![](https://webarchive.library.unt.edu/eot2008/20081030212808im_/http://cdc.gov/ncidod/eid/images/spacer.gif)
|
![](https://webarchive.library.unt.edu/eot2008/20081030212808im_/http://cdc.gov/ncidod/eid/images/spacer.gif) |
Dispatch
Costs of Surgical Site Infections
That Appear after Hospital Discharge
Nicholas Graves,*†
Kate Halton,*† Merrilyn Curtis,* Shane Doidge,* David Lairson,‡ Marylou
McLaws,§ and Michael Whitby*
*The Centre for Healthcare Related Infection Surveillance and Prevention,
Brisbane, Queensland, Australia; †Queensland University of Technology,
Brisbane, Queensland, Australia; ‡University of Texas Health Science Center
at Houston, Houston, Texas, USA; and §University of New South Wales, Sydney,
New South Wales, Australia
Appendix 3. Description of
Statistical Analyses
All analyses were carried out using Stata software (Stata Statistical
Software: Release 9.1, Stata Corp., College Station, TX, USA). Evidence
of multicollinearity was assessed by nested auxiliary regression, each
variable was dropped from the model and the R-squared values compared
to a complete model (i.e., the vif command in Stata was used to estimate
variance inflation factors and the tolerances). Nonnormality in dependent
variables was assessed by visual inspection of data plots, and the Breusch-Pagan
test was used as a formal test for heteroskedastic errors (i.e., the hettest
command in Stata). Terms that described interactions between socioeconomic
and clinical variables were included, and the constrained model was compared
with a number of unconstrained models; incremental F-tests were conducted
(i.e., the test command in Stata was used). Because the objective of the
analyses was to estimate the independent effect of surgical site infections
on cost outcomes, we included all available explanatory variables in a
general model and did not seeks a general-to-simple reduction based on
lack of statistical significance.
|