topic Identifying which individual observations are events in PROC PHREG to calculate person-time in SAS Programming

Identifying which individual observations are events in PROC PHREG to calculate person-time

jb915 — Thu, 14 Jul 2022 22:28:12 GMT

PROC PHREG data=XXX covs(aggregate);

CLASS [all categorical variables]

MODEL (exact age, exact age at hysterectomy)*cancer_event(0)=exposure v1 v2 v3 v4 v5;

ID [observation identifier];

HAZARDRATIO exposure / UNITS=10;

RUN;

Based on this model, I am trying to calculate person-time. When I run this model, I am able to see the the number of events in the 'Summary of the Number of Event and Censored Values' table, but I am not quite sure about how to identify which observations were considered to be events.

I think that if I was able to identify which observations are events that I would be able to calculate person time with a DATA step to subset the data and create a person-time variable before running PROC MEANS to sum all the observations for person-time. Any ideas about how to best calculate person time based on this model would be greatly appreciate.

Re: Identifying which individual observations are events in PROC PHREG to calculate person-time

ballardw — Thu, 14 Jul 2022 23:09:05 GMT

I think that if you add an OUTPUT statement you will see that the number of Cancer_event (censoring variable) values = 1 may match the number of events and 0 the number censored in the output set.

At least it is easy to test. Add

output out=work.somedatasetname;

Re: Identifying which individual observations are events in PROC PHREG to calculate person-time

jb915 — Thu, 14 Jul 2022 23:37:45 GMT

I'm not able to get the person-time calculations for the events only included in PROC PHREG. There are 403 events used in the PROC PHREG statement but I am getting the person-time calculation for all 404 cases.

Re: Identifying which individual observations are events in PROC PHREG to calculate person-time

jb915 — Fri, 15 Jul 2022 00:21:53 GMT

Is there a better way to display which observations were censored vs. which observations were events?