Disregarding the fact that I really like your question (since this is a field that I am very much interested in), I think the answer is not easy to give in one or two sentences.
When you say "learning disabilities", the term contains possibly dozens, if not hundreds of different variants of disability. A similar question would be "how to create a good survey for physically disabled people" - it makes a strong difference if the person has lost his arms or is blind: The solution has to be a very different one.
Exemplary, my solution for autism would completely differ from a solution for persons with reading problems. And even then, the form of autism differs from case to case.
Even if I do not like my answer this time - I think the case you want to cover might be too broad, the solution will not be 100% satisfying for all of your target group.
Rather, I would start with listing the disabilities' effects. To re-use the analogy: A blind person cannot use a screen. One without arms cannot use the keyboard. Once you identified the problems of each group you want to target, I'd try to group it, find similarities. This might help you finding a common set of issues you have to solve, which then brings you to a position where you can start solving the actual problems.
As an example, if you find that 9 of 10 participants tend to have problems with attending a problem for more than 3 minutes, this tells you about the maximum length or complexity of the question. If you see, that only 3 of 10 tend to be attracted to voice output of the question, you might reconsider the effort. I think you understand the direction I am going for: Instead of trying to solve the problem in a universal way, I would try to understand the differences by analyzing the problems each target group has. This, actually, is the standard usability improvement way and not particularly only fitting for your special audience.
I really would be interested in the results of such research - so if you are willing to share, I'd be happy to participate or see the results!
EDIT: One thought popped on my mind after posting - I'd suggest to more invest into usability research topics first, before you go for solutions - so "how to conduct usability research with learning disabilities", instead of interaction patterns, to begin with.