Identifying first and second re-admissions in a health dataset

[David Metcalfe]

I am working with Stata/MP 15.1 and have appended two datasets.

Database 1 includes a list of patients that had a specific operation. Each patient is only represented once in this dataset.

Database 2 includes the full medical record of every patient in database 1. Each patient will likely therefore be represented many times.

I think I have been able to identify the date of the first readmission after discharge from hospital using:

bysort patient: egen firstreadmit = min(readmission)
format firstreadmit %td

However, what I would really like to do is:

1. Return the cause of the first readmission, i.e. the "readmissioncausecode" variable.
2. Identify the date and cause of the second readmission, which I can't easily do using min(readmission).
3. Only identify patients being readmitted for a specific reason, e.g. readmissioncausecode==400.

I realise that this is three questions in one post but it seemed easier to present them all together with the context of what I am trying to achieve. If anyone can provide any hints or point me in the direction of how I could achieve (1), (2), or (3), I would be very grateful.

I have included a dummy dataset below, which broadly reflects the structure of the databases I am using.

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte(database patient) float(opdate dischargedate readmission) int readmissioncausecode
1 1 19002 19007     .   .
1 2 19400 19408     .   .
1 3 19389 19422     .   .
2 1     .     . 20739 400
2 1     .     . 21138 444
2 1     .     . 21268 999
2 2     .     . 21165 400
2 2     .     . 19919 233
2 2     .     . 20630 889
2 3     .     . 21268 889
2 3     .     . 21165 889
2 3     .     . 19919 123
2 3     .     . 20630 223
end
format %d opdate
format %d dischargedate
format %d readmission

[Rich Goldstein]

I'm not sure I fully understand but I think the following will help:

bys patient (readmission): gen byte readmitnum=_n

this gives you a counter within patient such that 1 is the first readmission, 2 is the second, etc. - you can then pick off what you want for any particular readmission; the third issue would seem to be handled by using something like:

gen code400 = readmissioncausecode==400

which will give you a variable that is equal to 1 if the code is 400 and equal to 0 otherwise; you don't say what you plan to do with this so I just mention that it can be used in an "if" clause; see

help if

if you want to repeat some analysis for each of several codes that can be more easily handled via -foreach-; see

help foreach