The SMR is the maximum likelihood estimator of the mortality rate ratio between the study population and the reference population, assuming:
1) The rates in the reference population are known without error
2) The rates in the study population are proportional to those in the reference population (otherwise it would not be meaningful to have a single rate ratio).
The "expected deaths" is the cumulative intensity in the study population, assuming the rates were as in the reference population. This is only approximately equal to an expected number of deaths; it is easy to devise an example where the "expected number of deaths" in the study population exceeds the actual number of persons in the study population.
Best regards,
Bendix Carstensen
> -----Original Message-----
> From: A UK-based worldwide e-mail broadcast system mailing
> list [mailto:[log in to unmask]] On Behalf Of Juergen Wellmann
> Sent: 19. juli 2010 11:07
> To: [log in to unmask]
> Subject: Query: standardized mortality ratio
>
> Dear all,
>
> the standardized mortality ratio (SMR) is computed as the
> number of observed deaths divided by the number of "expected
> deaths". The latter is computed as the observed person years
> times the mortality rate in some reference population.
> (Actually, nominator and denominator are usually split up as sum over
> age- and calendar intervals.)
>
> What does this statistic estimate --- what is the "true SMR"?
>
> I understand "expected deaths" as number of deaths that would
> have occurred if the study population had the mortality rate
> of the reference population. How is it justified then to use
> the observed person years which result from the mortality of
> the cohort under study rather than the mortality of the
> reference population?
>
> Any hints are appreciated.
> Juergen
>
> You may leave the list at any time by sending the command
>
> SIGNOFF allstat
>
> to [log in to unmask], leaving the subject line blank.
>
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|