Did I misread your numbers or does it really seem to indicate that 95% of them were released and
only 5% were considered serious enough for admission? If so that most certainly does seem to back
the figures I've been slowly gathering.
"Robert Chung" <
[email protected]> wrote in message
news:[email protected]...
>
> "Tom Kunich" <
[email protected]> wrote
> > The problem first appeared when I started trying to make heads or tails out of the DOT
> > statistics. As you've probably seen, they
report
> > fatalities and "injuries" which is any time anyone goes to the hospital. Well 95% or more of
> > those who go to the hospital have
minor
> > injuries and are treat and release (sounds like something from one
of
> > those wildlife shows).
> >
> > Finally I started pulling out a statistic here or there mostly
that
> > had to do with the numbers of hospital cases ADMITTED to the
hospital
> > from accidents.
> >
> > A lot of people go in, but the important number are those who
stay,
> > even if for just one night. Those are the serious cases.
>
> It turns out that what you've written is close, but not quite exact.
I used
> to be pretty familiar with hospital discharge data, and I drafted
one of
> many attempts to collect ER data. The ratio of hospital ER visits to hospital admissions isn't
> really known because there isn't
systematic
> collection of ER data. In addition, there are trends in hospital
admission
> behavior so that something that a physician might have considered a
reason
> for admission in the past might not be considered so today, or that
the
> decision to admit under one health system may not be the same under
a
> different system. For example: I've observed that the standard for
hospital
> admission appears to be set much lower in France than in the U.S.;
and U.S.
> researchers have been investigating whether the standard for
hospital
> admission is different for public and private hospitals, or rural
and urban
> hospitals, or for-profit and non-profit hospitals, or whether the
patient is
> poor or wealthy. It's clear that there are trends that vary between
seasons,
> and that admissions are sensitive not only to the demand from
patients but
> also the supply of beds (which one they're more sensitive to is the
big
> question). Whether these differences are a good thing or a bad thing
is
> arguable (and I have) but the fact is that they exist. That means
that
> "seriousness" isn't as easily determinable by whether or not a
patient was
> admitted to the hospital as one would hope, even among trauma cases
where
> one might think the decision would be a bit more straightforward.
>
> > I've never been able to get a hold on what I'd call real ratio
which
> > would require reading all of the available medical reports and
using
> > the injury indexes to determine what "serious" really is. But I
think
> > that if we were to be able to get the information it would
probably be
> > something like 3:1 serious injuries to fatalities.
>
> As I mentioned in my previous message, that depends on the mix of
types of
> injuries and the sampling distribution of forces. Anyway, that's a statistical issue (which is
> where I was originally going with my
research--I
> was looking at sampling distributions of different things, not bike accidents).
>
> I happen to have had the 1998 California hospital patient discharge
data on
> my disk for another project. Just for you, I ran a little query for admissions for APR-DRGs 55,
> 56, and 57 (basically, head trauma) and
got
> about 13,000. Then, I looked at these 13,000 admissions to see how
many of
> them had listed external injury codes having to do with bicycles
(the
> majority of head trauma appear to be folks over the age of 80 who
fall while
> walking--I have no idea whether the injury/fatality ratio is the
same for
> them as for cyclists, but I suspect not so I thought it more prudent
to try
> to separate them out). I got a bit under 600 (but note that E-codes
are
> notoriously poorly recorded, and nearly another 1000 cases were
blank so the
> real number can be anywhere from maybe 600 up to as high as 1600,
though
> this latter number I'd consider very unlikely; it would mean that
every
> single blank was bike-related). I have no information here on deaths
before
> someone ever got to the hospital, or the number of ER visits (I do
have
> whether a person died in-hospital, but I forgot to do the subquery
of head
> trauma admissions where the patient subsequently died), or certainly
the
> number of people who were so slightly injured that they didn't go to
a
> hospital at all. I describe how I did this because you can see that
at each
> step there is going to be some uncertainty in how cases got coded or
how
> people separate themselves into different categories.
>
> Anyway, FWIW, those are the numbers. I can't help you with the
"real" ratio,
> but I've given you some bounds on the numerator, at least for
California in
> 1998.
