On Mon, 03 Nov 2003 17:24:22 +0000, John E wrote:
> "Marc Schwartz" <[EMAIL PROTECTED]> wrote in message
> news:[EMAIL PROTECTED]
>>
>> There appears to be "something" fundamentally different about
>> group/hospital C, but without more information (ie. are the three groups
>> sampled differently), it is difficult to understand what that might be.
>
> Thanks for that Marc. I should have given more information on the data.
>
> They are the waiting times in the emergency room for three different
> hospitals. I have to write a report which identifies the similarities and
> differences between the relative performances of the hospitals.
That being the case, there is the potential for confounding based upon a
variety of hospital specific factors, some of which may be:
1. Time of day/day of week the measures were taken (which may affect both
staffing and patient mix)
2. General patient differences across the three hospitals
3. Type of ERs (ie. is each designated at different Trauma Levels)
4. Admission volumes at each
5. Urban versus community hospitals
6. Academic versus non-training hospitals
7. Call staff available
etc.
Unless you adjust for these and other relevant factors, univariate
comparisons can be highly misleading especially if one is to rank the
three based upon this measure of "performance".
Based upon the samples you gave ("in a vacuum") Hospital C could feasibly
be a busy trauma center, where the high outliers are low risk patients,
who might be better served going to their doctor or to an urgent care
center. The "bulk" of C's patients seem to get seen quickly. The data
could fit a trauma triage scenario.
A and B seem to have a more "normal" distribution of patients, but in
general, take longer to be seen. This might suggest a lower risk
patient profile. It could also be that the staff are generally not as
efficient...or there are lower staffing levels to handle the load.
It would be interesting to know what the admitting diagnoses are for C's
high outliers.
Those are first impressions, but again could be way off base without
knowing the other factors.
With sufficient information and proper sampling, you could feasibly
develop standards around a regression model, against which, each type of
hospital could be "risk-adjusted" and measured (within some confidence
level). That would at least level the playing field a bit.
HTH,
Marc Schwartz
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
