topic Re: Randomly assigning an Index Date to a patient in SAS Programming

Randomly assigning an Index Date to a patient

lisahoward — Mon, 26 Jan 2015 11:53:24 GMT

Hello,

I wondered if anyone could me in coming up with some SAS code help to randomly assign an index date to a patients diagnosis, (lets say diabetes), within a 6 month calender block?

Normally the index date would be taken as first occurrence of diagnosis, in this instance though the index date is is assigned by randomly selecting one of the patients physician visits within the 6 month calender block that the first diagnosis occurred.

For example patient X is diagnosed 02Jan2000 with Diabetes. He has 100 Physician visits and dates in the healthcare database , 90 of these dates/visits occur in the same 6 month block of time 01jan2000 - 30june2000 as the first diagnosis of diabetes. 1 of these 90 visit dates has to be randomly chosen and assigned to the patient as the index date. The other 10 visit dates can't be used as they fall outside of the cohort accrual 6 month block (either before or after 01jan2000 - 30june2000).

Does anyone have some code that would help me find and assign index dates this way please?

The over all cohort of patients spans 2004-2013 so there would be 20 blocks of 6 months to work through. Index are only assigned/matched within each block so if a diagnosis occurred in block 20 (1JUL2013-31DEC2013) then only visits for that patient also falling within the same time block are used to randomly find and assign an index date for that patient.

So far I have created a dataset with first diagnosis and have created a variable Index_Range that I have populated as follows:

PatID Dx Date Index_Range

1 20Jan2005 1JAN2005 - 30JUN2005

2 18JUN2006 1JAN2006 - 30JUN2006

3 02SEP2012 1JUL2012 - 31DEC2012

I next search and pull out all physician records for patients in dataset created above and assign a similar Index_Range type variable using the physician visit date.

I presume I then merge all visits in Diagnosis dataset and Physician visit Dataset matching on PatiD and Index_Range variable.

What I don't know is how to code after, when i have merged my datasets and have all the visits matched via index range to the patients diagnosis what is the code I use to randomly select only one visit per patient and assign that visit as the Index date. Any thoughts on helping me write the code to 'randomly select' would be great.

Thank you in advance.

Re: Randomly assigning an Index Date to a patient

data_null__ — Mon, 26 Jan 2015 12:15:38 GMT

Selecting the random sample of the targeted range of visits is the easiest step in your process and can be done with PROC SURVEYSELECT;

data visit;
   input PatID  (DxDate inx1-inx2 visitDate) (:date9.) visitnum;
   format dxdate inx1-inx2 visitdate date9.; 
   cards; 
1         20Jan2005    1JAN2005  30JUN2005 05Jan2005 4 
1         20Jan2005    1JAN2005  30JUN2005 25Jan2005 5
1         20Jan2005    1JAN2005  30JUN2005 03APR2005 6
1         20Jan2005    1JAN2005  30JUN2005 28MAY2005 7
1         20Jan2005    1JAN2005  30JUN2005 01JUN2005 8
2         20Jan2005    1JAN2005  30JUN2005 05Jan2005 4 
2         20Jan2005    1JAN2005  30JUN2005 25Jan2005 5
2         20Jan2005    1JAN2005  30JUN2005 03APR2005 6
2         20Jan2005    1JAN2005  30JUN2005 28MAY2005 7
2         20Jan2005    1JAN2005  30JUN2005 01JUN2005 8
;;;;
   run; 

proc surveyselect SEED=557754207 N=1 out=dxvisit;
   strata patid;
   run; 
proc print; 
   run;

Re: Randomly assigning an Index Date to a patient

lisahoward — Mon, 26 Jan 2015 13:12:07 GMT

Thank you so much for your quick response.

Re: Randomly assigning an Index Date to a patient

Babloo — Tue, 27 Jan 2015 07:38:19 GMT

What does the seedvalue (SEED=557754207) indicates?

Re: Randomly assigning an Index Date to a patient

data_null__ — Tue, 27 Jan 2015 13:01:43 GMT

SAS/STAT(R) 12.1 User's Guide

Re: Randomly assigning an Index Date to a patient

Phaneendra — Tue, 27 Jan 2015 14:33:29 GMT

It gives specifications for random seed generation.

Used for replicating results, i.e if we run the program more than once with same seed number, output will be the same; But if we change the seed number, output will be different.