Date: Wed, 22 Dec 2010 11:06:39 -0800
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: "Dennis G. Fisher, Ph.D." <dfisher@CSULB.EDU>
Organization: California State University, Long Beach
Subject: Re: Compare risk adjusted mortality rates
Content-Type: text/plain; charset="iso-8859-1"
I will take a stab at this. I would assert that what you have is two
percentages which is the same as two proportions. Assuming that we can
treat the two proportions as independent, then you can do a test of the
significance of the difference between two proportions as either a
chi-square or as a binomial which are mathematically equivalent. The
easiest way in SAS is as a chi-square using PROC FREQ.
Dennis G. Fisher, Ph.D.
Professor and Director
Center for Behavioral Research and Services
1090 Atlantic Avenue
Long Beach, CA 90813
tel: 562-495-2330 x121
From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of Morley
Sent: Wednesday, December 22, 2010 10:36 AM
Subject: Compare risk adjusted mortality rates
I had asked this question previously, but in the interests of brevity, I
probably did not make myself clear (or 'perfectly clear - as noted in A Few
In risk adjusting cardiac patient deaths, we use two factors. The raw
mortality rate (opmort) which counts the number of patients dying in a fixed
period of time expressed as a percent of the number of surgical cases . Thus
we might have 20 deaths in 1000 cases so opmort = 2.00%.
Surgical patients run the gamut from those relatively healthy and needing
possibly a single artery operation to those needing multiple bypasses, while
also having co-morbid diseases, eg: diabetes, obesity, increased age etc.
There is an algorithm that produces a predicted risk of mortality (prom) for
each patient based on a series of factors. Values range from 0 to 100%
theoretically (the bulk of the patients range 5-25%).
To compare performance between years, between centers etc, an
observed/expected (O/E) ratio is calculated. Numerator is the opmort rate
while the denominator is the prom value. A value of 1 indicates the
mortality rate is exactly as predicted based on patient disease severity,
while < 1 shows performance better than expected.
Question - how do I test between two O/E ratios?
Say in year 2008 I have 1000 patients with a prom of 2.07 ± 2.11 and opmort
rate of 22 deaths in the 1000 cases so the rate is 2.2%. Now the O/E ratio
Then in year 2009, there are 1500 patients, prom is 1.99 ± 1.86 while opmort
is 19/1500 or 1.27%. Now the O/E ratio is 0.64.
I want to know how to test if 2008 is statistically different from 2009. I
have an individual prom for each patient, but obviously the opmort is a
yes/no count variable.
Thanks in advance for any assistance
Wishing all list members the best of the season.