Datamining is likely to help detect and clarify negative patterns in large groups (like when stents don't actually reduce angina). But doing the same to find positive patterns to shape policy seems to be much harder due to the wild variation inherent in individual patient response to each therapy.

Existing therapy is refined via specific experiments which are not amenable to aggregation into large groups that can be 'mined' using AI or DM, thereby adding further value. It's not clear to me what AI can add to existing stat/prob methods here.

Inversely, when trying to characterize patients using their raw data, I've found individual patient data to be so horribly noisy and so often mislabeled as to be useless for more than trivial purposes of individual trend summary. This becomes especially woeful when aggregating digital health records, and is likely to remain so for decades to come (given their current glacial rate of improvement).

When it comes to the prospect of mining medical data to improve treatment, I can only say, "GIGO".

Agreed.

>Rather, [the government] should try to open the way for the private industry ...

Disagree, at least in part.

One source of personal frustration is the perception that rich techies who made their fortunes by writing programs without explicit government approval/legislation seem to spend a lot of time begging for government solutions (like UBI) or approvals for these problems without first attempting to use their money, brains, and tech with a little creativity to generate extra-governmental solutions.