How to create a new variable based on a series of variable?

[c8826024]

I have a dataset like this:

id	Hospital_dt	T1_date	T2_date	T3_date	T4_date	T5_date
1	3/4/2011	1/1/2011	3/2/3011	3/5/2011
2	1/1/2011	1/1/2011	2/1/2011
3	2/15/2011
4	10/1/2010	8/1/2010	9/1/2010

I want to create a new variable which will show the data of the first treatment AFTER hospital date. It will look like this:

id	Hospital_dt	T1_date	T2_date	T3_date	T4_date	T5_date	first_Treatment_dt
1	3/4/2011	1/1/2011	3/2/3011	3/5/2011			3/5/2011
2	1/1/2011	1/1/2011	2/1/2011				1/1/2011
3	2/15/2011
4	10/1/2010	8/1/2010	9/1/2010

[Tom]

How about using an array?

options ls=80;
data have;
  infile cards dlm='|' dsd truncover;
  input id  (Hospital_dt T1_date T2_date T3_date T4_date T5_date) (:mmddyy10.);
  format _all_ yymmdd10. id ;
cards;
1|3/4/2011|1/1/2011|3/2/3011|3/5/2011||
2|1/1/2011|1/1/2011|2/1/2011|||
3|2/15/2011|||||
4|10/1/2010|8/1/2010|9/1/2010|
run;
data want;
  set have ;
  array tdate t1_date t2_date t3_date t4_date t5_date ;
  do over tdate ;
    if tdate >= hospital_dt then first_treatment_dt = min(first_treatment_dt,tdate);
  end;
  format first_treatment_dt yymmdd10.;
run;

                                                                               f
                                                                               i
                                                                               r
                                                                               s
                                                                               t
                                                                               _
                                                                               t
             H                                                                 r
             o                                                                 e
             s                                                                 a
             p                                                                 t
             i          T          T          T          T          T          m
             t          1          2          3          4          5          e
             a          _          _          _          _          _          n
             l          d          d          d          d          d          t
O            _          a          a          a          a          a          _
b i          d          t          t          t          t          t          d
s d          t          e          e          e          e          e          t
1 1 2011-03-04 2011-01-01 3011-03-02 2011-03-05          .          . 2011-03-05
2 2 2011-01-01 2011-01-01 2011-02-01          .          .          . 2011-01-01
3 3 2011-02-15          .          .          .          .          .          .
4 4 2010-10-01 2010-08-01 2010-09-01          .          .          .          .

[c8826024]

That's quick and neat.

One more question, if I know each location where the patience was treated. But I only need the first location AFTER hospital date:

id	Hospital_dt	T1_date	T2_date	T3_date	T4_date	T5_date	Location1	Location2	Location3	first_Treatment_dt	location
1	3/4/2011	1/1/2011	3/2/3011	3/5/2011			a	b	c	3/5/2011	c
2	1/1/2011	1/1/2011	2/1/2011				b	d		1/1/2011	b
3	2/15/2011
4	10/1/2010	8/1/2010	9/1/2010				a	b

[Linlin]

you can add an array for locations to Art's code.

data want (drop=i);

  set have;

  array dates(*) T1_date--T5_date;

  array location(*) location1--location5;

  format first_Treatment_dt mmddyy10.;

  do i=1 to dim(dates);

    if dates(i) ge Hospital_dt then do;

      first_Treatment_dt=dates(i);

            first_location=location(i);

      leave;

    end;

  end;

run;

[art297]

Linlin, Agreed!  I was about to propose the exact same code, but didn't, as the results didn't come out as expected.  Then I realized that it wasn't working as expected because the second date, in the example data, was 3/2/3011.  Doubt if I'll be alive in 3011, but I couldn't figure out why 3/2 was being selected.

[art297]

Same suggestion as Tom's, but I like to leave loops once a condition is met.  Conversely, if your data isn't in date order, then I'd definitely go with Tom's suggested code:

data want (drop=i);

  set have;

  array dates(*) T1_date--T5_date;

  format first_Treatment_dt mmddyy10.;

  do i=1 to dim(dates);

    if dates(i) ge Hospital_dt then do;

      first_Treatment_dt=dates(i);

      leave;

    end;

  end;

run;