Prev: PROC GREPLAY and BY lists
Next: Join data horizontally
From: Dave on 11 Feb 2010 05:45
I am looking to this group for suggestions to improve or suggest a better
way to come up with a solution to my problem.

Background:
I have one record per icd9 code per patient id. I have a list of 50 icd9
codes that I need to see if the patient ever was diagnosed with that code by
creating a new variable (0=no, 1=yes) for each wanted code. Output only one
record per patient id containing only those new 0/1 variables and id.

Example (for sake of brevity, I am only looking for 3 specific icd9 codes,
but in the real run I will be looking for 50):

I am searching the records for icd9 code = 49300, 49301 and 49302 (icd9 is a
character variable). Create another new variable "exclusion" if we found any
of the specific icd9 codes. Output only one record per patient id.

data have;
length id icd9 $5 ;
input id $ icd9 $;
cards;
1 33333
1 22
1 49300
2 356
2 123
3 49321
4 49300
4 49301
4 49302
;
run;

data want;
set have;
by id;

retain dx49300 dx49301 dx49302 0;
if first.id then
do;
dx49300 = 0;
dx49301 = 0;
dx49302 = 0;
end;

if icd9 = '49300' then dx49300 = 1 ;
else if icd9 = '49301' then dx49301 = 1 ;
else if icd9 = '49302' then dx49302 = 1 ;

if sum(of dx49300 dx49301 dx49302) > 0 then exclusion = 1;
else exclusion = 0;

drop icd9;
if last.id then output;
run;

proc print data=want noobs;
run;

As you can see, this can get quite tedious when I need to check for 50
codes. What is a better, more efficient way of solving my problem?

Thanks much.
Dave
 | 
Pages: 1
Prev: PROC GREPLAY and BY lists
Next: Join data horizontally