If the data is rigorously anonymized, then it loses all value for the health insurance / advertising sector, who want to exclude high-cost patients from coverage and target individuals with health conditions, respectively.
Do you think Big Tech shareholders (with additional holdings in pharmaceuticals, petrochemical, agribusiness etc.) would be interested in a database of rigorously anonymized patient data intended to discover relationships between things like pesticide exposure and Parkinson's disease, or unpleaseant side effects of currently profitable pharmaceutical products?
I basically don't believe non-degenerate (psued)anonymization is possible, although that complicated af homomorphic encryption stuff makes me a little uncertain.
* Hypothesis generation sped up
* Candidate participant discovery sped up - including finding controls
* large classes of analysis possible on just the data, including difference in difference and natural experiment