Date:         Thu, 2 May 2002 13:38:29 -0400
Reply-To:     "Huang, Ya" <ya.huang@PFIZER.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "Huang, Ya" <ya.huang@PFIZER.COM>
Subject:      Re: Adverse Event - Problem
Comments: To: Mohan Menon <mohmenin@YAHOO.COM>
Content-Type: text/plain

Mohan,

Combine the number of rel and sev into one,
give rel a bigger weight. Assuming the higher the
number of sev, the severer is the severity.
Also holds true for relationship. For example,
if the relation code can be 1,2,3, where 3 is the
closest related to drug; severity code can be 1,2,3,4
where 4 is the most severe AE, then we can combine
these two number as:

comb=rel*10+sev

now if you have four cases of AE and all of them
have rel=2, the combined number will be 21,22,23,24,
all we need to do is to pick the largest number 24,
which is the most severe AE. If you have 4 obs,
where 3 of them are rel=2, one has rel=3, and sev=1,
the largest comb=31, and that's what we want.

Since your sample data used formatted sev and rel,
I have to use my own proc format to convert them back
to numeric. As you mentioned, your sev and rel are
already numeric, you only need to make sure they
are defined the correct way, otherwise, you may need
a transform.


proc format;
invalue sev
'mod'=1
'sev'=2
;
invalue rel
'unreltd'=1
'reltd'=2
;

data ae;
input Pat primary $ bodysys $ severe $ relation $;
nsev=input(severe,sev.);
nrel=input(relation,rel.);
ncomb=nrel*10+nsev;
cards;
1  rash  skin  mod  unreltd
1  rash  skin  sev  reltd
1  rash  skin  sev  unreltd
1  rash  skin  mod  reltd
2  rash  skin  mod  unreltd
2  rash  skin  sev  unreltd
;

proc sort;
by pat ncomb;

data ae;
 set ae;
by pat ncomb;
if last.pat;

proc print;
run;
---------------------
Obs    Pat    primary    bodysys    severe    relation    nsev    nrel
ncomb

 1      1      rash       skin       sev      reltd         2       2
22
 2      2      rash       skin       sev      unreltd       2       1
12

HTH

Ya Huang

-----Original Message-----
From: Mohan Menon [mailto:mohmenin@YAHOO.COM]
Sent: Thursday, May 02, 2002 9:39 AM
To: SAS-L@LISTSERV.UGA.EDU
Subject: Adverse Event - Problem


Hi Everybody
    I have an issue that I am trying to resolve,
please help.  Here is the situation

Pat       primary       bodysys    severe    relation

1          rash          skin        mod       unreltd
1          rash          skin        sev        reltd
1          rash          skin        sev       unreltd
1          rash          skin        mod       reltd

The condition is, if a subject has multiple occurences
of a primary term, then the event with the most severe
intensity within the strongest relationship to study
drug should be included.  Relationship is the first
criteria for selection followed by severity.
Therefore I want to pick only the second observation.

I am not quite sure on how to accomplish this.  Both
severity and intensity are numeric variables with
formats attached.

Thanks for your help.

Mohan Menon

__________________________________________________
Do You Yahoo!?
Yahoo! Health - your guide to health and wellness
http://health.yahoo.com