| |
The denominator
for end-of-life care analyses is always difficult because it should
at least be limited to persons who die in a given time period.
Ideally, the denominator includes only persons whose death could
be anticipated. That is, persons who die in an accident or of
sudden coronary death do not have an opportunity for hospice care.
This is difficult for the reasons noted above (primarily lack
of information on cause of death). Often researchers are forced
to limit their denominators to people with X disease who die and
assume that all people with X disease who die actually die of
X disease. In other cases, it may be sufficient to include all
deaths as the denominator and assume that the proportion of deaths
that could be anticipated is constant across population groups
(e.g. males and females). Clearly, the ideal choice would involve
merging cause of death information with claims to identify the
most precise denominator. In the absence of the ideal--explicit
choices and their rationale--strengths and weaknesses should be
part of the documentation of methods.
|