topic Join tables (one-to-many relationship) where many are dummy variables? in SAS Programming

Join tables (one-to-many relationship) where many are dummy variables?

PharmlyDoc — Fri, 10 Sep 2021 18:03:32 GMT

I have two tables, table 1 has one record per patient, table 2 has multiple records per patient with dummy variables. I want to join the tables so that the dummy variables are merged with table 1.

Table 1.

PAT_ID	AGE	SEX
93704	20	F
93705	26	F
93706	28	M
93707	27	F
93708	47	M
93709	38	F
93710	44	F
93711	38	M
93712	24	F
93713	40	F
93714	40	F
93715	25	M
93716	23	M
93717	48	F
93718	41	F

Table 2.

PAT_ID	DRUG	STATIN	BETA_BLOCKER	ACEI	DIURETIC	STEROID
93705	atorvastatin	1	0	0	0	0
93705	metoprolol	0	1	0	0	0
93705	lisinopril	0	0	1	0	0
93709	hydrochlorothiazide	0	0	0	1	0
93709	rosuvastatin	1	0	0	0	0
93709	prednisone	0	0	0	0	1
93713	pravastatin	1	0	0	0	0
93715	carvedilol	0	1	0	0	0
93715	prednisone	0	0	0	0	1
93715	atorvastatin	1	0	0	0	0
93716	spironolactone	0	0	0	1	0
93717	lisinopril	0	0	1	0	0
93717	propranolol	0	1	0	0	0
93717	hydrochlorothiazide	0	0	0	1	0
93718	lisinopril	0	0	1	0	0

Want:

PAT_ID	AGE	SEX	STATIN	BETA_BLOCKER	ACEI	DIURETIC	STEROID
93704	20	F	0	0	0	0	0
93705	26	F	1	1	1	0	0
93706	28	M	0	0	0	0	0
93707	27	F	0	0	0	0	0
93708	47	M	0	0	0	0	0
93709	38	F	1	0	0	1	1
93710	44	F	0	0	0	0	0
93711	38	M	0	0	0	0	0
93712	24	F	0	0	0	0	0
93713	40	F	1	0	0	0	0
93714	40	F	0	0	0	0	0
93715	25	M	1	1	0	0	1
93716	23	M	0	0	0	1	0
93717	48	F	0	1	1	1	0
93718	41	F	0	0	1	0	0

data table_1;
input PAT_ID AGE SEX $;
datalines;
93704 20 F
93705 26 F
93706 28 M
93707 27 F
93708 47 M
93709 38 F
93710 44 F
93711 38 M
93712 24 F
93713 40 F
93714 40 F
93715 25 M
93716 23 M
93717 48 F
93718 41 F
;

data table_2;
infile datalines dlm=',';
input PAT_ID DRUG : $19. STATIN BETA_BLOCKER ACEI DIURETIC STEROID;
datalines;
93705,atorvastatin,1,0,0,0,0
93705,metoprolol,0,1,0,0,0
93705,lisinopril,0,0,1,0,0
93709,hydrochlorothiazide,0,0,0,1,0
93709,rosuvastatin,1,0,0,0,0
93709,prednisone,0,0,0,0,1
93713,pravastatin,1,0,0,0,0
93715,carvedilol,0,1,0,0,0
93715,prednisone,0,0,0,0,1
93715,atorvastatin,1,0,0,0,0
93716,spironolactone,0,0,0,1,0
93717,lisinopril,0,0,1,0,0
93717,propranolol,0,1,0,0,0
93717,hydrochlorothiazide,0,0,0,1,0
93718,lisinopril,0,0,1,0,0
;

Re: Join tables (one-to-many relationship) where many are dummy variables?

Shmuel — Fri, 10 Sep 2021 18:36:31 GMT

Having one to many relationship between the two tables. you can use sas base, merge statement:

data want;
merge table1
      table2(drop=drug);
  by pat_id;
run;

Re: Join tables (one-to-many relationship) where many are dummy variables?

tarheel13 — Fri, 10 Sep 2021 18:43:03 GMT

That gets more than 1 row per patient though.

Re: Join tables (one-to-many relationship) where many are dummy variables?

Patrick — Sat, 11 Sep 2021 08:27:26 GMT

Using the SQL below should do the job.

data table_1;
input PAT_ID AGE SEX $;
datalines;
93704 20 F
93705 26 F
93706 28 M
93707 27 F
93708 47 M
93709 38 F
93710 44 F
93711 38 M
93712 24 F
93713 40 F
93714 40 F
93715 25 M
93716 23 M
93717 48 F
93718 41 F
;

data table_2;
infile datalines dlm=',';
input PAT_ID DRUG : $19. STATIN BETA_BLOCKER ACEI DIURETIC STEROID;
datalines;
93705,atorvastatin,1,0,0,0,0
93705,metoprolol,0,1,0,0,0
93705,lisinopril,0,0,1,0,0
93709,hydrochlorothiazide,0,0,0,1,0
93709,rosuvastatin,1,0,0,0,0
93709,prednisone,0,0,0,0,1
93713,pravastatin,1,0,0,0,0
93715,carvedilol,0,1,0,0,0
93715,prednisone,0,0,0,0,1
93715,atorvastatin,1,0,0,0,0
93716,spironolactone,0,0,0,1,0
93717,lisinopril,0,0,1,0,0
93717,propranolol,0,1,0,0,0
93717,hydrochlorothiazide,0,0,0,1,0
93718,lisinopril,0,0,1,0,0
;

proc sql;
  create table want as
  select 
    l.PAT_ID,
    l.AGE, 
    l.SEX,
    max(coalesce(r.STATIN,0)) as STATIN,
    max(coalesce(r.BETA_BLOCKER,0)) as BETA_BLOCKER,
    max(coalesce(r.ACEI,0)) as ACEI,
    max(coalesce(r.DIURETIC,0)) as DIURETIC,
    max(coalesce(r.STEROID,0)) as STEROID 
  from 
    table_1 l
    left join
    table_2 r
    on
      l.PAT_ID=r.PAT_ID
  group by l.PAT_ID, l.AGE, l.SEX
  ;
quit;

Re: Join tables (one-to-many relationship) where many are dummy variables?

PharmlyDoc — Sat, 11 Sep 2021 15:59:59 GMT

@tarheel13 Thanks for answering, I noticed that your method does not provide the correct output if a patient is taking more than one medication in the same drug class, for example 2 diuretics, and then distinct values for DIURETIC would be .,0,1,2 instead of just .,0,1.

@Patrick

Your method works. Though lets say table_1 has 30 columns instead of 3, in which case I modified your code grouping only by PAT_ID and then used proc sort with the nodupkey option.

proc sql;
  create table want as
  select 
    l.*,
    max(coalesce(r.STATIN,0)) as STATIN,
    max(coalesce(r.BETA_BLOCKER,0)) as BETA_BLOCKER,
    max(coalesce(r.ACEI,0)) as ACEI,
    max(coalesce(r.DIURETIC,0)) as DIURETIC,
    max(coalesce(r.STEROID,0)) as STEROID 
  from 
    table_1 l
    left join
    table_2 r
    on
      l.PAT_ID=r.PAT_ID
  group by l.PAT_ID
  ;
quit;

PROC SORT DATA=want OUT=want_NODUP NODUPKEY;
BY PAT_ID;
RUN;