[EpiData-list] Re: Dealing with 'Ongoing' duration in epicurves

epidata-list at lists.umanitoba.ca epidata-list at lists.umanitoba.ca
Tue Dec 2 13:40:59 CST 2008


"I would appreciate it if those working with outbreaks could reply about 
how they enter and display "ongoing" duration of illness." We create a 
graph showing at what time (date) the symptoms start and graph this 
according to exposure or time window.

For epicurve the standard way of thinking is:
..............dayonset (start) .................................
...... dayonset (start) .....................................
command could be: epicurve case dayonset

In a different instance we would like to analyse "pattern of disease 
development" or pattern of how long time disease lasts
begin ----------------------- end                       
begin -----------------------------------------                      
(still under observation)

The typical analysis here could be a Survival curve indicating length of 
disease period:
Assume we have cases and string variables for start and end in the 
analysis. We would :
read ourdata (or read from clipboard since these are copied from a 
spreadsheet).
* generate two date variables for start and end of period:
gen d stop = .
if end <> "ongoing" then stop = date(end, "%dmy")
gen d start = date(begin, "%dmy")
* generate a case variable indicating if people were cured or are still 
under observation:
gen i well = 0
if end <> "ongoing" then well = 1

* now we are ready for the analysis. notice the /mt which will assume 
that everyone with end time missing were observed until the latest of 
those where "stop" was "non-missing".
lifetable well start stop     /mt

* an alternative would be to give the stop as a precise value: e.g. Dec 
1st 2008
if end = "ongoing" then stop = dmy(1,12,2008)  
lifetable well start stop 

Notice that we did not change the "well" to 1 since keeping this with 
the value 0 indicates this observation was observed until given stop 
date, but did not develop disease.

The lifetable would show median survival, possibly grouped  or compared 
to a reference with the options "/by=sex  /p25 ....."

Regards
Jens Lauritsen
EpiData Association


More information about the EpiData-list mailing list