Liyan,

The first SQL solution I offered a couple days ago may delete observations crossover ID groups.

The one below has three WHERE conditions appiled to the Cartesian product of a given dataset. conditions 1) and 2) subset the big Cartesian dataset into the original one. And the third one delete the transfers within a ID group. If you have other concerns, you may modify or add additional conditions into the where clause.

data t1; input @1 id @3 admis ddmmyy8. @12 disch ddmmyy8.; format admis ddmmyy8. disch ddmmyy8.; cards; 1 23/01/97 25/01/97 1 26/01/97 30/01/97 2 13/03/97 20/03/97 2 25/03/97 27/04/97 3 08/01/97 09/01/97 3 14/09/97 17/09/97 3 18/09/97 20/09/97 ; proc print; run; proc sql; select a.id, a.admis, a.disch from t1 a, t1 b where a.id=b.id and a.admis = b.admis and (b.disch+1) not in (select a.admis from t1 as a where a.id=b.id) order by a.id, a.admis ; quit; run;

Liu, Liyan HSURC wrote:

> Dear List-members, > > I am working with hospital discharge data. I'd like to create a subset > excluding transfers(admitted to another hospital within one day of > discharging from the former hospital, in the data below, ID 1 and 3 were > transferred cases). > > My question is: how to delete the 2 records with * attached. Retain or lag > command? > > id date of admission date of discharge > 1 23/01/97 25/01/97 * > 1 26/01/97 30/01/97 > 2 13/03/97 20/03/97 > 2 25/03/97 27/04/97 > 3 08/01/97 09/01/97 > 3 14/09/97 17/09/97 * > 3 18/09/97 20/09/97 > > Thanks. > > Liyan > > Liyan Liu > Research Officer > Health Services Utilization and Research Commission (HSURC) > Box 46, 103 Hospital Drive > Saskatoon, SK Canada S7N 0W8 > > tel: (306) 655-1523 > fax: (306) 655-1462 > e-mail: liul@sdh.sk.ca > website: http://www.sdh.sk.ca/hsurc/