Longitudinal analysis advice

[Harris, Betty A.]

Hi All,

I need some suggestions for how to approach for some analyses I need to do.

We have two groups (treatment and control) with a staggered entry into the study.  Groups of participants entered each month throughout the year and were randomly assigned to group.

We have inpatient and outpatient claims data that I've cleaned up to where it I can aggregate length of stay and total changes by month for each participant.

I plan to aggregate the inpatient and outpatient data to reflect 3 month blocks of time starting 6 months before each participant entered the study through for six months after completion of the 12 months of treatment.

I have dose (hours of the treatment and number of treatment contacts) that I can also roll up into the same time intervals as the inpatient and outpatient data.

I also have two other dichotomous indicators that should conceptually be related to the claims data that I can roll up into the same time intervals as the inpatient and outpatient data.

The question is, what would be the best statistical test for this type of data?
I plan to do descriptives and correlations to see how the data look.
My inclination would be a 2 x 8 group by time anova, however the 8 time periods seem troublesome to me in terms of pinpointing trends over time between the groups.

Any suggestions would be appreciated!

Thanks,
Betty Harris
University of Oklahoma E-TEAM
http://eteam.ou.edu/

[Bruce Weaver]

Harris, Betty A. wrote

Hi All,

I need some suggestions for how to approach for some analyses I need to do.

We have two groups (treatment and control) with a staggered entry into the study.  Groups of participants entered each month throughout the year and were randomly assigned to group.

We have inpatient and outpatient claims data that I've cleaned up to where it I can aggregate length of stay and total changes by month for each participant.

I plan to aggregate the inpatient and outpatient data to reflect 3 month blocks of time starting 6 months before each participant entered the study through for six months after completion of the 12 months of treatment.

I have dose (hours of the treatment and number of treatment contacts) that I can also roll up into the same time intervals as the inpatient and outpatient data.

I also have two other dichotomous indicators that should conceptually be related to the claims data that I can roll up into the same time intervals as the inpatient and outpatient data.

The question is, what would be the best statistical test for this type of data?
I plan to do descriptives and correlations to see how the data look.
My inclination would be a 2 x 8 group by time anova, however the 8 time periods seem troublesome to me in terms of pinpointing trends over time between the groups.

Any suggestions would be appreciated!

Thanks,
Betty Harris
University of Oklahoma E-TEAM
http://eteam.ou.edu/

<http://listserv.uga.edu/cgi-bin/wa?LIST=SPSSX-L>

Have you considered multilevel linear regression?  Take a look at the very good book by Singer & Willett (Applied Longitudinal Data Analysis).  You will find SPSS syntax for many of their examples on the UCLA Computing website.

HTH.