Matching cases to controls using risk set matching - Ranking cases by index time

[Schtroumpfette]

Hello,

I am new to complex SAS coding and I would appreciate your help to have a code to match cases to controls, while using risk set sampling, and not Propensity score matching. 

I have two datasets, cases and controls-please see below some examples.  Each dataset has the min(discharge_time) as index_date and I already formatted the date.  I also grouped by Patient_id.  Each patient_id is a unique patient identifier and the patient can have multiple encounters. 

I created age_low and age_high variables, as well as index_date_low and index_date_high in the Controls datasest. 

• A control may become a case, as long as it is prior to the index date of the would be case. 
• I can select the controls within a range of the index date (example between index_date_low and index_date of the case) to provide a suitable range for index_date. 
• I also added age_low and age_high to have a range for age
• I need to match exactly on sex and hospital_id.
• I would keep all cases even if the case has only 1 control. 

I need to do the following based on the required procedure:

1. Rank the cases by index_date.

2. temp: For each case, pull all encounters within an interval around index date, except for the case (remove the patient_id for the case)

3. From temp, draw encounters that match to sex, hospitalID and age, while taking the closest encounter_id to the index. 

4. Randomly sample matching encounters.

Unfortunately, I don't know how to do that.  Would you be able to provide a code for this? 

Here is an example for each dataset. 

Cases dataset is as follows:

EncounterID	Patient_id	age	sex	hospitalID	index_date	Diabetes_Event
123	1		M	44	20053	1
124	1		M	44	20053	1
125	1		M	45	20053	1
12345	2		F	67	19875	1
12367	2		F	67	19875	1

Controls Dataset is:

EncounterID	Patient_id	age	sex	hospitalID	index_date	Diabetes_Event	Age_low	Age_high	Index_date low
333	4		M	40	20049	0	38	42	20019
33667	5		F	66	19849	0	64	68	19819
33885	6		F	66	19849	0	64	68	19819
33889	6		F	66	19849	0	64	68	19819
33999	6		F	66	19849	0	64	68	19819
1238	7		M	46	18920	0	44	48	18890
1239	7		M	46	18920	0	44	48	18890
1300	7		M	46	18920	0	44	48	18890
145	8		F	30	20207	0	28	32	20177
146	8		F	30	20207	0	28	32	20177

Thanks, 

Schtroumpfette

[japelin]

If possible, after filling in the AGE, can you clarify the results of each of the processes 1-3 as data, based on the Control data set you presented?

I am not fully aware of what kind of processing you are assuming.

For example, what kind of process is "Rank the cases by index_date"?
Is it enough to rank them in order from 1?
How are the results of this process stored?

[Schtroumpfette]

Thanks Kawakami,

I will get back to you with more information, hopefully shortly. There was a change of plan to this.

Thanks,