The ideal solution would be to have a risk sharing system, with minimal administrative overhead. The core component that people need is catastrophic coverage , as that's not a level of un-hedged risk that it makes sense to have, and because people are generally bad at estimating risk, preventative care. If people weren't bad at that, the preventative care would be superfluous, as that's not really an unpredictable risk.
Plans should be standardized and easy comparable ('apples to apples') to limit the
confusopoly that develops around complex risk management systems. This is something that the MA exchanges (Romneycare) and Obamacare have done a decent job at.
Further risk pooling is an option, via single payer, but that relies on central planning, that can be expensive to do correctly, but it's still a net benefit if those publicly borne administration and market distortions costs are less than the private administration and market distortion costs (yes private risk management also generates market distortion, but in some fairly complex and counter-intuitive ways). The big benefit here is the general gain in opportunity cost by not forcing everyone under the state's umbrella to do their own coverage calculations, for the same reason that some states choose to handle personal taxes centrally rather than have every try to color within the lines each year. It also lets you do cost shifting onto those who can pay while also incentivizing those who can't to use the services correctly.
Things that are more or less a net drag on good heath insurance policies are the policies with hundreds of line item exemptions and subsidies; there is no reason each class of care need it's own hedge, you're unnecessarily running premiums up with them. People buy them because they like getting free x/y/z but it's not, they've just shaped the market for specific things. This is one place that I take issue with the ACA package on, as it embeds several of these into the standardized packages. But it's probably still a net benefit as you can do better comparative analysis, even if the plans are suboptimal.