Date: Fri, 25 Jun 1999 17:03:26 -0400
Reply-To: "Baker, Harley E,, DMDCWEST" <BAKERHE@OSD.PENTAGON.MIL>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: "Baker, Harley E,, DMDCWEST" <BAKERHE@OSD.PENTAGON.MIL>
Subject: Re: comparison to a benchmark
I think we need more information to answer this question. I am guessing that
the complication rate is some type of ratio (patients with
complications/patients without complications; patients with
complications/total number of patients; etc.) if this is the case, SPSS
offers two relatively easy ways to answer your question.
First, assuming that the rates are based on at least 100 patients in the
denominator, and assuming that the complication rates are between .20 and
.80, you could use a one-way ANOVA with MD as the independent variable, and
patient complication status (yes = 1, no = 0) as the dependent variable. If
the overall F is significant at your prespecified level, then you could
conduct either post hoc Bonferroni tests, or you could specify a
Scheffe-type complex test of the one MD vs the mean of the other five MDs
(as David suggests.) This is a standard approach and is quite easy to do in
SPSS. Since the unit of analysis is now the individual patient (rather than
the MD) it will likely not suffer from low power.
Or, if you prefer, you can do the same thing as a chi-squared problem. It
would be a 6 (MD) X 2 (individual patient outcome) chi-squared. If the
overall is significant, you could then collapse it to a 2 (MD vs other five
MDs) X 2 (patient outcome: complication vs no complication.)
Both of these approaches start with the raw patient-level data, and not with
the complication rate. If all you have is the rates and number on which the
rates were derived, it could still be done using the ANOVA procedure that
allows you to enter the means, SDs, and sample sizes.
If I am wrong about the nature of the complication rate, of course, neither
of these two approaches will be useful.
Defense Manpower Data Center and
School of Education, University of San Francisco
> -----Original Message-----
> From: David F. Greenberg [SMTP:dg4@IS3.NYU.EDU]
> Sent: Friday, June 25, 1999 11:36 AM
> To: SPSSX-L@LISTSERV.UGA.EDU
> Subject: Re: comparison to a benchmark
> If you use the mean complication rate for all six and use that for
> purposes of comparison, the two rates will not be independent, because the
> mean rate will incorporate data for the one MD you want to compare with.
> It would be better to compare the one with the five. However, your samples
> are extremely small for this purpose. - David Greenberg, Sociology
> Department, New York University.
> On Fri, 25 Jun 1999, Dina Wilke wrote:
> > I'm interested in testing if one MD's complication rate is significantly
> different from the rest of the department's (6 other MDs). I thought I
> would use the mean complication rate for the 6 MDs as the benchmark, and
> then do a test of proportions. Is there a way to do this in SPSS? I have
> the formula to calculate it by hand, but I wondered what I was missing in
> the program.