Dear Statalisters
A non-Stata question, if I may, regarding the appropriate covariate form for
longitudinal analysis.
I am looking at the 28 day mortality of patients admitted to an intensive
care unit (ICU) and am assessing the predictive utility (for both ICU and
hospital mortality) of only one variable called the "lod score" which is a
composite organ failure score recorded daily from ICU admission (day 1)
until discharge from ICU, or death or 28 days in ICU.
In looking at the "lod score" I have used:
1. the raw score
2. differenced or delta score (plus the day 1 score); such that each score
(beyond day 1) is the difference between two consecutive days
3. serial score; difference between each day (for day 2 onwards) and day 1
(which is the baseline)
It would appear that, in this data set, the "serial score" is the _best_
predictor of outcome (for a variety of different models).
Now looking at (more than) a few references on "longitudinal analysis";
there is a paucity of discussion of the virtues of form 1 vs 2 vs 3.
Such is pretty fundamental, but does any-one have some key references on
this specific question?
Much obliged for any responses / references
john moran
John L Moran
Intensive Care Unit
The Queen Elizabeth Hospital
28 Woodville Road Woodville
SA 5011
Australia
Tel 61 08 8222 6463
Fax 61 08 8222 6045
E-mail: [email protected]
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
