My firm is considering participating in the Heritage Health Prize, an X-prize inspired project seeking to reward a means to predict 1 year in advance which patients will be hospitalized. They provide claims data, which is an interesting building block for predictions because of how ambiguous claims data (Oh, we'll this patient as billed for a LVL 3, diagnosis being VXX.XX) seems - at least to me.
I think what is most interesting about this is how we're searching for a discovery, for previously ignored or unknown patterns. I think we can guarantee we'll soon have an abundance of data. If we get good at extracting lessons and building knowledge, these cost centers can become like in-silico repositories of generations and generations of human biology.

