Dark Patterns Didn't Just Enter Healthcare. They Were Engineered Into It.

In behavioral economics, there's a useful pair of terms. A nudge makes a behavior easier — a default enrollment in a retirement plan, a prominent recycling bin, a pre-filled form. A sludge makes a behavior harder — extra steps, confusing options, friction designed to wear you down until you give up. Both are forms of intentional design. One is in your interest. The other usually isn't.

A recent Freakonomics Radio episode, "Sludge, Part 1: The World Is Drowning in It," made a claim that should stop you cold: a lot of healthcare sludge isn't accidental. It's intentional. It's how insurance companies ration care in order to profit-maximize. The friction you experience trying to understand your deductible, find an in-network provider, navigate a prior authorization, or get a denial reversed — that friction was designed. Not by accident, not by bureaucratic inertia, but as a deliberate strategy.

This is dark patterns at their most consequential scale. Not a pre-checked newsletter box. Not a hidden unsubscribe link. A system built to deny lifesaving medical care through engineered exhaustion.

The industrialization of "no"

To understand how this happened, it helps to look at what the industry actually built.

In 2022, a ProPublica investigation revealed a company called EviCore — a subsidiary of Evernorth, which is owned by Cigna. EviCore's business model is straightforward: insurers like Cigna, UnitedHealthcare, and Aetna outsource their prior authorization reviews to EviCore. EviCore reviews the claims and decides whether to approve or deny them. The insurers pay EviCore for this service. The incentive structure does not require stating explicitly — it is sufficient to observe that EviCore's revenue comes from insurers who benefit from denials, and that its denial rates were high enough to attract a federal investigation.

Prior authorization is the process by which a doctor must get a insurer's approval before a treatment, drug, or procedure is covered. Originally framed as a cost-control mechanism, it has evolved into something else. A 2023 KFF survey found that 16% of insured adults experienced prior authorization problems in the past year. Among Medicare Advantage enrollees, the rate was 22%. More importantly: when denials were appealed, they were overturned at a rate of 82% in Medicare Advantage plans. That figure is worth pausing on. More than 4 in 5 denials that were appealed were found to be wrong. And the vast majority of people who receive a denial never appeal.

That gap — between the overturn rate and the appeal rate — is where the money is.

1.2 seconds per case

In 2023, a ProPublica investigation documented what Cigna's internal review process actually looked like. The company had developed a system called PXDX — shorthand for "procedure to diagnosis" — that automatically cross-referenced diagnosis codes with procedure codes to flag claims for denial. Medical reviewers using the system spent an average of 1.2 seconds per case. In two months in 2022, the system denied more than 300,000 claims.

To be precise about what 1.2 seconds means: it is less time than it takes to read a single sentence. It is not a review. It is a rubber stamp on a computer-generated decision. The human reviewer in this process is not a safeguard — they are a legal formality, present to allow the company to say that a physician reviewed the claim.

Cigna's response to the reporting was that the process was appropriate for straightforward, low-complexity denials. The question the response doesn't answer is what the error rate was — and whether the 82% overturn-on-appeal figure from Medicare Advantage gives any indication.

The algorithm that was wrong 90% of the time

UnitedHealth Group, the largest health insurer in the United States, developed an AI model called nH Predict for managing post-acute care decisions — specifically, how long patients recovering from surgery or illness would be allowed to stay in a rehabilitation facility before Medicare Advantage coverage was cut off. A federal class-action lawsuit alleged that the algorithm had a 90% error rate: that is, when patients appealed denials made by nH Predict, the original decision was overturned nine times out of ten.

The lawsuit further alleged that UnitedHealth's internal data showed the algorithm was systematically wrong, and that rather than correcting it, the company continued to deploy it because the denial rates it produced aligned with corporate financial targets. Human reviewers who overrode the system's recommendations reportedly faced internal pressure to conform to its outputs.

The 90% figure is not a rounding error or a methodological artifact. It is a description of a system that was correct less often than a coin flip — and that was used to deny care to elderly patients recovering from major medical events, until a court intervened.

The architecture of abandonment

Dark patterns on websites rely on a specific psychological mechanism: they make the "wrong" path (wrong for you, right for the company) easy, and the "right" path (right for you, wrong for the company) hard enough that most people give up. The confusing menu. The buried cancellation link. The shame-based opt-out. The pre-checked box. Each of these works because the majority of users, facing friction, choose the path of least resistance.

Health insurance denial systems work on exactly the same principle, at vastly higher stakes.

The denial letter arrives. It is written in a language that is technically informative but practically impenetrable. It references an appeals process with a deadline. The appeals process requires documentation — from you, from your doctor, sometimes from a specialist. Your doctor's office has to spend time on this. You have to spend time on this. Many people are sick, or managing a sick family member, or working, or simply exhausted. The path of least resistance is to accept the denial, change plans, go without the treatment, or pay out of pocket. The insurance company collects the premium either way.

This is sludge as a business model. The friction isn't incidental to the product — it is the product.

What this means for everyone building digital products

Health insurance is an extreme case. The stakes — medical care, financial ruin, sometimes life itself — are not typical of the industries where dark patterns usually appear. But the mechanism is identical to what happens in subscription services, e-commerce checkouts, and cookie consent flows. The difference is scale and consequence, not kind.

The Freakonomics framing of "sludge" is useful precisely because it names the category without limiting it to any one industry. Intentional friction, deployed to benefit the entity creating it at the expense of the person experiencing it, is sludge. It appears in bureaucracies, in software, in insurance claims processing, in subscription cancellation flows. The tools differ. The intent is the same.

Understanding this is the first step toward building something different. Every product decision that makes the honest path easier — the cancellation as easy as the signup, the privacy setting as prominent as the share button, the total cost as visible as the monthly rate — is a choice against this model. Those choices compound. They are also, it turns out, what trust is made of.