Insurance is the wrong way to pay for chronic care.

Insurance is a bet against unlikely events

Insurance is a bet against unlikely events. You pay a premium so that if something rare and expensive happens, you're covered. The math works because most people paying in won't need to collect.

Auto insurance is the clean example. You pay your premium, you almost never interact with the insurer, and when you do it's because something genuinely expensive happened. The administrative machinery activates rarely, on events that warrant it. The admin-to-value ratio is reasonable.

Chronic care is not unlikely

60% of American adults have at least one chronic condition. 51% have two or more. Once you have one, the probability of developing another is roughly 70%. Among adults between 18 and 34, 27% already have multiple chronic conditions.

This is the baseline of American health. CMS recognizes it: annual wellness visits, chronic care management codes, transitional care management codes all exist because the agency understands chronic care as recurring, scheduled, ongoing. A rough but not inaccurate analog is that chronic care is similar to routine maintenance on your car. New brakes, oil changes, AC recharges etc.

Chronic care is inexpensive

Did you really just say that? Yes, I did. Call it clickbait.

The question isn't whether chronic care is expensive, it obviously is. The question to ask is why it is. Costs concentrate at two places: high utilizers seeing clinicians frequently, and pharmaceuticals. The pharmaceutical piece is real and worth its own treatment. But the clinical utilization piece deserves scrutiny, because seeing a clinician frequently shouldn't be expensive.

Two reasons.

Intelligence doesn’t compound.

Early autonomous vehicle systems were expensive not because driving is inherently costly, but because the cost was in the hardware and the human operator sitting in it. You weren't paying for transportation. You were paying for the apparatus required to produce it.

Over time, as those systems accumulated experience and got progressively better at the task, the major cost center could be removed. The marginal cost of transportation got cheaper because the cost center shifted.

Medicine doesn't work that way. Every encounter starts from scratch. Clipboards. Rushed conversations. The same questions, the same baseline, re-established from zero. You pay not for the care itself but for the overhead required to produce it, every single time, without accumulation. Systems that don't compound stay expensive.

Routing predictable care through adjudication exponentially increases admin cost

But unlike routine auto maintenance, insurance is involved in nearly every healthcare decision. Every prescription, every referral, every lab order. The same machinery that makes sense for a $200,000 hospitalization is processing a $35 blood draw. Administrative cost scales with frequency, not severity. The less catastrophic the event, the more admin it generates per dollar of care delivered.

The shape matters more than any single point on the curve. As the coverage threshold moves left into the bulk of the utilization distribution, admin burden grows exponentially. Each new category of routine care pulled into adjudication adds friction faster than it adds value.

"Okay, but we can solve the actuarial problem with duct tape and zip-ties."

That's the industry's answer. More prior-auth staff. Utilization-management software. AI for claims processing. Each generation of investment tries to lower the marginal cost of adjudicating a $35 blood draw, which makes it economically viable to adjudicate even more $35 blood draws.

The infrastructure cost gets baked into prices. Higher prices justify higher reimbursement. Higher reimbursement justifies more infrastructure to manage it. Every actor inside the loop is responding rationally to local incentives. The system as a whole gets more expensive each cycle.

The first-principles answer is to stop routing predictable care through that path at all. You cannot optimize your way out of a category error.

What insurance is actually good for

Risk pooling has a job. Its job is the tail.

Within chronic populations, the tail is real and concentrated. A small fraction of patients accounts for the majority of spending. Inpatient costs rise sharply as conditions accumulate. Preventable hospitalizations run into the tens of billions annually. Some patients will progress, require high-cost interventions, and generate spending no subscription model absorbs.

That's what insurance was designed for. Genuine, unlikely, expensive events.

The other 70 to 80% of chronic care is predictable, recurring, and manageable. The right instrument for it isn't a claims apparatus. Direct primary care already demonstrates this: significant reductions in emergency utilization, lower risk-adjusted total cost, measurably better chronic disease control.

And DPC is a relatively blunt instrument. It removes administrative friction and adds continuity, but it doesn't compound intelligence across a patient's history, and it doesn't extend into specialty.

The results it produces before doing any of that should tell you something about where the ceiling is.

Insurance covers what's left. The actual tail. The thing it was built for all along.