On Jul 16, 2012, at 6:55 PM, Noah Silverman wrote:
> Hello,
>
> I'm tasked with putting together a cost effectiveness analysis on a
proposed medical treatment.
>
> The "standard" suggested by someone is an expensive commercial
package names "TreeAge Pro" which looks like its just a fancy GUI on
top of a decision tree. However, we don't have it and it is expensive to
buy.
>
> Questions for the group:
>
> 1) Are there any interesting R packages for decision trees?
>
> 2) Has anyone here done a cost effectiveness analysis using R that I can
show to the group as an example?
>
> Thanks!
Hi Noah,
I have not been involved directly in performing CEA. However, where studies that
I have been involved with have a CEA component, that part has typically been
carved out to pharmacoeconomics folks, in the case of drug studies. Many
industry sponsors (drug and device) will have reimbursement strategy folks
either in-house or via outside consulting and CEA is frequently a component of
their modus operandi, since FDA approval is only one part of getting a medical
product to market. Insurance companies also have to be willing to pay for them
and that decision can be independent of FDA approval. Thus, these types of folks
will have the focused expertise to guide the design and analytic process and I
might suggest that you seek out such expertise for collaboration.
A standard reference text that seems to commonly come up in this domain is:
Cost-Effectiveness in Health and Medicine.
M.R. Gold, J.E. Siegel, L.B. Russell, and M.C. Weinstein (eds).
New York: Oxford University Press, 1996
http://www.amazon.com/Cost-Effectiveness-Health-Medicine-Marthe-Gold/dp/0195108248
I am a bit confused by the consideration that TreeAge Pro is considered a
standard, notably in light of the fact that they seem to consider MS Excel to be
their primary competition
(http://www.treeage.com/support/TreeAgeProVsExcel.html) and have even created an
Excel plug-in package. That seems counterintuitive to me.
That being said, they do appear to have academic and student licenses available,
which might make it a more affordable choice, if your collaboration would be
facilitated by using a common tool. Consider that cost versus the
"opportunity cost" of you developing something new or perhaps having
to modify an existing tool to make it suitable for use in your study.
A search for R related resources did come up with some possibilities:
ArvoRe Package
http://arvore.r-forge.r-project.org/
ICEinfer package:
http://cran.r-project.org/web/packages/ICEinfer/
Lastly, if the above two are not suitable, you might want to review the Machine
Learning Task View, since these analyses seem overlap to an extent with that
domain:
http://cran.us.r-project.org/web/views/MachineLearning.html
Regards,
Marc Schwartz