery (CRS) with surgeon epidemiological data, performing partition grouping to look for significant relationships.
Methods: Data were used from a survey of members of Spanish coloproctology associations on perioperative care in colorectal surgery, and analysing the responses associated with mechanical bowel preparation (MBP), nasogastric intubation (NGI), drainages (D), and early feeding (EF), over the existing scientific evidence (SE) which shows that the first ones are unnecessary and the importance of the last one. We applied a variant of particle swarm optimization (PSO), to group data conglomerates, optimising variables with statistical grouping criteria.
Results: A total of 130 surveys were analysed, finding 2 clear groups which included 21.5% and 78.5% of the sample, respectively. Sixty eight per cent of the surgeons in Group A belonged to the European Board in Coloproctology, compared to none in Group B, and the former performed 80% of the coloproctology activity, compared to 60% of the rest. A responded homogeneously to questions on MBP, NGI, D and EF, those of group A following the SE, while the others did it randomly and without following it. Age, work position or academic range were not significant in the grouping.
Conclusions: The evolutionary algorithm was shown to be able to identify groups according to the use of perioperative care in CRS. Accreditation and dedication was associated with behaviour based on the SE.
Copyright © 2010 AEC. Published by Elsevier Espana. All rights reserved.