Prev: Model fit for ordinal multinomial models in GENMOD?
Next: Proc Tabulate-adding a row
From: Sigurd Hermansen on 12 Feb 2010 18:14
After looking closer at the problem you describe I can't say that I understand the study design. Are you following many individuals beginning at different points in time and following them for periods of time much shorter than 7.5 years? Also, are you counting exposure of an individual as elapsed time from a starting point or work hours?

While an average annual man hours of ~17,000 looks large at first glance, in terms of elapsed time it represents less than two person-years. I wouldn't think that you would have a sample of two persons and three cases over a 7.5 year interval.

Given that you want to express your results as man hours lost, I'd suspect that you are talking about a disease that a person can catch and then recover his or her health and then catch again. If you are measuring hours lost to illness, that doesn't sound like prevalence or incidence, but instead an estimate of days per year per 1,000, or a ratio of hours lost to an illness and hours worked plus hour lost to that illness.\

Incidence measures require summing of individual times of exposure up to the point of illness or loss of follow-up. If your estimates don't require that level of complexity, why make it more complex than necessary.
S

-----Original Message-----
From: SAS(r) Discussion [mailto:SAS-L(a)LISTSERV.UGA.EDU] On Behalf Of Billy Thompson
Sent: Wednesday, February 10, 2010 3:24 PM
To: SAS-L(a)LISTSERV.UGA.EDU
Subject: Prevalence vs Incidence

This may be more of a biostatistical theory question than anything else.
If we know that the general population has an annual prevalence rate of
2.7% and we have 3 cases in our sample population over 7.5 years, should
we compare the rates in terms of prevalence or incidence? Within our
sample population we have an average annual man hours of ~ 17,000 and want
to express our results in terms of man hours lost to this disease. My
colleague wants to use prevalence while the reviewer is caught between
prevalence and incidence since we are working with a 7.5 year sample
population.

Hope this is clear enough. Thanks for any help.
 | 
Pages: 1
Prev: Model fit for ordinal multinomial models in GENMOD?
Next: Proc Tabulate-adding a row