Date: Thu, 3 Feb 2011 03:48:37 -0500
Reply-To: "Kirby, Ted" <ted.kirby@LEWIN.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: "Kirby, Ted" <ted.kirby@LEWIN.COM>
Subject: Re: Calculating a 30-day window from changing dates
Content-Type: text/plain; charset="iso-8859-1"
Max,
If Tom was in hospital A for one year, then there can be no other hospital stays within that year. A person cannot be in two hospitals at the same time.
________________________________
From: SAS(r) Discussion on behalf of bbser 2009
Sent: Wed 2/2/2011 10:40 PM
To: SAS-L@LISTSERV.UGA.EDU
Subject: Re: Calculating a 30-day window from changing dates
OK. It's clear on that point now. But here is another issue. Think about an
ultra-most case where Tom was with hospital A for like one year or even 10
years. Then any other treatments for Tom from other hospitals which felt in
the admission and discharge period with hospital A would not be counted as
index admissions. This sounds not reasonable. Using admit dates as "anchors"
makes much sense to me.
Anyway, if you insist on discharge dates, I guess it is not hard to modify
my original code for this purpose.
Max
From: Emily Sim [mailto:emilyssim@gmail.com]
Sent: February-02-11 9:42 PM
To: bbser 2009
Cc: SAS-L@listserv.uga.edu
Subject: Re: Calculating a 30-day window from changing dates
Sorry, that surely is confusing. Each observation represents a hospital
stay, so each observation has an admit date and a discharge date. An
observation is considered an "index admit" if the admit date for that
observation is greater than 30 days from the discharge date of the previous
index admit. Does that clear it up?
Thanks,
Emily
On Wed, Feb 2, 2011 at 7:27 PM, bbser 2009 <bbser2009@gmail.com> wrote:
Emily
In your original post, you did mention using admit date instead of discharge
date to be candidates of index dates.
I thought it made sense that way.
Because of your last email, I was confused. Did you mean the first admit
date after 30 days of "index discharge date" would be another index admit
date?
Thanks.
Max
From: Emily Sim [mailto:emilyssim@gmail.com]
Sent: February-02-11 8:54 PM
To: bbser 2009
Cc: SAS-L@listserv.uga.edu
Subject: Re: [SAS-L] Calculating a 30-day window from changing dates
Thanks Max.
One note of clarification, the 30 day window is from the discharge date, not
the admit date. So I think I would need to amend your code to be
index_admit = disch_date after the first do statement. But is seems to make
sense after that.
Does your code work for index admits after the second one? Wouldn't this
just give me the first index admit and the second one, but not any after
that?
Thanks so much,
Emily
On Wed, Feb 2, 2011 at 6:13 PM, bbser 2009 <bbser2009@gmail.com> wrote:
Given an index admission, the first admission coming right after 30 days of
the index admission would be another index admission.
If this understanding is right, please try the code below.
Max
==== code after your first data step ====
proc sort data=hypo;
by ID ADMIT_DATE;
run;
data index (drop=index_admit);
set hypo;
by id admit_date;
retain index_admit;
if first.id <http://first.id/> then
do;
index_admit=admit_date;
output;
end;
else if admit_date-index_admit >=30 then
do;
index_admit=admit_date;
output;
end;
run;
proc print;
run;
-----Original Message-----
From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of Emily
Sim
Sent: February-02-11 6:58 PM
To: SAS-L@LISTSERV.UGA.EDU
Subject: [SAS-L] Calculating a 30-day window from changing dates
My data consists of hospital admissions. Each observation is one hospital
admission for a patient (includes patient ID, Hospital Name, Admit Date,
Discharge Date, and Discharge Status)
I'm trying to determine which hospital admissions for each patient
are "index admissions." There have to be at least 30 days between index
admissions. The first admission for any patient is an index admission and
then you count 30 days before another admission can be coded as an index
admission. Any other admissions for a patient are either readmissions (if
it is the first admission occurring within 30 days of an index admission)
or "not a readmission" (if it is another admit after a readmission but
before another index admission).
PROBLEM: How can I determine which hospital admissions are "index
admissions" in a streamlined way? I have written code that will do it if
the patient has up to 4 hospital admissions, but it is cumbersome and will
become too unwieldy for higher numbers of admissions (the highest number
of admissions for one patient in my real dataset is 46).
Thanks for any help (my data and code follow below),
-Emily
My real dataset has over 350,000 obs. I've created a miniature version to
learn on.
Each obs is a hospital admission. Variables are ID (patient ID),
HOSP_NAME (hospital name), ADMIT_DATE (date of admission), DISCH_DATE
(date of discharge), DISCH_STATUS (discharge status of patient).
Here is the code I have written so far. First I read in the dataset, then
I get rid of hospital transfers (DISCH_STATUS = 02), then I start
identifying index admissions and readmissions. I use IF THEN statements
to identify index admissions and I'll be the first to admit it isn't
elegant!
data hypo;
input ID $ 1
HOSP_NAME $ 2-11
@12 ADMIT_DATE MMDDYY10.
@22 DISCH_DATE MMDDYY10.
DISCH_STATUS $ 32-33
;
format ADMIT_DATE DISCH_DATE mmddyy10.;
cards;
1StRoseSan 05/23/200906/24/200909
1StRoseSie 05/10/200905/23/200902
1NorthVista05/10/200905/10/200902
1NorthVista06/28/200907/01/200909
2Sunrise 01/05/200901/07/200902
2NorthVista01/07/200901/09/200909
2Sunrise 01/15/200901/17/200909
2Sunrise 01/22/200901/24/200909
2Sunrise 02/05/200902/09/200909
2Sunrise 06/01/200906/10/200909
3NorthVista06/11/200906/11/200902
3NorthVista06/11/200906/18/200909
4Sunrise 09/25/200909/28/200909
4Sunrise 09/30/200909/30/200920
;
run;
proc sort data=hypo;
by ID ADMIT_DATE DISCH_DATE;
run;
/*This next section recodes hospital transfers as only one visit - so if a
patient is transfered twice, it still appears as one hospital admit
instead of 3*/
data hypo2;
set hypo;
PREV_DISCH_STATUS = lag1(DISCH_STATUS);
PREV_ADMIT_DATE = lag1(ADMIT_DATE);
format PREV_ADMIT_DATE mmddyy10.;
run;
/*deleting middle transfers - when a patient was transfered more than
once*/
data hypo3;
set hypo2;
if DISCH_STATUS = '02' and PREV_DISCH_STATUS = '02' then delete;
run;
data hypo4;
set hypo3;
if PREV_DISCH_STATUS = '02' then ADMIT_DATE = PREV_ADMIT_DATE;
run;
/*deleting the initial admit before the transfer - the transferred admit
now has the
admit date from the initial admit*/
data hypo5;
set hypo4;
if DISCH_STATUS = '02' then delete;
if PREV_DISCH_STATUS = '02' then COLLAPSED = 'Y';
else COLLAPSED = 'N';
/*HERE IS WHERE I START TO COUNT INDEX ADMITS (AND READMISSIONS)*/
/*a patient can only have one index admit within each 30-day period. The
30-day count starts from each index admit*/
/*first counting the number of admits for each patient - COUNT = 1 is the
first admit, COUNT = 2 is the second, etc.*/
data hypo6;
set hypo5;
COUNT + 1;
by ID;
if first.ID then COUNT=1;
run;
data hypo9;
set hypo6;
DAYS_LAG1 = ADMIT_DATE - lag1(DISCH_DATE);
DAYS_LAG2 = ADMIT_DATE - lag2(DISCH_DATE);
DAYS_LAG3 = ADMIT_DATE - lag3(DISCH_DATE);
DAYS_LAG4 = ADMIT_DATE - lag4(DISCH_DATE);
run;
data hypo10;
set hypo9;
if COUNT = 1 then DAYS_FROM_INDEX = .;
if COUNT = 2 then DAYS_FROM_INDEX = DAYS_LAG1;
if COUNT = 3 then DAYS_FROM_INDEX = DAYS_LAG2;
if COUNT = 4 then DAYS_FROM_INDEX = DAYS_LAG3;
if COUNT = 5 then DAYS_FROM_INDEX = DAYS_LAG4;
run;
/*CODING INDEX ADMITS FOR COUNT = 2 (PATIENTS WITH 2 ADMISSIONS)*/
data hypo11;
set hypo10;
if COUNT = 1 then INDEX_ADMIT = 'Y';
if COUNT = 2 and DAYS_FROM_INDEX >30 then INDEX_ADMIT = 'Y';
if COUNT = 2 and DAYS_FROM_INDEX <= 30 then READMISSION = 'Y';
run;
/*Now dealing with COUNT = 3 (PATIENTS WITH 3 ADMISSIONS)*/
data hypo12;
set hypo11;
DAYS_FROM_SECOND = ADMIT_DATE - lag1(DISCH_DATE);
THIRD_PREV_INDEX = lag1(INDEX_ADMIT);
THIRD_PREV_READMISSION = lag1(READMISSION);
if COUNT = 3 and THIRD_PREV_INDEX = 'Y' and DAYS_FROM_SECOND >30
then INDEX_ADMIT = 'Y';
if COUNT = 3 and THIRD_PREV_INDEX = 'Y' and DAYS_FROM_SECOND <=30
then READMISSION = 'Y';
if COUNT = 3 and THIRD_PREV_READMISSION = 'Y' and DAYS_FROM_INDEX
>30 then INDEX_ADMIT = 'Y';
if COUNT = 3 and THIRD_PREV_READMISSION = 'Y' and DAYS_FROM_INDEX
<=30 then READMISSION = 'N';
run;
/*Now dealing with COUNT = 4 (PATIENTS WITH 4 ADMISSIONS)*/
data hypo13;
set hypo12;
DAYS_FROM_THIRD = ADMIT_DATE - lag1(DISCH_DATE);
DAYS_FROM_SECOND = ADMIT_DATE - lag2(DISCH_DATE);
FOURTH_PREV_INDEX = lag1(INDEX_ADMIT);
FOURTH_PREV_READMISSION = lag1(READMISSION);
FOURTH_TWO_PREV_INDEX = lag2(INDEX_ADMIT);
FOURTH_TWO_PREV_READMISSION = lag2(READMISSION);
if COUNT = 4 and FOURTH_TWO_PREV_INDEX = 'Y' and FOURTH_PREV_INDEX
= 'Y' and DAYS_FROM_THIRD >30 then INDEX_ADMIT = 'Y';
if COUNT = 4 and FOURTH_TWO_PREV_INDEX = 'Y' and FOURTH_PREV_INDEX
= 'Y' and DAYS_FROM_THIRD <=30 then READMISSION = 'Y';
if COUNT = 4 and FOURTH_TWO_PREV_INDEX = 'Y' and
FOURTH_PREV_READMISSION = 'Y' and DAYS_FROM_SECOND >30 then INDEX_ADMIT
= 'Y';
if COUNT = 4 and FOURTH_TWO_PREV_INDEX = 'Y' and
FOURTH_PREV_READMISSION = 'Y' and DAYS_FROM_SECOND <=30 then READMISSION
= 'N';
if COUNT = 4 and FOURTH_TWO_PREV_READMISSION = 'Y' and
FOURTH_PREV_READMISSION = 'N' and DAYS_FROM_INDEX <30 then READMISSION
= 'N';
if COUNT = 4 and FOURTH_TWO_PREV_READMISSION = 'Y' and
FOURTH_PREV_READMISSION = 'N' and DAYS_FROM_INDEX >=30 then INDEX_ADMIT
= 'Y';
if COUNT = 4 and FOURTH_TWO_PREV_READMISSION = 'Y' and
FOURTH_PREV_INDEX = 'Y' and DAYS_FROM_THIRD >30 then INDEX_ADMIT = 'Y';
if COUNT = 4 and FOURTH_TWO_PREV_READMISSION = 'Y' and
FOURTH_PREV_INDEX = 'Y' and DAYS_FROM_THIRD <=30 then READMISSION = 'Y';
run;
************* IMPORTANT - PLEASE READ ********************
This e-mail, including attachments, may include confidential and/or proprietary information,
and may be used only by the person or entity to which it is addressed. If the reader of this
e-mail is not the intended recipient or his or her authorized agent, the reader is hereby
notified that any dissemination, distribution or copying of this e-mail is prohibited. If you
have received this e-mail in error, please notify the sender by replying to this message
and delete this e-mail immediately.
