HOLISTIC APPROACH TO ANALYSIS OF MEDICAL DATA - CANCER OF THE CORPUS UTERI

Besides the information regarding his/her disease, each hospitalized c ancer patient also provides the variety of data regarding his/her psyc hological, cultural, social, economical, genetic, constitutional and m edical background. The aim of this study was to introduce a holistic a pproach to analysis of medical data, in this case clinical data regard ing cancer of the corpus uteri. Such approach requires the collection of data regarding different aspects of the cancer patient, and after t he satisfactory sample size is obtained (which should be at least five times greater than the number of examined patient characteristics), t he performance of factor analysis. In this study, the authors have pro cessed the data regarding 25 characteristics of 928 corpus uteri cance r patients treated between 1980 and 1990 at the Department for Gynecol ogical Oncology of the University Hospital for Gynecology and Obstetri cs, Zagreb, Croatia. In factor analysis, the principal components were rotated after the initial extraction (the authors recommended the use of oblimin rotation) in order to obtain better ground for interpretat ion of the obtained results. The next step in this approach was the st epwise exclusion of characteristics with smallest communalities accord ing to Kaiser-Meyer-Olkin. criteria, and retaining the characteristics and components with the most significant impact on the explained syst em variance. When the number of principal components and initial analy zed characteristics was reduced to 3-4 and 7-10, respectively, the ult imate interpretations and conclusions were made. This approach outline d some clusters of correlations between medical data which are difficu lt to identify using other statistical procedures, primarily the impac ts of various socioeconomic and hereditary-constitutional variables on overall survival.