Variable to identify individuals whose type of diabetes changes from one hospital visit to another

[Wylyann]

Hello, can anyone help me?
I have some medico-administrative data and I want to create a binomial variable that will code an individual as 1, if the type of diabetes diagnosed is constant from one hospital visit to another; and code 0, if the type of diabetes diagnosed varies according to the hospital visit.
This is what my medico-administrative data looks like:

Observation (numeric)	Id (numeric)	Hospital visit Number (alphanumeric)	Type of diabetes diagnosed (numeric)
1	3	04	2
2	3	05	2
3	3	07	2
4	15	01	1
5	15	02	2
6	18	03	1
7	22	06	2
8	30	01	2
9	30	03	1
10	30	04	1

[PeterClemmensen]

Across all distinct hospital visits?

[Wylyann]

No, for each individual.

[ballardw]

Can you show what you expect the result to look like?

Are the ONLY values for the diabetes variable 1 and 2?

[Wylyann]

Yes, the only values for the diabetes variable are 1 (type 1 diabetes mellitus) and 2 (type 2 diabetes mellitus).

This involves writing a command that automatically creates the variable "Type of diabetes evolution" (TDevolution) that assigns code 1 to individuals (Id) "3", "18" and "22"; then code 0 to Id "15" and "30".

[ballardw]

---

@Wylyann wrote:
Yes, the only values for the diabetes variable are 1 (type 1 diabetes mellitus) and 2 (type 2 diabetes mellitus).

This involves writing a command that automatically creates the variable "Type of diabetes evolution" (TDevolution) that assigns code 1 to individuals (Id) "3", "18" and "22"; then code 0 to Id "15" and "30".

---

Still have NOT shown what the resulting data set is supposed to look like.

[Wylyann]

Maybe I didn't understand you correctly. Can you reformulate your question, please?

[Wylyann]

My data contains more than 20,000 observations. What I present here is only a Sample.

[mkeintz]

Read each ID twice, the first time to set a flag if there is a change in diabetes type.  The second time to output data if the flag was set:

But the original code I sent included the NOTSORTED option in the by statement - which is not supported when SET has more than one input dataset.  So  "by id diabetes_type nosorted" needs to be reset to "by id", and the subsequent code needs to be changed, as here:

data want (drop=_:);
  set have (in=firstpass)  have (in=secondpass);
  by id;
  retain _keep_this_id 0;
  if first.id=1 then _keep_this_id=0;
  if first.id=0 and diabetes_type^=lag(diabetes_type) then _keep_this_id=1;
  if secondpass=1 and _keep_this_id=1;
run;

The original code (below) should not be used, and will be flagged by SAS:

data want (drop=_:);
  set have (in=firstpass)  have (in=secondpass);
  by id diabetes_type notsorted;
  retain _keep_this_id 0;
  if first.id=1 then _keep_this_id=0;
  if first.id=0 and first.diabetes_type=1 then _keep_this_id=1;
  if secondpass=1 and _keep_this_id=1;
run;

The first.id=0 and first.diabetes=1 condition tests for a change in diabetes others than at the beginning of the ID.   Because diabetes can change up, or change down., the BY statement has the notsorted parameter.