topic Determine how linkage flag variables are overlapped in SAS Programming

Determine how linkage flag variables are overlapped

Cruise — Tue, 19 Jun 2018 04:33:05 GMT

How to create 10 by 10 matrix to map out the overlapped flag variables?

I'd like to assess the overlapping pattern 10 different flag variables. Here is some background why I'm here. I have 17,811 rows of unique 'med_codes' column (see excel data attached if you have time) which is a mixture of literally all categories of medical procedure, diagnosis, billing and administrative codes. My goal is to define each code row. I'm kinda getting there since I have only 466 codes undefined. How did I do that? I have flagged if complete and source list of i.e., ICD10, ICD9, CPT and HCPCS et.c linked to my 'med_codes' column. If unknown code is linked to ICD-9cm then flag_icd9cm=1; it takes 0. As a result, I have 10 different flags below for each code categories. Unfortunately, they overlapped. Data=check has 466 rows even though sum of all linked variables reached 27449 exceeding total row number 17,811.

So now, my question is: How to map out the pattern how flag variables are crossed altogether? Had I known how many different flag variables are overlapped to what extent then I would look at them and weed them out using certain rules that makes sense to the nature of my project.

Thanks zillionz in advance. Any brainstorming or direct coding hints are appreciated.

flag_icd10pc

flag_icd10cm

flag_hcpc

flag_icd9_pc

flag_cpt

flag_icd9cm

flag_cpt_del

flag_hipps

flag_abc

flag_bill

flag_dent

flag_D

data check; set cruise;
linked=0;
if flag_cpt=1 then linked=1; 
if flag_hcpc=1 then linked=1; 
if flag_icd9_pc=1 then linked=1; 
if flag_icd10pc=1 then linked=1; 
if flag_icd9cm=1 then linked=1; 
if flag_icd10cm=1 then linked=1; 
if flag_cpt_del=1 then linked=1; 
if flag_hipps=1 then linked=1; 
if flag_other=1 then linked=1; 
if flag_abc=1 then linked=1; 
if flag_bill=1 then linked=1;
if flag_dent=1 then linked=1;
if flag_D=1 then linked=1;
if linked=0; 
run; 

/*have time to have a close
look at my data?, plz use .xlsx data 
attached*/

proc import 
datafile="cruise.xlsx"
out=cruise
dbms=xlsx 
replace;
getnames=yes;
run;

Re: Determine how linkage flag variables are overlapped

ChrisNZ — Tue, 19 Jun 2018 05:08:49 GMT

Not too sure I understand the output you are after (where is your example?) but his may help.

data HAVE; 
  do I= 1 to 100;
    FLAG_ICD10PC =round(ranuni(0)*.56);
    FLAG_ICD10CM =round(ranuni(0)*.56);	
    FLAG_HCPC    =round(ranuni(0)*.56);	
    FLAG_ICD9_PC =round(ranuni(0)*.56);	
    FLAG_CPT     =round(ranuni(0)*.56);	
    FLAG_ICD9CM  =round(ranuni(0)*.56);	
    FLAG_CPT_DEL =round(ranuni(0)*.56);	
    FLAG_HIPPS   =round(ranuni(0)*.56);
    FLAG_ABC     =round(ranuni(0)*.56);
    FLAG_BILL	   =round(ranuni(0)*.56);
    output; 
  end;
run;

data REPORT; 
  set HAVE end=LASTOBS;
  array F [10] FLAG:;
  array X [10,10] 8;
  do I= 1 to 10;
    do J=2 to 10;
      if I = J then continue;
      if F[I] & F[J] then do;
        X[I,J]+1;
        X[J,I]+1;
      end;
    end;
  end;
  if LASTOBS then do;
    do I= 1 to 10;
      do J=1 to 10;
        if I = J then continue;
        VAL=coalesce(X[I,J],0);
        output;
      end;
    end;
  end;
  keep I J VAL; 
run;

proc format ;
  value varnames 1= 'FLAG_ICD10PC'  
                 2= 'FLAG_ICD10CM' 
                 3= 'FLAG_HCPC   ' 
                 4= 'FLAG_ICD9_PC' 
                 5= 'FLAG_CPT    ' 
                 6= 'FLAG_ICD9CM ' 
                 7= 'FLAG_CPT_DEL' 
                 8= 'FLAG_HIPPS  ' 
                 9= 'FLAG_ABC    ' 
                10= 'FLAG_BILL	 ';  
run;

proc tabulate data=REPORT;
  class I J;
  var VAL;
  tables I=' ',J=' '*VAL=' ' *sum=' '*f=5.0 ;
  format I J varnames.;
 run;

	FLAG_ICD10PC	FLAG_ICD10CM	FLAG_HCPC	FLAG_ICD9_PC	FLAG_CPT	FLAG_ICD9CM	FLAG_CPT_DEL	FLAG_HIPPS	FLAG_ABC	FLAG_BILL
FLAG_ICD10PC	.	1	1	1	2	3	2	1	2	1
FLAG_ICD10CM	1	.	6	2	2	0	2	8	8	4
FLAG_HCPC	1	6	.	0	4	0	2	4	6	6
FLAG_ICD9_PC	1	2	0	.	2	4	0	0	0	0
FLAG_CPT	2	2	4	2	.	4	2	4	6	6
FLAG_ICD9CM	3	0	0	4	4	.	2	2	0	2
FLAG_CPT_DEL	2	2	2	0	2	2	.	2	8	0
FLAG_HIPPS	1	8	4	0	4	2	2	.	6	2
FLAG_ABC	2	8	6	0	6	0	8	6	.	8
FLAG_BILL	1	4	6	0	6	2	0	2	8	.

Re: Determine how linkage flag variables are overlapped

Cruise — Tue, 19 Jun 2018 11:41:20 GMT

Exactamento!!!

Re: Determine how linkage flag variables are overlapped

Cruise — Tue, 19 Jun 2018 20:44:45 GMT

Hi Chris,

Great approach. I applied your code to my actual data. resulted numbers in the matrix are doubled the amount that would be expected expected simple proc freq 2 by 2 and misplaced across the cells. Any idea how it happened?

Thanks zillions!!!


data cruise1; 
  set cruise end=LASTOBS;
  array F [12] FLAG:;
  array X [12,12] 8;
  do I= 1 to 12;
    do J=2 to 12;
      if I = J then continue;
      if F[I] & F[J] then do;
        X[I,J]+1;
        X[J,I]+1;
      end;
    end;
  end;
  if LASTOBS then do;
    do I= 1 to 12;
      do J=1 to 12;
        if I = J then continue;
        VAL=coalesce(X[I,J],0);
        output;
      end;
    end;
  end;
  keep I J VAL; 
run;

proc format ;
  value varnames 1= 'FLAG_ICD10PC'  
                 2= 'FLAG_ICD10CM' 
                 3= 'FLAG_HCPC' 
                 4= 'FLAG_ICD9_PC' 
                 5= 'FLAG_CPT' 
                 6= 'FLAG_ICD9CM' 
                 7= 'FLAG_CPT_DEL' 
                 8= 'FLAG_HIPPS' 
                 9= 'FLAG_ABC' 
                10= 'FLAG_BILL' 
		11= 'FLAG_D' 
		12= 'FLAG_dent' ;
run;

proc tabulate data=cruise1;
  class I J;
  var VAL;
  tables I=' ',J=' '*VAL=' ' *sum=' '*f=5.0 ;
  format I J varnames.;
 run;

proc freq data=cruise;
tables flag_hcpc*flag_cpt/norow nocol nopercent nocum; 
run; /*6433*/