Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Re: brace counting for person-time

From   Nabarun Dasgupta <>
To   <>
Subject   st: Re: brace counting for person-time
Date   Sun, 03 Mar 2013 22:58:50 -0500

Of course the answer comes to you after you reach out.

The solution that is working for me:
1. Convert ender to 1/-1 instead of 0/1 and use sum() to create a running
2. Figure out the number of NON-EXPOSED days (including ends of the month)
using the running total
3. Subtract from the known number of days in the month

On 3/2/13 5:26 PM, Nabarun Dasgupta typed:

>I would greatly appreciate your help and haven't seen exactly what I need
>in the archive. I am trying to count the person-days exposed for a single
>patient in a given month over multiple overlapping Rx records. The correct
>number of days exposed is 26 in the example data below. I tried a "brace
>counting" approach, so the variable _ender_ is zero at the start of the Rx
>period, and one for the last day of the period. Hence there are eight
>records for four prescriptions. One solution would be to calculate the
>days in between each _bracedate_ and only add up the unique days between
>braces, but I can't figure out how to create a counter that identifies
>only the exposed periods. Note that rxno 5061 was only for one day, making
>things trickier.
>start	stop	rxno	ender	bracedate
>5-Jan-11	12-Jan-11	2248	0	5-Jan-11
>12-Jan-11	26-Jan-11	3645	0	12-Jan-11
>5-Jan-11	12-Jan-11	2248	1	12-Jan-11
>19-Jan-11	30-Jan-11	2261	0	19-Jan-11
>25-Jan-11	25-Jan-11	5061	0	25-Jan-11
>25-Jan-11	25-Jan-11	5061	1	25-Jan-11
>12-Jan-11	26-Jan-11	3645	1	26-Jan-11
>19-Jan-11	30-Jan-11	2261	1	30-Jan-11
>Generalized schematic of the brace counting approach is below, but I am
>open to other solutions. Creating a dummy variable for each date and using
>0,1 is not an option as far as I can tell because of the number of
>prescription records (55 million) and length of time (3 years).
>Overlapping intervals:
> (-------------------)
>                      (----------------------)
>                                                    (---)
>      (---------------------)
>                                               (-----------------)
> Consolidated:
> (-----(-------------)-(-----)----------------)      (----(---)--------)
>Your help is genuinely appreciated!
>Nabarun Dasgupta
>Department of Epidemiology
>Gillings School of Global Public Health
>& Injury Prevention Research Center
>University of North Carolina
>Chapel Hill, NC 27599 USA

*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index