topic Re: How to identify the last record? in SAS Programming

How to identify the last record?

Ivy — Fri, 16 Aug 2019 16:34:48 GMT

Dear friends,

I have a data as below.

I try to derive patient's treatment line name based on the algorithm in "Newline" column of the table below:

1. Start treatment: 1st line, 20 days drugs used in this period is the line name

2. Advance line: a: treatment Gap > 3 months or b: a new Chemo drug after 20 days of treatment or C: A new bio drug advance or not after 20 days of treatment based on the condition listed in the table

The complex part is to code the part when I see a new BIO. I am wondering whether there is good solution.

Thank you very much!

ID	DrugName	Dt	DrugCat	LineName	LineNumber	NewLine
1	A	1/1/2019	Chemo	A	1	Line name is generated based on 20 days period when patient starting treatment
1	B	1/12/2019	Chemo	A+B	1
1	A	1/26/2019	Chemo	A+B	1
1	B	1/27/2019	Chemo	A+B	1
1	A	2/20/2019	Chemo	A+B	1
1	B	02/29/2019	Chemo	A+B	1
1	A	02/30/2019	Chemo	A+B	1
1	D	3/15/2019	Bio		1	Not new line When see a new BIO drug after 20 days of starting treatment A, we will change the line based on the drugs applied in 28 days period of after seeing BIO: 1. New chemo comparing previous line name which is retaining from above or 2. Without previous line drugs For this case, we see previous line drug, and, without new chemo, so keep previous line number but add D, the new BIO, to line name A+B so to change to A+B+C
1	A	3/30/2019	Chemo		1
1	D	4/5/2019	Bio	A+B+C	1	MAKE DECISION
1	A	4/15/2019	Chemo	A+B+C	1
1	A	5/5/2019	Chemo	A+B+C	1
1	C	9/9/2019	Bio		2	New line, because during 28 days period, no new chemo, and, stopping all previous line drugs
1	C	9/19/2019	Bio	C	2	MAKE DECISION
1	C	10/20/2019	Bio	C	2
1	C	11/5/2019	Bio	C	2
1	C	12/1/2020	Bio	C	3	New line: Due to treatment GAP > 6 months
1	C	12/2/2020	Bio	C	3
1	C	12/3/2020	Bio	C	3
1	F	12/29/2020	Chemo	C	4	New line: Due to new chemo
1	F	12/30/2020	Chemo	C	4

Re: How to identify the last record?

novinosrin — Fri, 16 Aug 2019 01:10:48 GMT

Hello @Ivy This solution is just purely for fun as i wanted some excitement, I'll prolly post the simple one later.

data have;
infile cards truncover;
input ID	LineName :$75.	samedaydrugs :$75.	AdministeredDate :date9.	RefDt :date9.;
format AdministeredDate 	RefDt date9.;
cards;
1	.	rituximab+gemcitabine	9-Sep-15	.
1	.	rituximab+gemcitabine	29-Sep-15	.
1	.	rituximab+gemcitabine	30-Sep-15	.
1	rituximab+gemcitabine	rituximab+gemcitabine	6-Oct-15	.
1	.	crizotinib	2-Nov-15	29-Nov-15
1	.	rituximab	16-Nov-15	29-Nov-15
1	.	rituximab	23-Nov-15	29-Nov-15
1	.	rituximab	30-Nov-15	.
1	.	rituximab	7-Dec-15	.
;


data want;
if _n_=1 then do;
	if 0 then set have(rename=(AdministeredDate=_d));
   dcl hash H (dataset:'have(rename=(AdministeredDate=_d))',ordered: "A") ;
   h.definekey  ("id","_d") ;
   h.definedata ("id","_d") ;
   h.definedone () ;
   dcl hiter hi('h');
end;
do until(last.id);
 set have ;
 by id;
 if samedaydrugs='crizotinib' then do;
  do _rc=hi.setcur(key:id,key:AdministeredDate) by 0 while(_rc=0);
    if intck('days',AdministeredDate,_d)<28 then _iorc_=_d;
	else leave;
    _rc = hi.next();
  end;
 end;
 combo=ifc(_iorc_=AdministeredDate,catx('+','crizotinib',samedaydrugs),' ');
 output;
end;
call missing(_iorc_, _d);
drop _:;
run;

Re: How to identify the last record?

novinosrin — Fri, 16 Aug 2019 00:55:20 GMT

data have;
infile cards truncover;
input ID	LineName :$75.	samedaydrugs :$75.	AdministeredDate :date9.	RefDt :date9.;
format AdministeredDate 	RefDt date9.;
cards;
1	.	rituximab+gemcitabine	9-Sep-15	.
1	.	rituximab+gemcitabine	29-Sep-15	.
1	.	rituximab+gemcitabine	30-Sep-15	.
1	rituximab+gemcitabine	rituximab+gemcitabine	6-Oct-15	.
1	.	crizotinib	2-Nov-15	29-Nov-15
1	.	rituximab	16-Nov-15	29-Nov-15
1	.	rituximab	23-Nov-15	29-Nov-15
1	.	rituximab	30-Nov-15	.
1	.	rituximab	7-Dec-15	.
;

data want;
do _n_=1 by 1 until(last.id);
 set have ;
 by id;
 array t(999999) _temporary_;
 if samedaydrugs='crizotinib' then _f=1;
 if _f then t(_n_)=AdministeredDate;
end;
do _n_=1 to _n_;
 set have ;
 if samedaydrugs='crizotinib' then do;
 _iorc_=AdministeredDate;
  do while(intck('day',AdministeredDate,_iorc_)<27);
   _iorc_=intnx('day',_iorc_,1);
   if _iorc_ in t then _j=_iorc_;
  end;
 end;
 combo=ifc(_j=AdministeredDate,catx('+','crizotinib',samedaydrugs),' ');
 output;
end;
call missing(_iorc_,of t(*));
drop _:;
run;

Re: How to identify the last record?

novinosrin — Fri, 16 Aug 2019 01:38:34 GMT

Sorry, ignore the previous

data have;
infile cards truncover;
input ID	LineName :$75.	samedaydrugs :$75.	AdministeredDate :date9.	RefDt :date9.;
format AdministeredDate 	RefDt date9.;
cards;
1	.	rituximab+gemcitabine	9-Sep-15	.
1	.	rituximab+gemcitabine	29-Sep-15	.
1	.	rituximab+gemcitabine	30-Sep-15	.
1	rituximab+gemcitabine	rituximab+gemcitabine	6-Oct-15	.
1	.	crizotinib	2-Nov-15	29-Nov-15
1	.	rituximab	16-Nov-15	29-Nov-15
1	.	rituximab	23-Nov-15	29-Nov-15
1	.	rituximab	30-Nov-15	.
1	.	rituximab	7-Dec-15	.
;

data want;
do _n_=1 by 1  until(last.id);
 set have ;
 by id;
 if samedaydrugs='crizotinib' then do; _d= AdministeredDate;_f=1;end;
 else if _f then if intck('days',_d,AdministeredDate)<28  then  __d=AdministeredDate;
end;
do _n_=1 to _n_;
 set have;
 by id;
 combo=ifc(__d=AdministeredDate,catx('+','crizotinib',samedaydrugs),' ');
 output;
end;
drop _:;
run;

Re: How to identify the last record?

novinosrin — Fri, 16 Aug 2019 02:07:15 GMT

SQL:

data have;
infile cards truncover;
input ID	LineName :$75.	samedaydrugs :$75.	AdministeredDate :date9.	RefDt :date9.;
format AdministeredDate 	RefDt date9.;
cards;
1	.	rituximab+gemcitabine	9-Sep-15	.
1	.	rituximab+gemcitabine	29-Sep-15	.
1	.	rituximab+gemcitabine	30-Sep-15	.
1	rituximab+gemcitabine	rituximab+gemcitabine	6-Oct-15	.
1	.	crizotinib	2-Nov-15	29-Nov-15
1	.	rituximab	16-Nov-15	29-Nov-15
1	.	rituximab	23-Nov-15	29-Nov-15
1	.	rituximab	30-Nov-15	.
1	.	rituximab	7-Dec-15	.
;


proc sql;
create table want(drop=t:) as
select *, ifc(max( AdministeredDate)=AdministeredDate and t1,catx('+','crizotinib',samedaydrugs),' ') as Combo
from
(select *,max((samedaydrugs='crizotinib')* AdministeredDate) as t, intck('days',calculated t,AdministeredDate)<28 as t1
from have
group by id)
group by id, t1
order by id, AdministeredDate;
quit;

Re: How to identify the last record?

Ivy — Fri, 16 Aug 2019 15:19:35 GMT

Hello, @novinosrin

Thank you so much for your help.

This works for this case.

However, in my real data situation, this part is part of my data processing to generate the LineName, there are other conditions lining up with this, which I should process together in one data processing step. I tried to identify the last date of this period and then output the combination drugs in one condition such as : If (ref data - AdministeredDate) <= 28 days then do: ---- end; Maybe there is no way to do it ? Thank you. I am guessing I am confusing you.....

Re: How to identify the last record?

novinosrin — Fri, 16 Aug 2019 15:36:27 GMT

Hi @Ivy To your " there are other conditions lining up with this, which I should process together in one data processing step. ". Well this sounds like your requirement is much more. However, I wouldn't know. But if i can guess those are similar, you could perhaps modify along the same lines. So, what's in your real data and the real requirement is something you and your org alone have control over.

My general opinion/approach would be:

1. If there are too many conditions making it too tedious let alone complex, I 'd split and approach in piecemeal.

2. It's also easier to maintain, modify and update

3. The high end processing involving very advance techniques might surface but the trade off should be considered what you prioritize.

4. A lot of times, the kind of fancy solutions I post or take from others here on the community are not always well comprehended by other users in my team.

5. First break it down, get a working solution, and then see if you can consolidate into fewer steps if possible

Re: How to identify the last record?

Ivy — Fri, 16 Aug 2019 16:15:35 GMT

Hi, @novinosrin,

Thank you very much for your suggestions. I really appreciate your help.

I edit the initial post and list a pretty completed data picture,

I am stuck in for a few days. 😞

Thanks again
,
Ivy