Measuring change in a binary outcome following different exposure levels

[Michael Kruger]

I am trying to determine what is the best way to look at a fairly simple
question. I am looking at self-reported survey data concerning pregnancy
outcomes. The women report the number of pregnancies, stillbirths, and
live births before and after a certain year (1991). Some of the women
were exposed to certain environmental toxins prior to this data while
some were not. I would like to be able to get ORs & 95% CIs for each
outcome (Negative, Positive - No, Yes) for both before and after this
date and compare those who were exposed to those who were not and test
for differences in these ORs both before and after 1991 as well as
between Exposed vs . Not Exposed.

So I essentially have a mixed design with a 2 level within factor
(Before and After 1991) and a 2 level between factor (Exposed vs Not
Exposed) and a set of binary outcome variables
(stillbirth, pregnant, live birth - all coded as Negative or Positive).
It seems a simple enough design but I can't figure out how to compare
OR's within individuals and that seems to be the best way of handling
it. Any advice would be appreciated.

Michael Kruger
Statistical Analyst
Dept. of OB/GYN
WSU School of Medicine

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Bruce Weaver]

Hi Michael.  Two options to consider are generalized estimating equations (GEE), via GENLIN, and multilevel logistic regression (with occasion clustered within ID) via GENLINMIXED.

HTH.

Michael Kruger wrote

I am trying to determine what is the best way to look at a fairly simple
question. I am looking at self-reported survey data concerning pregnancy
outcomes. The women report the number of pregnancies, stillbirths, and
live births before and after a certain year (1991). Some of the women
were exposed to certain environmental toxins prior to this data while
some were not. I would like to be able to get ORs & 95% CIs for each
outcome (Negative, Positive - No, Yes) for both before and after this
date and compare those who were exposed to those who were not and test
for differences in these ORs both before and after 1991 as well as
between Exposed vs . Not Exposed.

So I essentially have a mixed design with a 2 level within factor
(Before and After 1991) and a 2 level between factor (Exposed vs Not
Exposed) and a set of binary outcome variables
(stillbirth, pregnant, live birth - all coded as Negative or Positive).
It seems a simple enough design but I can't figure out how to compare
OR's within individuals and that seems to be the best way of handling
it. Any advice would be appreciated.

Michael Kruger
Statistical Analyst
Dept. of OB/GYN
WSU School of Medicine

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD