Avtor/Urednik		Pohar, Maja; Stare, Janez; Henderson, Robin
Naslov		An individual measure of relative survival
Tip		članek
Vir		In: ROeS (Region Oesterreich - Schweiz) seminar; 2003 Sep 28-Oct 2; Gallen. Gallen: Internationale Biometrische Gesellschaft,
Leto izdaje		2003
Obseg		str. 1-13
Jezik		eng
Abstrakt		The ratio between the observed and the expected survival is often used in studies of survival, cancer being the most common example. This ratio is called the relative survival and it is a measure of the net effect of the disease on survival. While the expected survival is in principle, a simple concept, there are some problems as to calculating it in a way that it controls for heterogeneity in the potential follow-up times and yet remains independent of the observed mortality at the salne time. Of the methods proposed the hest approach is that of Hakulinen. Still, all these methods focus on group experience, and say nothing about individual values. For example, they do not answer a very natural question "How long relatively to the general population, did a certain person live?" We introduce an individual measure of relative survival that provides answers to such and similar qustions. But, more importanly, it enables silnple and effective modelling of excess risk via regression models. In this presentation we focus on properties of the measure, and the comparison of our regression approach with those of Hakulinen and Tenkanen (1989), and Andersen et al (1985). The main advantage of our approach is that it accounts for the population risk without assuming anything about its relation to the excess risk. Furthermore, in regression models, variables other than those defining population tables can be used. In fact, the method enables usage of all known regression approaches without restrictions, using only standard statistical software. The approach is illustrated using data on survival of patients after acute myocardial infarction.
Deskriptorji		SURVIVAL ANALYSIS NEOPLASMS AGE FACTORS SEX FACTORS COHORT STUDIES REGRESSION ANALYSIS MODELS, STATISTICAL PROPORTIONAL HAZARDS MODELS