topic Re: Identifying onset in longitudinal data in SAS Programming

Identifying onset in longitudinal data

lalaktgrau — Mon, 15 Oct 2018 18:42:21 GMT

Hi all,

I have longitudinal data (one obs per person per minute) with the variables: ID, DateTime, and Disease. Disease is binary.

How can I identify the first occurrence of disease in this longitudinal data?

Thanks

Re: Identifying onset in longitudinal data

novinosrin — Mon, 15 Oct 2018 18:44:00 GMT

Plz post samples of your input and your required output?

Sure some genuises may be gauge from sentence based description but samples help test the solutions from responder's end

Re: Identifying onset in longitudinal data

lalaktgrau — Mon, 15 Oct 2018 18:46:23 GMT

data WORK.CLASS;
infile datalines truncover;
input Name:$8. disease datetime;
informat datetime datetime20.;
format datetime datetime20.;
datalines;
Alfred 0 04MAR14:23:55:00
Alfred 0 04MAR14:23:56:00
Alfred 1 04MAR14:23:57:00
Alfred 1 04MAR14:23:58:00
Alfred 1 04MAR14:23:59:00
Alfred 1 04MAR14:00:00:00
Alfred 1 04MAR14:01:01:00
Alfred 1 04MAR14:01:02:00
Alfred 1 04MAR14:01:03:00
Alfred 1 04MAR14:01:04:00

mary 0 04MAR14:23:55:00
mary 0 04MAR14:23:56:00
mary 0 04MAR14:23:57:00
mary 0 04MAR14:23:58:00
mary 0 04MAR14:23:59:00
mary 1 04MAR14:00:00:00
mary 1 04MAR14:01:01:00
mary 1 04MAR14:01:02:00
mary 1 04MAR14:01:03:00
mary 1 04MAR14:01:04:00
;

I'd like a variable that =1 just on the FIRST time disease=1. For example, the variable would =1 for Alfred at '04MAR14:23:57:00' and =1 for Mary at ' 04MAR14:00:00:00'

Re: Identifying onset in longitudinal data

novinosrin — Mon, 15 Oct 2018 18:52:56 GMT

 
data WORK.CLASS;
infile datalines truncover;
input Name:$8. disease datetime;
informat datetime datetime20.;
format datetime datetime20.;
datalines;
Alfred 0 04MAR14:23:55:00
Alfred 0 04MAR14:23:56:00
Alfred 1 04MAR14:23:57:00
Alfred 1 04MAR14:23:58:00
Alfred 1 04MAR14:23:59:00
Alfred 1 04MAR14:00:00:00
Alfred 1 04MAR14:01:01:00
Alfred 1 04MAR14:01:02:00
Alfred 1 04MAR14:01:03:00
Alfred 1 04MAR14:01:04:00
mary 0 04MAR14:23:55:00
mary 0 04MAR14:23:56:00
mary 0 04MAR14:23:57:00
mary 0 04MAR14:23:58:00
mary 0 04MAR14:23:59:00
mary 1 04MAR14:00:00:00
mary 1 04MAR14:01:01:00
mary 1 04MAR14:01:02:00
mary 1 04MAR14:01:03:00
mary 1 04MAR14:01:04:00
;

data want;
set class;
by name;
if first.name then temp=1;
temp+disease;
if temp=2;
drop temp;
run;

Re: Identifying onset in longitudinal data

lalaktgrau — Mon, 15 Oct 2018 19:33:15 GMT

This works... but what if a person can fall in and out of disease? How can I identify each individual onset?

Re: Identifying onset in longitudinal data

novinosrin — Mon, 15 Oct 2018 19:34:49 GMT

Can you clarify that point and post your required output plz

Re: Identifying onset in longitudinal data

lalaktgrau — Mon, 15 Oct 2018 19:36:28 GMT

data WORK.CLASS;
infile datalines truncover;
input Name:$8. disease datetime;
informat datetime datetime20.;
format datetime datetime20.;
datalines;
Alfred 0 04MAR14:23:55:00
Alfred 0 04MAR14:23:56:00
Alfred 1 04MAR14:23:57:00
Alfred 1 04MAR14:23:58:00
Alfred 0 04MAR14:23:59:00
Alfred 0 04MAR14:00:00:00
Alfred 1 04MAR14:01:01:00
Alfred 1 04MAR14:01:02:00
Alfred 1 04MAR14:01:03:00
Alfred 1 04MAR14:01:04:00

mary 0 04MAR14:23:55:00
mary 0 04MAR14:23:56:00
mary 0 04MAR14:23:57:00
mary 0 04MAR14:23:58:00
mary 0 04MAR14:23:59:00
mary 1 04MAR14:00:00:00
mary 0 04MAR14:01:01:00
mary 0 04MAR14:01:02:00
mary 1 04MAR14:01:03:00
mary 1 04MAR14:01:04:00
;

If we wanted the variable to =1 each time there is a new onset of disease? Changing from 0 to 1?

Re: Identifying onset in longitudinal data

novinosrin — Mon, 15 Oct 2018 19:39:15 GMT

Do you mean you want to pick all the 1's?

An output sample for the input would help

Re: Identifying onset in longitudinal data

lalaktgrau — Mon, 15 Oct 2018 19:39:39 GMT

I would want a variable to equal 1 whenever there is a change from 0 to 1.

Re: Identifying onset in longitudinal data

novinosrin — Mon, 15 Oct 2018 19:42:13 GMT

may be this?

data WORK.CLASS;
infile datalines truncover;
input Name:$8. disease datetime;
informat datetime datetime20.;
format datetime datetime20.;
datalines;
Alfred 0 04MAR14:23:55:00
Alfred 0 04MAR14:23:56:00
Alfred 1 04MAR14:23:57:00
Alfred 1 04MAR14:23:58:00
Alfred 0 04MAR14:23:59:00
Alfred 0 04MAR14:00:00:00
Alfred 1 04MAR14:01:01:00
Alfred 1 04MAR14:01:02:00
Alfred 1 04MAR14:01:03:00
Alfred 1 04MAR14:01:04:00
mary 0 04MAR14:23:55:00
mary 0 04MAR14:23:56:00
mary 0 04MAR14:23:57:00
mary 0 04MAR14:23:58:00
mary 0 04MAR14:23:59:00
mary 1 04MAR14:00:00:00
mary 0 04MAR14:01:01:00
mary 0 04MAR14:01:02:00
mary 1 04MAR14:01:03:00
mary 1 04MAR14:01:04:00
;

data want;
set class;
by name disease notsorted;
if first.disease and disease then flag=1;
run;

Re: Identifying onset in longitudinal data

mkeintz — Mon, 15 Oct 2018 21:03:41 GMT

data WORK.CLASS;
infile datalines truncover;
input Name:$8. disease datetime;
informat datetime datetime20.;
format datetime datetime20.;
datalines;
Alfred 0 04MAR14:23:55:00
Alfred 0 04MAR14:23:56:00
Alfred 1 04MAR14:23:57:00
Alfred 1 04MAR14:23:58:00
Alfred 0 04MAR14:23:59:00
Alfred 0 04MAR14:00:00:00
Alfred 1 04MAR14:01:01:00
Alfred 1 04MAR14:01:02:00
Alfred 1 04MAR14:01:03:00
Alfred 1 04MAR14:01:04:00
mary 0 04MAR14:23:55:00
mary 0 04MAR14:23:56:00
mary 0 04MAR14:23:57:00
mary 0 04MAR14:23:58:00
mary 0 04MAR14:23:59:00
mary 1 04MAR14:00:00:00
mary 0 04MAR14:01:01:00
mary 0 04MAR14:01:02:00
mary 1 04MAR14:01:03:00
mary 1 04MAR14:01:04:00
;

data want;
  set class;
  by name;
  flag= (first.name=1 and disease=1) or (dif(disease)=1);
run;

The DIF function is defined as dif(x)=x-lag(x). So the condition "dif(disease)=1" picks up all instances of disease=1 that immediately follow a disease=0.