topic Re: Coding missing value indicator for longitudinal studies in SAS Programming

Coding missing value indicator for longitudinal studies

BochengJing — Mon, 19 Jul 2021 23:48:41 GMT

Hello SAS users, I am working on a longitudinal study data where I need to code missing value indicator for an outcome. So for each person ID, the outcome was measured several times throughout different time point. If the outcome is not missing, we say the outcome is observed (Indicator="O"), if some of the outcomes are missing but later the outcome appeared within the same person ID, then we say the outcome is intermittently missing (Indicator="I"). If the outcomes are missing at one point till the end of the person ID, we say this person is dropout (Indicator="D"). Here is how the data looks like:

ID	Time	Outcome	Indicator
1	3	2	O
1	6	.	I
1	9	3	O
1	12	1	O
1	15	.	I
1	18	1	O
2	3	2	O
2	6	3	O
2	9	.	I
2	12	.	I
2	15	3	O
2	18	2	O
3	3	1	O
3	6	1	O
3	9	.	D
3	12	.	D
3	15	.	D
3	18	.	D
4	3	3	O
4	6	.	I
4	9	2	O
4	12	.	D
4	15	.	D
4	18	.	D

Would you mind helping me figure a way to code the indicator variable that contains "observed", "Intermittently missing", "Dropout"? The original data only has ID, time, and outcome; I added the indicator variable for the demonstration.

Thank you.

Re: Coding missing value indicator for longitudinal studies

mkeintz — Tue, 20 Jul 2021 01:37:51 GMT

Read all records for each ID twice:

First pass determine which record number is the last non-missing record, call it last_observed.
Second pass
1. if the record-in-hand is after last_observed, then you have a 'D'
2. all other missing are 'I'
3. all non-missings are 'O'
4. then output the record.

This is untested, in the absence of a working DATA step with sample data.

data want (drop=i last_observed);
  do i=1 by 1 until (last.id);
    set have;
    by id;
    if outcome^=. then last_observed=i;
  end;

  do i=1 by 1 until (last.id);
    if      i>last_observed  then indicator='D'; 
    else if missing(outcome) then indicator='I';
    else                          indicator='O';
    output;
  end;
run;

Re: Coding missing value indicator for longitudinal studies

ChrisNZ — Tue, 20 Jul 2021 01:54:19 GMT

Like this?

data HAVE;
  input ID TIME OUTCOME ;
cards;
1 3 2 O
1 6 . I
1 9 3 O
1 12 1 O
1 15 . I
1 18 1 O
2 3 2 O
2 6 3 O
2 9 . I
2 12 . I
2 15 3 O
2 18 2 O
3 3 1 O
3 6 1 O
3 9 . D
3 12 . D
3 15 . D
3 18 . D
4 3 3 O
4 6 . I
4 9 2 O
4 12 . D
4 15 . D
4 18 . D
run;

proc sql nothreads;
  create table LASTOUTCOME as 
  select ID, max(N*(OUTCOME>0)) as LASTOUTCOME  
  from (select *, monotonic() as N from HAVE) 
  group by ID 
  order by ID ;
quit;

data WANT;
  merge HAVE LASTOUTCOME;
  by ID;
  length INDICATOR $1;
  INDICATOR=ifc(OUTCOME,'O', ifc(_N_<=LASTOUTCOME ,'I', 'D' ));
run;

ID	TIME	OUTCOME	LASTOUTCOME	INDICATOR
1	3	2	6	O
1	6	.	6	I
1	9	3	6	O
1	12	1	6	O
1	15	.	6	I
1	18	1	6	O
2	3	2	12	O
2	6	3	12	O
2	9	.	12	I
2	12	.	12	I
2	15	3	12	O
2	18	2	12	O
3	3	1	14	O
3	6	1	14	O
3	9	.	14	D
3	12	.	14	D
3	15	.	14	D
3	18	.	14	D
4	3	3	21	O
4	6	.	21	I
4	9	2	21	O
4	12	.	21	D
4	15	.	21	D
4	18	.	21	D

ID	TIME	OUTCOME	LASTOUTCOME	INDICATOR
1	3	2	6	O
1	6	.	6	I
1	9	3	6	O
1	12	1	6	O
1	15	.	6	I
1	18	1	6	O
2	3	2	12	O
2	6	3	12	O
2	9	.	12	I
2	12	.	12	I
2	15	3	12	O
2	18	2	12	O
3	3	1	14	O
3	6	1	14	O
3	9	.	14	D
3	12	.	14	D
3	15	.	14	D
3	18	.	14	D
4	3	3	21	O
4	6	.	21	I
4	9	2	21	O
4	12	.	21	D
4	15	.	21	D
4	18	.	21	D

ID	TIME	OUTCOME	LASTOUTCOME	INDICATOR
1	3	2	6	O
1	6	.	6	I
1	9	3	6	O
1	12	1	6	O
1	15	.	6	I
1	18	1	6	O
2	3	2	12	O
2	6	3	12	O
2	9	.	12	I
2	12	.	12	I
2	15	3	12	O
2	18	2	12	O
3	3	1	14	O
3	6	1	14	O
3	9	.	14	D
3	12	.	14	D
3	15	.	14	D
3	18	.	14	D
4	3	3	21	O
4	6	.	21	I
4	9	2	21	O
4	12	.	21	D
4	15	.	21	D
4	18	.	21	D

ID	TIME	OUTCOME	LASTOUTCOME	INDICATOR
1	3	2	6	O
1	6	.	6	I
1	9	3	6	O
1	12	1	6	O
1	15	.	6	I
1	18	1	6	O
2	3	2	12	O
2	6	3	12	O
2	9	.	12	I
2	12	.	12	I
2	15	3	12	O
2	18	2	12	O
3	3	1	14	O
3	6	1	14	O
3	9	.	14	D
3	12	.	14	D
3	15	.	14	D
3	18	.	14	D
4	3	3	21	O
4	6	.	21	I
4	9	2	21	O
4	12	.	21	D
4	15	.	21	D
4	18	.	21	D