topic Re: Determining the number of subjects who chose from select all that apply question in SAS Programming

Determining the number of subjects who chose from select all that apply question

GregBond — Thu, 17 Apr 2014 19:00:04 GMT

Hopefully this wont confuse those who attempt to help...

Here is the question:

variable names in "( )"

X% of persons with lupus receiving corticosteroids (q131) reported having at least 1 co-morbidity (q141r1 q141r2 q141r3......)other than lupus; of which X% had at least one outpatient visit with a healthcare provider in the past three months (q111), X% were admitted at least once to the hospital in the past 12 months (q115), and X% have been to the emergency room at least once in the past 12 months (q118).

I suppose I have to make q141r1 q141r2... into one variable then run a crosstab, but I'm not entirely sure.

Any suggestions? Thanks!

Re: Determining the number of subjects who chose from select all that apply question

data_null__ — Thu, 17 Apr 2014 19:12:23 GMT

It could be as simple as

new=max(of q141r1 q141r2 q141r3);

assuming numeric 0/1 variables. Example and details of your data will make it easier to determine exactly.

Re: Determining the number of subjects who chose from select all that apply question

GregBond — Fri, 18 Apr 2014 13:03:12 GMT

Thank for your response!

You are correct in that the data are binary 0/1. Essentially, i need to find out if any patients reported not having a comorbidic condition which are listed (check all that apply) as q141r1 q141r2 q141r3.... and are 0=no 1=yes. Hope this helps and thanks again!

Regards,

Greg

Re: Determining the number of subjects who chose from select all that apply question

GregBond — Fri, 18 Apr 2014 13:05:44 GMT

In addition to having XXXX; have you ever been told by a healthcare provider that you have:

Eye problems more than needing glasses q141r1

Brain or spinal cord problems q141r2

Kidney problems q141r3

Lung problems

Heart problems

Blood vessels problems

Stomach, intestine or liver problems

Muscle, bone, and joint problems

Skin problems

Diabetes

Cancer

Mental health problems

I don't have any health problems other than lupus

TOTAL RESPONSES

I would need to know how to tell if an individual patient did not chose any of these items.

Re: Determining the number of subjects who chose from select all that apply question

Haikuo — Fri, 18 Apr 2014 13:35:49 GMT

The answer from should work for your problem if they are binary:

new=max(of q141r1 q141r2 q141r3);

or for more information to use:

new=sum(q141r1, q141r2, q141r3);

So when new=0 meaning the patient has not chosen any of these options, and the value of 'new' indicates how many options have been selected.

Haikuo

Re: Determining the number of subjects who chose from select all that apply question

GregBond — Fri, 18 Apr 2014 14:54:43 GMT

Thanks for you help Haikuo!

There are 827 people in the dataset

I ran the following code and had a question as to how I can tell the percentage of people who did no chose one answer. Notice in the output there are 11 variables when the code there are 13 because two options must be '0'.

data work.pwl_spss;

set work.pwl_spss;

new2=sum(q42r1 , q42r2 , q42r3 , q42r4 , q42r5 , q42r6 , q42r7 , q42r8 , q42r9 , q42r10 , q42r11, q42r12 , q42r13);

run;

proc freq data=work.pwl_spss;

tables new2;

run;

and got the following output...

The FREQ Procedure

new2 Frequency Percent Cumulative Frequency Cumulative Percent

1 110 13.30 110 13.30

2 100 12.09 210 2 5.39

3 134 16.20 344 41.60

4 118 14.27 462 55.86

5 129 15.60 591 71.46

6 92 11.12 683 82.59

7 69 8.34 752 90.93

8 37 4.47 789 95.41

9 24 2.90 813 98.31

10 10 1.21 823 99.52

11 4 0.48 827 100.00

Re: Determining the number of subjects who chose from select all that apply question

GregBond — Fri, 18 Apr 2014 14:58:21 GMT

NVM Haikuo, I seemed to have figured out the issue, thank you for your assistance!

Re: Determining the number of subjects who chose from select all that apply question

data_null__ — Fri, 18 Apr 2014 15:20:00 GMT

I usually use PROC SUMMARY when I have binary data: n=Denom sum=Count mean=percent. You can also use PROC TABULATE but the output from tabulated is not suited to the arcane requirements of my reporting environment, so I use PROC REPORT.

The variable labels are the row labels with the stats displayed in columns defining groups if necessary.

I would be easier to give you ideas about how you might do it if you supplied some sample data that more or less realistically models your actual data.