> The U.S. Department of Health and Human Services uses Palantir to detect Medicare fraud. Having worked in this field, I can say that this is hilarious overkill. The vast majority (in dollar terms) of Medicare/Medicaid fraud is drug companies misreporting their Average Market Price (AMP), the critical variable in how much they get paid. In 2012, Glaxo settled(!) a small bundle of cases, some of which I worked on, fo…
Medicare fraud goes well beyond price reporting by manufacturers. Physicians can bill for more expensive procedures, bill phantom patients, etc. That's a huge part of the multi billion dollar fraud that Medicare has. I assume being able to identify physicians will outlier billing patterns would be helpful in figuring out where to look closer.
The second largest cases are typically against hospital chains that encourage (read: basically force) their coding teams to use particular billing codes. It's easy enough to model this from claims data (and we did). When a chain buys a new hospital, suddenly patients that present with the same diagnosis codes are all getting extensive exams that ostensibly require 45+ minutes of a physician's time and cost 4-10x more.
Again - the large operators (hospital chains) absolutely dwarf individuals (doctors). DOJ barely has the resources to prosecute the largest offenders, and have faced repeated budget cuts. Wasting time on individual doctors or private practices (beyond maybe making examples of a few egregious actors) would be an inefficient allocation of their resources.