topic Re: Identifying the number of subjects who had treatment switch in SAS Programming

Identifying the number of subjects who had treatment switch

SASFREAK — Tue, 10 May 2016 15:11:51 GMT

Hi SAS folks,

I am working on some reseach project using hospital visit data. I have to identify the subjects who had switched their treatment. the defination of treatment switch is like:

Treatments are grouped by topical, non-biologic systemic, and biologic. A switching event is defined as the withdrawal of a medication that was not dispensed at the time of the last treatment withdrawal. There is a hierarchy, with biologics at the top, then non-bio systemics, then topical.

If a patient is first observed withdrawing a topical, any withdrawal of a treatment after that to a bio, non-bio systemic, or new topical is a switch
If a patient is first observed withdrawing a non-bio systemic, any withdrawal of a treatment after that to a bio or a new non-bio systemic is a switch
- A withdrawal of any topical is not a switch
If a patient is first observed withdrawing a bio, a switch can only be made to another bio (which is different from the first)
- A withdrawal of any topical or non-bio systemic is not a switch

According to the logic above, once a patient has switched to a higher class of therapy, the lower classes are considered concomitant medication and a withdrawal of a concomitant medication does not indicate a treatment switch.

In nutshell, If a patient moves from lower class of treatment to upper or withing the same class(different ATC code) then it will consider as a Switch.

Treatment hierarchy

Biologics> Non_Bio_Sys>Topical

I am adding sample dataset for the reference, with variable information :

LPNR: unique subjID

EDATUM: prescrition date ( treatment recieved date)

ATC_CODE: codes for each of treament type

'D05AX52' ,
'D05AX02' = PSorasis(Topical)

'L01XC02' ,
'L04AA24',
'L04AB01',
'L04AB02',
'L04AB04',
'L04AB05',
'L04AB06',
'L04AC03',
'L04AC07' =Non_Bio_systemic

'A11CC03' ,
'D05BB02',
'L01BA01',
'L01BB03' ,
'L01XX05' ,
'L04AA21' ,
'L04AD01' ,
'L04AX03' =Biologics

TYPE: Treatment type

It would be great , if you can help me with SAS code to indentify the Subjects with Switches .

Re: Identifying the number of subjects who had treatment switch

DartRodrigo — Tue, 10 May 2016 19:12:59 GMT

Hi mate.

I couldn't understand what you mean exactly but i think you are looking for something to check where are these dependencies changed.

My guess is you can use the LAST. or FIRST. options.

Check the code below:

FILENAME REFFILE "/PATH/sample_data.xlsx" TERMSTR=CR;

PROC IMPORT DATAFILE=REFFILE
	DBMS=XLSX
	OUT=WORK.IMPORT;
	GETNAMES=YES;
RUN;

PROC SORT DATA=IMPORT;
  BY LPNR ATC_CODE TYPE;
RUN;

DATA NEED;
	LENGTH INFO_CHANGE $ 20; 
	SET IMPORT;
	BY LPNR ATC_CODE TYPE;
	IF LAST.LPNR AND LAST.ATC_CODE AND LAST.TYPE THEN INFO_CHANGE = "CHANGED";
	ELSE INFO_CHANGE = "UNCHANGED";
RUN;

The last data step in this code prints the changed if the dependencies occured in the last time they changed.

And unchanged for the rest.

If you need the first time jus switch to FIRST. and checkout the result.

Hope this helps

Re: Identifying the number of subjects who had treatment switch

SASFREAK — Wed, 11 May 2016 04:21:31 GMT

@DartRodrigo Hi , Thanks for the quick reply. I am sorry for the unclear explaination, but what i exactly means is:

as you see the data, we have three treatment type in TYPE variable. Now for each Type variable we have different ATC_CODEs. I need to identify the event (SWITCH) , whenever subject changes his treatment type from Lower treatment type( i,e Topical) to Higher treatment group(i,e Non_bio_sys or Biologics) during his treatment period. Same is true for Non_sys_Bio , if Subject changes to Biologics, then it will also consider as Switch. Event is also to be counted whenever subjects changes his ATC_CODEs within his treatment TYPE , ie in case, when Treatment TYPE is "Psorasis" and if Its' ATC_CODE changes , then we have to identify it as SWITCH.

same is true for other treatment arm.

order of treatment: BIOLOGICS> NON_BIO_SYS> TOPICAL.

Hope i am able to clear you doubt. please help. thanks

Re: Identifying the number of subjects who had treatment switch

ramchinna24 — Wed, 11 May 2016 08:24:28 GMT

data have;
length TYPE $20.;
input pnr ATC_CODE $ type $ EDATUM $ 10. ;
date=input(edatum,mmddyy10.);
format date mmddyy10.;
cards;
478 L04AX03 Nonbiologic_systemic 2/10/2010
478 D05AX52 Psoriasis 2/23/2010
478 L04AX03 Nonbiologic_systemic 3/5/2010
478 L04AX03 Nonbiologic_systemic 5/18/2010
478 L04AX03 Nonbiologic_systemic 8/2/2010
478 L04AX03 Nonbiologic_systemic 11/3/2010
478 L04AX03 Nonbiologic_systemic 2/7/2011
478 L04AX03 Nonbiologic_systemic 8/16/2011
478 L04AX03 Nonbiologic_systemic 11/22/2011
603 D05AX02 Psoriasis 8/24/2005
603 D05AX02 Psoriasis 10/13/2005
603 D05AX02 Psoriasis 6/2/2006
603 D05AX02 Psoriasis 7/19/2006
603 D05AX02 Psoriasis 4/3/2007
603 D05AX02 Psoriasis 12/7/2007
603 D05AX02 Psoriasis 1/4/2008
603 D05AX02 Psoriasis 1/4/2008
603 D05AX02 Psoriasis 10/2/2008
603 D05AX52 Psoriasis 3/16/2009
603 D05AX52 Psoriasis 3/16/2009
603 D05AX52 Psoriasis 5/7/2009
603 D05AX52 Psoriasis 8/21/2009
603 D05AX52 Psoriasis 11/9/2009
603 D05AX52 Psoriasis 3/19/2010
603 D05AX52 Psoriasis 7/30/2010
603 D05AX52 Psoriasis 7/30/2010
603 D05AX52 Psoriasis 1/13/2011
603 D05AX52 Psoriasis 5/3/2011
603 D05AX52 Psoriasis 12/12/2011
1103 L04AX03 Nonbiologic_systemic 1/17/2006
1103 L04AX03 Nonbiologic_systemic 8/23/2006
1103 L04AX03 Nonbiologic_systemic 11/9/2006
1103 D05AX52 Psoriasis 4/19/2007
1103 L04AX03 Nonbiologic_systemic 4/19/2007
1103 D05AX52 Psoriasis 10/26/2007
1103 D05AX52 Psoriasis 4/1/2008
1103 L04AX03 Nonbiologic_systemic 10/20/2008
1103 D05AX52 Psoriasis 10/20/2008
1103 L04AX03 Nonbiologic_systemic 4/7/2009
1103 L04AX03 Nonbiologic_systemic 11/20/2009
1103 L04AX03 Nonbiologic_systemic 11/12/2010
1103 L04AX03 Nonbiologic_systemic 8/12/2011
1103 D05AX52 Psoriasis 10/22/2011
1103 D05AX52 Psoriasis 12/29/2011
;
run;

/*sort the data to keep the date in increase order to know subject moved to other treatment or not*/

proc sort data=have;
by pnr date;
run;

data need;
set have;
length change $ 10.;
by pnr;
/*create new variable change and keep constant if it is topical for first observation*/
/*compare topical values with next values if it changes to other non bio or bio then switched*/

if first.pnr and atc_code in ( 'D05AX52','D05AX02') then change="";
else if not first.pnr and atc_code not in('D05AX52','D05AX02','L01XC02','L04AA24','L04AB01','L04AB02','L04AB04','L04AB05','L04AB06','L04AC03','L04AC07','A11CC03','D05BB02','L01BA01','L01BB03' ,'L01XX05' ,'L04AA21' ,'L04AD01' ,'L04AX03')
then change='switched';

/*compare non bio values with next values if it changes to other non bio or bio then switched*/
if first.pnr and atc_code in ('L01XC02','L04AA24','L04AB01','L04AB02','L04AB04','L04AB05','L04AB06','L04AC03','L04AC07') then change="";
else if not first.pnr and atc_code not in( 'D05AX52','D05AX02') and atc_code in('L01XC02','L04AA24','L04AB01','L04AB02','L04AB04','L04AB05','L04AB06','L04AC03','L04AC07','A11CC03','D05BB02','L01BA01','L01BB03' ,'L01XX05' ,'L04AA21' ,'L04AD01' ,'L04AX03')
then change='switched';

/*compare bio with next values if it changes to other bio then switched*/

if first.pnr and atc_code in('A11CC03','D05BB02','L01BA01','L01BB03' ,'L01XX05' ,'L04AA21' ,'L04AD01' ,'L04AX03') then change='';
else if not first.pnr and atc_code not in( 'D05AX52','D05AX02') and atc_code in('A11CC03','D05BB02','L01BA01','L01BB03' ,'L01XX05' ,'L04AA21' ,'L04AD01' ,'L04AX03') then change='switched';
if atc_code=lag(atc_code) then change="";

run;

Hope this may help you 🙂

Re: Identifying the number of subjects who had treatment switch

SASFREAK — Wed, 11 May 2016 08:44:58 GMT

@ramchinna24 Hi, Thanks for the logic, but what if ATC code withing the psorasis or within other treatment group changes, we also need to identify those changes as switches.

i want this :

lpnr	ATC_CODE	TYPE	EDATUM	Flag_switch	Comment
478	L04AX03	Nonbiologic_systemic	2/10/2010
478	D05AX52	Psoriasis	2/23/2010
478	L04AX03	Nonbiologic_systemic	3/5/2010	Y	This is switch because subject is changing his treatment from lower(psorais) to higher treatment(non_bio_sys/bio)
478	L04AX03	Nonbiologic_systemic	5/18/2010
478	L04AX03	Nonbiologic_systemic	8/2/2010
478	L04AX03	Nonbiologic_systemic	11/3/2010
478	L04AX03	Nonbiologic_systemic	2/7/2011
478	L04AX03	Nonbiologic_systemic	8/16/2011
478	L04AX03	Nonbiologic_systemic	11/22/2011
603	D05AX02	Psoriasis	8/24/2005
603	D05AX02	Psoriasis	10/13/2005
603	D05AX02	Psoriasis	6/2/2006
603	D05AX02	Psoriasis	7/19/2006
603	D05AX02	Psoriasis	4/3/2007
603	D05AX02	Psoriasis	12/7/2007
603	D05AX02	Psoriasis	1/4/2008
603	D05AX02	Psoriasis	10/2/2008
603	D05AX52	Psoriasis	3/16/2009	Y	This is switch because subject is changing his ATC code within Psorasis(or any other treatment group)
603	D05AX52	Psoriasis	5/7/2009
603	D05AX52	Psoriasis	8/21/2009
603	D05AX52	Psoriasis	11/9/2009
603	D05AX52	Psoriasis	3/19/2010
603	D05AX52	Psoriasis	7/30/2010
603	D05AX52	Psoriasis	1/13/2011
603	D05AX52	Psoriasis	5/3/2011
603	D05AX52	Psoriasis	12/12/2011
1103	L04AX03	Nonbiologic_systemic	1/17/2006
1103	L04AX03	Nonbiologic_systemic	8/23/2006
1103	L04AX03	Nonbiologic_systemic	11/9/2006
1103	L04AX03	Nonbiologic_systemic	4/19/2007
1103	D05AX52	Psoriasis	10/26/2007
1103	D05AX52	Psoriasis	4/1/2008
1103	L04AX03	Nonbiologic_systemic	10/20/2008	Y	This is switch because subject is changing his treatment from lower(psorais) to higher treatment(non_bio_sys/bio)
1103	D05AX52	Psoriasis	10/20/2008
1103	L04AX03	Nonbiologic_systemic	4/7/2009	Y	This is switch because subject is changing his treatment from lower(psorais) to higher treatment(non_bio_sys/bio)
1103	L04AX03	Nonbiologic_systemic	11/20/2009
1103	L04AX03	Nonbiologic_systemic	11/12/2010
1103	L04AX03	Nonbiologic_systemic	8/12/2011
1103	D05AX52	Psoriasis	10/22/2011
1103	D05AX52	Psoriasis	12/29/2011

Re: Identifying the number of subjects who had treatment switch

ramchinna24 — Wed, 11 May 2016 08:42:10 GMT

even if code changes with in the group, it works. Just try it once. hope it works. If it doesnt work, tell me .

Re: Identifying the number of subjects who had treatment switch

SASFREAK — Wed, 11 May 2016 08:58:21 GMT

@ramchinna24 Hi , code is working on treatment Type level, however whenever atc_codes are changing withing the treatment type, it's not working. I have added the desired output in previous reply, please kindly check.

Many thanks for your's help!!

Re: Identifying the number of subjects who had treatment switch

ramchinna24 — Wed, 11 May 2016 09:05:20 GMT

proc sort data=have;
by pnr atc_code type date;
run;

data need;
set have;
length change $ 10.;
by pnr atc_code;
/*create new variable change and keep constant if it is topical for first observation*/
/*compare topical values with next values if it changes to other non bio or bio then switched*/

if first.pnr and atc_code in ( 'D05AX52','D05AX02') then change="";
else if not first.pnr and atc_code in('D05AX52','D05AX02','L01XC02','L04AA24','L04AB01','L04AB02','L04AB04','L04AB05','L04AB06','L04AC03','L04AC07','A11CC03','D05BB02','L01BA01','L01BB03' ,'L01XX05' ,'L04AA21' ,'L04AD01' ,'L04AX03')
then change='switched';
/*compare non bio values with next values if it changes to other non bio or bio then switched*/
if first.pnr and atc_code in ('L01XC02','L04AA24','L04AB01','L04AB02','L04AB04','L04AB05','L04AB06','L04AC03','L04AC07') then change="";
else if not first.pnr and atc_code not in( 'D05AX52','D05AX02') and atc_code in('L01XC02','L04AA24','L04AB01','L04AB02','L04AB04','L04AB05','L04AB06','L04AC03','L04AC07','A11CC03','D05BB02','L01BA01','L01BB03' ,'L01XX05' ,'L04AA21' ,'L04AD01' ,'L04AX03')
then change='switched';
/*compare bio with next values if it changes to other bio then switched*/
if first.pnr and atc_code in('A11CC03','D05BB02','L01BA01','L01BB03' ,'L01XX05' ,'L04AA21' ,'L04AD01' ,'L04AX03') then change='';
else if not first.pnr and atc_code not in( 'D05AX52','D05AX02','L01XC02','L04AA24','L04AB01','L04AB02','L04AB04','L04AB05','L04AB06','L04AC03','L04AC07') and atc_code in('A11CC03','D05BB02','L01BA01','L01BB03' ,'L01XX05' ,'L04AA21' ,'L04AD01' ,'L04AX03') then change='switched';
if atc_code=lag(atc_code) then change="";

run;

Re: Identifying the number of subjects who had treatment switch

ramchinna24 — Wed, 11 May 2016 09:12:55 GMT

Did u try last updated solution? I sent updated code and it works for you

Re: Identifying the number of subjects who had treatment switch

SASFREAK — Wed, 11 May 2016 09:37:03 GMT

@ramchinna24 I think, sorting order is conflicting with data, for example on: PNR 478, Switch was recorded at 5th March 2010 ,

Nonbiologic_systemic	2/10/2010
Psoriasis	2/23/2010
Nonbiologic_systemic	3/5/2010

but when we are sorting it on atc_code and then type and followed by edatum , switch is recording at 2/10/2010.

Re: Identifying the number of subjects who had treatment switch

SASFREAK — Wed, 11 May 2016 16:27:14 GMT

@ramchinna24 Hi I have tried this:

data test;

    set new_have;/*sorted on pnr date*/
    by pnr;
    if lag(type)='Psoriasis' and type in ('Nonbiologic_systemic' 'Biologics')   then f=1;
    if lag(type)='Psoriasis' and type='Psoriasis' and lag(atc_code) ne atc_code then f=1;

    if lag(type)='Nonbiologic_systemic' and type in ('Biologics') then f=1;
    if lag(type)='Nonbiologic_systemic' and type='Nonbiologic_systemic' and lag(atc_code) ne atc_code then f=1;

   if lag(type) in ('Nonbiologic_systemic' 'Psoriasis') and type='Biologics' then f=1;
   if lag(type)='Biologics' and type='Biologics' and lag(atc_code) ne atc_code then f=1;

/*    if lag(pnr) =pnr and lag(atc_code) ne atc_code then f=1; */

run;

But it's ignoring the by group and adding flagging for situations when patient by group is changes.

Re: Identifying the number of subjects who had treatment switch

ramchinna24 — Thu, 12 May 2016 04:25:12 GMT

proc format;
value treatType 1 = "Topical"
2 = "Non-biologic systemics"
3 = "Biologics";
run;

data have;
format pnr 8. tType treatType. tCode $10. date YYMMDDs8.;
informat tType 1. date yymmdd10.;
input pnr tCode tType date;
datalines;
478 L04AX03 2 2010-02-10
478 D05AX52 1 2010-02-23
478 L04AX03 2 2010-03-05
478 L04AX03 2 2010-05-18
478 L04AX03 2 2010-08-02
478 L04AX03 2 2010-11-03
478 L04AX03 2 2011-02-07
478 L04AX03 2 2011-08-16
478 L04AX03 2 2011-11-22
603 D05AX02 1 2005-08-24
603 D05AX02 1 2005-10-13
603 D05AX02 1 2006-06-02
603 D05AX02 1 2006-07-19
603 D05AX02 1 2007-04-03
603 D05AX02 1 2007-12-07
603 D05AX02 1 2008-01-04
603 D05AX02 1 2008-01-04
603 D05AX02 1 2008-10-02
603 D05AX52 1 2009-03-16
603 D05AX52 1 2009-03-16
603 D05AX52 1 2009-05-07
603 D05AX52 1 2009-08-21
603 D05AX52 1 2009-11-09
603 D05AX52 1 2010-03-19
603 D05AX52 1 2010-07-30
603 D05AX52 1 2010-07-30
603 D05AX52 1 2011-01-13
603 D05AX52 1 2011-05-03
603 D05AX52 1 2011-12-12
1103 L04AX03 2 2006-01-17
1103 L04AX03 2 2006-08-23
1103 L04AX03 2 2006-11-09
1103 D05AX52 1 2007-04-19
1103 L04AX03 2 2007-04-19
1103 D05AX52 1 2007-10-26
1103 D05AX52 1 2008-04-01
1103 L04AX03 2 2008-10-20
1103 D05AX52 1 2008-10-20
1103 L04AX03 2 2009-04-07
1103 L04AX03 2 2009-11-20
1103 L04AX03 2 2010-11-12
1103 L04AX03 2 2011-08-12
1103 D05AX52 1 2011-10-22
1103 D05AX52 1 2011-12-29
run;
proc sort data=have;
by pnr date;
run;

data want;
set have;
by pnr;
format switch 1. tCode_prev $10. tType_prev 3.;
retain tCode_prev '' tType_prev .;
format switch 1.;

if first.pnr then do;
switch = 0;
end;

else if tType_prev = 1 then do;
if tType > 1 or tCode ne tCode_prev then switch = 1;
else switch = 0;
end;

else if tType_prev = 2 then do;
if tType > 2 or (tType = 2 and tCode ne tCode_prev) then switch = 1;
else switch = 0;
end;

else if tType_prev = 3 then do;
/* Here's your homework */
end;

output;
tCode_prev = tCode;
tType_prev = tType;

drop tCode_prev tType_prev;
run;

Re: Identifying the number of subjects who had treatment switch

SASFREAK — Mon, 16 May 2016 09:42:47 GMT

@ramchinna24 Many Thanks for the help and time! . It worked