Hello,
We are conducting a study in a French area about doctors prescribing Cox
selective NSAID (a drug to prevent inflammation).
So, we selected 2600 doctors with highest number of patients with NSAID
among 5200 general practitioners and rheumatologists (variable spe). We
randomly picked a 1300 sample and mailed 3 successive information to
each of them.
We measured several criterions :
* number of patients with Cox2 selective NSAID (variable named cox)
* number of patients with other classical NSAID (nsaidc variable)
* number of patients with acetaminophen (paracetamol) prescribed for at
least 12 weeks (para variable)
* % of patients with H2 receptor antagonist or proton pump inhibitors
(ulcer prevention) among cox and platelet activation inhibitors (Aspirin
or Plavix) users (ulcer variable) (we know the absolute numbers for
calculating percentage)
* % of patients older than 70 with serum creatinine measure (renal
function) among cox users (creat variable) (we know the absolute numbers
for calculating percentage)
The patients population of interest is cox plus nsaid plus para. We have
no specific information about the patients other than above.
The number of patients may vary from time to time for each doctor
(opened patient population) but the doctors do not vary (closed doctor
population). We have a little bit more information about the doctors
(age, urban or rural...).
So, these data are xt panel. We have three measures (time 1, time2,
time3) of each variable in two groups :
- cox1, cox2, cox3
- nsaidc1 , nsaidc2 , nsaidc3
- para1, para2, para3
- ulcer1 , ulcer2 , ulcer3
- creat1, creat2, creat3
I would like which analysis I should use :
� xi: xtpoisson cox spe i.time, i(doctor) irr � � � � � � � � � � � �for
modelling the count of cox users
� expanding the data and using a hierarchical logit model (gllamm) with
first unit beeing doctors and second unit beeing patients (cox + nsaidc
+ para) and depvar recoded 0/1 for users of Cox2 NSAID (coxuser). But
how can I consider the three time measures.
� reversing the model, id modelling the group instead of the result and
writing something like
� xi: xtlogit group i.spe i.time cox nsaidc para ulcer creat, i(doctor)
The problem is how to use in the model the percentages considering
precision depending on the doctor's patients population.
Is there any other way I could analyse these data?
Thank you very much.
Bien amicalement.
Dr Jean-Fran�ois Blatier
*
* 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/
