topic Re: How do I extract a range of ICD codes from several variables to create a 2-level indicator? in SAS Programming

How do I extract a range of ICD codes from several variables to create a 2-level indicator?

SAS93 — Tue, 22 Oct 2024 17:31:37 GMT

I'm working with ICD-10/ICD-10M codes, with about 22 different variables to pull the codes from.

I want to create 2-level indicator variable for each of the main ICD groups, so that if a record has any of the diagnosis codes, they'll fall into the respective bucket. For example:

'A00' - 'B99' = 'Intestinal Infectious Disease'

But what I'm doing is only returning DX_intest = 0.

ARRAY old [22] h_Diag1 - h_Diag18   h_Admit_Diag_Cd  h_Patient_Reason_Visit1 - h_Patient_Reason_Visit3 ;

Do i = 1 to 22;		
	If substr(OLD(i),1,3) in ("A:") OR  substr(OLD(i),1,3) in ("B:") then DX_intest=1;
	Else DX_all=0;
end;

Ultimately, I want to return "1" for DX_intest if any of the variables listed in the array contain ICD codes of A00 through B99; otherwise, return DX_intest = 0;

Re: How do I extract a range of ICD codes from several variables to create a 2-level indicator?

ballardw — Tue, 22 Oct 2024 17:51:45 GMT

If you are trying to find "begins with" don't use

in ("A:")

that tries to match 3 selected characters with exactly 2 and includes the : . So nothing matches.

Consider:

data example;
   input code $;
   if code in: ('A' 'B') then put 'Found ' code=;
datalines;
A00456
B12456
C00000
;

However you current program would return only the value of the LAST variable compared: h_Patient_Reason_Visit3

So you might want:

ARRAY old [22] h_Diag1 - h_Diag18   h_Admit_Diag_Cd  h_Patient_Reason_Visit1 - h_Patient_Reason_Visit3 ;

dx_all=0;
Do i = 1 to 22;		
   dx_intest= substr(OLD(i),1,3) in: ("A" "B");
   if DX_intest=1 then do;
      dx_all = .;
      leave;
   end;
	
end;

You don't say what dx_all is supposed to represent so I am guessing on that point that you don't want dx_all to be 0 when you find one of the codes.

The LEAVE instruction exits a do loop (caution with nested loops as to which it leaves). So when you find the first A or B this sets the value of DX_intest to 1 then ends the loop and stops searching. A side effect of this code is that the variable i will have the index of the variable in the array where the value was found.

SAS will return 1 for True comparisons and 0 for false. So if none of the variables have the condition true the result at completion of the array tested will be 0.

Re: How do I extract a range of ICD codes from several variables to create a 2-level indicator?

Tom — Tue, 22 Oct 2024 17:55:23 GMT

To detect ANY set the result FALSE before the loop. The loop over the array storing the result of the test, stopping when it is TRUE.

dx_intest=0;
do i = 1 to dim(old) while (not dx_intest);
  dx_intest= ( old[i] in: ('A' 'B') );
end;

Or even easier just stop once it is TRUE.

do i = 1 to dim(old) until(dx_intest);
  dx_intest= ( old[i] in: ('A' 'B') );
end;

Re: How do I extract a range of ICD codes from several variables to create a 2-level indicator?

Kurt_Bremser — Tue, 22 Oct 2024 18:02:23 GMT

Move the zero assignment out of the loop:

DX_intest = 0;
do i = 1 to 22;		
  if substr(OLD{i},1,1) eq "A" or substr(OLD{i},1,1) eq "B" then DX_intest = 1;
end;

Note that you can not use wildcards to build a list for the IN operator. All values in the IN list must be explicitly coded. If necessary, you can store all wanted ICD codes in a dataset and use it to build code dynamically.