Date of Death vs. Date of Registration

Never confuse the two (I haven’t so far).  This is an interesting problem in real-time surveillance of public health, for both morbidity and mortality.  The fact is that health quantification is largely administrative, and there is some lag time associated with almost all of the available databases.  The closest exception in BC is the PharmaNet database, which logs medications as they are dispensed.  These data are reliably captured in near-real-time for all pharmacies hooked into the system, which is the vast majority of them.

Deaths get registered less quickly.  The rule-of-thumb figure seems to be that 95% are registered within 7 days, but that’s of little value to us if we want to know that something weird happened yesterday.  Earlier this week I cobbled together some code that will automatically plot mortality vs. temperature for us on a daily basis, but now I need to start generating the data that will help us to assess whether yesterday was weird today, rather than a week from now.  I am working under the assumption that there will be some deviation from the day-after-death registration baseline on days when death is significantly elevated, but the first step is to establish the baseline.  This is a different sort of problem than I used to tackling in R, but I think that I have worked it out:

Advertisement