# Survival analysis - help with determining source of error

I am trying to determine the effect of four interventions/exposures upon the acquisition of an infection (the interventions aimed at preventing the infection). I'm using registry data. My analysis time is limited to 4 years as my primary outcome is at 4 years (infection/or not). It is a relatively new registry and so individuals' data have been entered contemporaneously (for those born after the registry started) and other data for which individuals' data commence later, many of whom do not contribute to my analysis as they were older than 4 years of age at inception of the registry.

In STATA I have set out my origin (birth = Time 0), entry (first date entered on registry), and failure (infection).

As acquisition of infection is an absolute failure and my analysis time is age I would expect that numbers at risk of infection should reduce over time. However when I plot a Kaplan-Meier survival curve with an at-risk table, at time 0, 0 individuals are at risk and later individuals become at risk.

Any ideas why this is not the case?

(By the way, I usually add scale(365.25) to the stset statement in my Stata code in order to display analysis time in years rather than days.)
• +1 I usually add the scale() option too, but when I want to express days as years, I use scale(365.24). This is because there are 24 leap years per century. (I know this is probably just a trivial detail, and 365.25 is very often used.) – boscovich Mar 1 '12 at 10:13