A lot of that trove of existing data is locked up behind privacy law barriers, and there's no incentive for doing the work to strip out the PII and make it available for aggregate analysis. And since we're talking about the costs for performing medical treatments, the details that would need to be stripped out could be significant.

For example, if Procedure A tends to be more costly when the patient also needs Procedure B, that's significant. But if the aggregate information includes the fact that the same patient needed both A and B, that might be usable to identify the patient, and so the relationship can't be shared in the data.