Calculating cumulative incidence of multiple simultaneous events within groups

[Jayne Williams]

l am analysing a database of 1850 patients with longitudinal data on anything from 1 to 11 years of annual follow ups. I am interested in calculating the cumulative incidence (CI) of 42 different 'damage' events e.g. cataracts, fractures etc. within the whole cohort. As well as calculating the CI of individual items, I also want to calculate the CI of items within groups (i.e. how many events have occurred within 5 or 10 items). I had done this by generating a variable (new_group_event) which was 1 if any new item had occurred within that group of items and then stset-ing the data with the new_group_event as the failure event. I then calculated the CI manually by subtracting the survivor function from 1 in sts list. The problem with this is that there is no differentiation between one , two or more items occurring simultaneously (at a follow up assessment) within a group. I could of course calculate the CI for each of the 42 items and then sum them for the groupings but this does not allow reproducibility. Any ides on how I could get around this?

[Steve Samuels]

Events take place in the year prior to the annual followup, so quite possibly you need a grouped data analysis. Leaving this issue aside, why does summing the CIs for individual events "not allow reproducibility"?