From Jay Kaufman To statalist@hsphsun2.harvard.edu Subject Re: st: adjusted percentages and adjusted odds ratios Date Fri, 19 Sep 2003 13:47:47 -0400

```Magnitude of confounding is a function of 3 parameters:  the
association between the covariate and the exposure, the association
between the covariate and the outcome, and the prevalence of the
covariate.  You have only specified one of these parameters.  If
the other association parameter is null (i.e., OR = 1), then there
will be no confounding at all.  If the other association parameter
is modest, then you will have the potential for some modest degree
of confounding, as you actually observe in your data.  For details,
see:

Flanders WD, Khoury MJ.  Indirect assessment of  confounding: graphic
description and limits on effect of adjusting for covariates.  Epidemiology
1990; 1(3): 239-46.

Lin DY, Psaty BM, Kronmal RA.  Assessing the sensitivity of regression results
to unmeasured confounders in observational studies.  Biometrics 1998; 54:948-63.

Yanagawa T. Case-control studies: assessing the effect of a confounding factor.
Biometrika 1984; 71: 191-4.

- JK

Elizabeth.EBY@med.va.gov wrote:
>
> Hi-
> Could someone help me out with understanding the relatively small difference
> in adjusted percentages (20% vs 17%) compared to a more robust adjusted odds
> ration of 1.7?  The adjusted percentages are controlling for covariates, and
> were obtained using the predict sequence.
>
> Thank you,
>
> Elizabeth Eby
> Research Health Science Specialist
> Ann Arbor VAMC
> 734-769-7100 X6248
>
> -----Original Message-----
> From: Copeland, Laurel
> Sent: Friday, September 19, 2003 11:26 AM
> To: Eby, Elizabeth
> Cc: VHAANN HSRD Statistical Group
>
> The raw odds ratio can be calculated...17% of 334=277, 20% of 2741=2193
>         cut back?
>             no  yes
> priv no   277  57
> ins? yes 2193 548
>
> OR = (277/57) / (2193/548)
>    =  4.86  /  4.00
>    = approx 1.22
> The remainder of the effect in the adjusted OR must be due to the
> covariates.
>
> -----Original Message-----
> From: Eby, Elizabeth
> Sent: Friday, September 19, 2003 11:05 AM
> To: VHAANN HSRD Statistical Group
>
> How would you explain the small difference in the adjusted percentages (17%
> vs 20%) with an odds ratio of 1.7 (p=.04)?  Does it have anything to do with
> the underlying sample sizes of the 2 groups (n=334 VA patients and n=2741
> with private insurance)?
>
> Thanks.
>
> -----Original Message-----
> From: jpiette@umich.edu [mailto:jpiette@umich.edu]
> Sent: Friday, September 19, 2003 10:51 AM
> To: elizabeth.eby@med.va.gov
> Subject:
>
> Liz - can you double check this?  THe difference in %'s seem small relative
> to
> the AOR.
>
> Controlling for their sociodemographic characteristics, number of chronic
> diseases, and number of prescription medications, 17% of VA patients cut
> back
> on medication use due to cost compared to 20 % of those with private
> insurance
> (adjusted odds ratio [AOR]: 1.7, p=.04),

--
Jay S. Kaufman, Ph.D
-----------------------------
email: Jay_Kaufman@unc.edu
-----------------------------
Department of Epidemiology
UNC School of Public Health
2104C McGavran-Greenberg Hall
Chapel Hill, NC 27599-7435
phone:  919-966-7435
fax:    919-966-2089
-----------------------------
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```