Your intuition that this may be different than SNP to SNP LD is correct. In fact, it is already known that using traditional SNP-SNP LD metrics can lead to incorrect results. Thus, specific methodologies to study this have already been published for dealing with this issue, although there is not a lot out there, yet ...
You might try this resource: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3109359/ - it is a framework for determining SNP CNV LD that uses a custom method, and if it is unable to determine it with the new method it will, otherwise the program resorts to default SNV - SNV LD determination metrics.
Furthermore, in the Web Resources section, it says that the code is available on request. I believe that last author, Dr. Hemant Tiwari, is at UAB and is an active researcher, so you would likely be able to get it from him or the first author.
This paper describes the two in relation to haplotype, which would also be very good background for you: http://bioinformatics.oxfordjournals.org/content/26/11/1437.full
Beyond this, I think combing through Google and Pubmed Results will get you up to speed on what has and has not been done in this area. Good luck and let us know what you find.