Another interesting way to look at it is that insurance is fundamentally based on future uncertainty.
Imagine if it somehow became possible to predict with perfect accuracy a person's future payouts under a given insurance policy. Bob goes to buy house insurance and they can predict with certainty that he'll eventually get $400,000 paid out (presumably because his house got destroyed somehow) while Steve is predicted to have nothing paid out.
If the insurance company has access to these predictions, they will naturally charge Bob about $400,000, plus a profit margin, while they'll charge Steve next to nothing. In this situation, insurance policies become nothing more than elaborate savings accounts.
On the other hand, if the clients have access to these predictions, Bob will buy the best policy he can find, while Steve will naturally not buy any policy. In this scenario, insurance companies all go out of business instantaneously because it becomes impossible to make any money, or even break even.
Yet both sides have huge incentives to try to approximate these scenarios as best they can. The better insurance companies can predict people's insurance usage, the more profitable they can be. The better clients can predict their own usage, the less they spend. But let both parties achieve their predictive goals and the whole thing falls apart utterly.
Medical science is driving relentlessly toward that goal, to the extent that many people were being shut out. The ACA tries to attack both sides of it, by preventing insurance companies from charging differing prices based on certain predictive information, and by preventing clients from refusing to purchase the product. In theory, this returns you to a flatter risk pool where all people contribute roughly equally and everybody sort of pretends they don't know that Bob's house is going to burn down.