Indices are a problem across ebooks. Page numbers are no longer relevant in a digital space, which in itself makes 90% of an index immediately useless. To address this problem in books with simple indices (one page reference per item), publishers sometimes delete the page number and attempt to link the item to its corresponding reference (links being the digital equivalent to page numbers: faster, but not smarter). Creating these links is manual work that is both expensive and prone to error.
The more complicated the index, the more complicated the solutions to translate it for digital use. For example, the indices often contain multiple page references per item. Should publishers keep the page numbers as a means of having individual items to link to each particular reference, knowing that page numbers no longer apply? Regardless of what solution is reached here, since the linking is manual work, the more complicated the task becomes, the more expensive and the more prone to error the task becomes.
To be honest, of course, creating a digital index is not any more manual or prone to labor than creating a physical index, but digital publishing, though booming, is nowhere near the point of paying the sums of money that index experts (fairly) demand, and the basic reading experience that current devices aim for actually preempt putting anything close to this effort into a product. (Tablet publishing is another beast, perhaps to come to bear, but perhaps not, consider my solution below.) Therefore, we rely on the same companies to whom we outsource our digitization work, which itself even poses problems in titles that contain only narrative text and share the same escalating problems of complexity.
To respond to the index problem, I recommend implementing a solution to aggressively remove all indices from all titles. You may not feel like this is a perfect solution for your books, but let me offer the primary positive reason for which we made this decision.
All ebook readers (devices, apps, and other instances) have search functionality. Readers will have varying comfort levels searching for important terms, trained by their use of Google and other search engines. The solution is not perfect but is actually more likely to return what the user was looking for than an index, especially when you consider the human error that goes into making the digital index.
The ability to search for relevant terms is synonymous with a level of comfort with technology assumed in owning an e-reading device and provides an opportunity to improve reader experience without notifying them that a more frustrating option (a digital index) could also have been available.