How to assess switch from one value to another

[lillymaginta]

id       rx1     rx_date             h_date 
1       a         1/1/2005           1/20/2005
1       b         2/2/2005
1       c         3/3/2005
2       a         1/2/2003           1/12/2003
2       a         2/2/2003
2       a         4/4/2003
3       b         1/2/2005           1/18/2005 
3       c         2/3/2005
3       b         3/4/2005

I have a data similar to this. Patients can have multiple drug therapy rx1 in the same rx_Date or in different rx_date, I wanted to assess the patterns of change. For example I wanted to know how many patients switched one drug to another after h_date. Patient 1 was on a then after h_Date switched to b, patient 2 was on a and cont on a after h_Date. 

[Ksharp]

Your data is not completed . There are lots of missing value . 

And what output do you want ?

[lillymaginta]

There is only 1 h_date per patient. I want the proprtion of patient who change rx1 (in a time period of 30 days) after h_date

proprtion of those who remain on a after h_date (patient 2), those who change from a to b (patient a) and so on. 

[Ksharp · Accepted solution]

OK. There is only one h_Date for each patient.

data have;
input id       rx1   $  (rx_date             h_date) (: mmddyy10.); 
format rx_date             h_date mmddyy10.;
cards;
1       a         1/1/2005           1/20/2005
1       b         2/2/2005 .
1       c         3/3/2005 .
2       a         1/2/2003           1/12/2003
2       a         2/2/2003 .
2       a         4/4/2003 .
3       b         1/2/2005           1/18/2005 
3       c         2/3/2005 .
3       b         3/4/2005 .
;
run;
data temp;
 set have(keep=id rx1 rx_date ) 
     have(keep=id rx1 h_date rename=(h_date=rx_date) where=(rx_date is not missing) in=inb);
 by id rx_date ;
 if inb=0 and lag(inb)=1 or first.id;
run;
proc sql;
select * 
 from temp
  group by id
   having count(distinct rx1) gt 1
    order by id,rx_date ;
quit;

[lillymaginta]

Thank you!