Their profit margins are higher than average but not exceptional.
https://www.financecharts.com/compare/UNH,CI,ELV,CVS,MOH/sum...
But even that doesn't mean anything.
I worked for a company that used ML to fight health insurance fraud ("mistakes") in the German market - where doctors/administrators would require payment for extra procedures that they never actually performed. So "denial rate" or even "profit margin" would go up with better AI that discovers more fraud!
Maybe a good metric would be something like "expected lifespan"? But even that can be misleading (e.g. maybe different insurers serve different populations...)