Diskriminační analýza jako nástroj pro hodnocení chirurgických rizik.

Abstract

This doctoral thesis deals with an analysis of colorectal operations results. These operations were carried out at the Faculty Hospital Ostrava from 2001 to 2009. Discriminant analysis methods were used as the main tool for statistical assessment of the medical data. The thesis is dedicated to a detailed description of these methods as well as to their use in the following cases: prediction of postoperative complications (morbidity) with open operation technique, prediction of conversion of laparoscopic to open colorectal surgery, identifying important risk factors associated with conversion, comparison of open and laparoscopic operation techniques with the view of long-term survival. This work is divided into nine chapters. The first chapter provides a brief introduction to the goals of the thesis and to problems being investigated from medical and statistical point of view. The second chapter is aimed at a detailed description of the operation techniques that are used in colorectal surgery and at their comparison in general. Chapter 3 presents the principles of scoring systems that are used in non-randomized medical studies to ensure objectivity of the results. Scoring system POSSUM and propensity score system, which were used in this work, were described in detail. The content of the fourth chapter is a description of the main principles of survival analysis, particularly the Kaplan-Meier method, which was used to compare survival times of laparoscopic and open colorectal surgery. Chapter 5 is a comprehensive overview of the most frequently used discriminant methods - normal discriminant analysis, logistic regression and mixtures of multivariate normal distributions. Discriminant analysis methods were applied to the source data file and the results of the analyses are performed in Chapter 6. Chapter 7 brings the final summary.