Intermediation in Healthcare Delivery

The chain is usually drawn as a line: an innovation is manufactured, distributed, ordered by a provider, paid for by a payer, and used by a patient. It is a useful sketch and a poor description.

Real delivery is rarely that linear. The party that approves a product is often not the party that buys it, the party that bills for it is often not the party that made it, and the party that benefits economically when it works is often none of them. A business model built on the sketch tends to discover the real chain after the money is spent.

It is more accurate to read the chain as nine flows that happen to travel together: clinical authority, payment, reimbursement, distribution, workflow, data, regulatory responsibility, risk and patient access.

Each flow can stop at a different party. Clinical authority may sit with a physician who spends none of the money. Reimbursement may sit with a supplier who touches none of the clinical work. Risk may sit with a health system, an employer or a risk-bearing group that never appears in the sales conversation.

A business model becomes interesting when these flows can be reorganized: when the party that carries the risk can also be the party that pays, or when the entity that bills can also be the entity that runs the workflow. Most durable healthcare businesses are a rearrangement of these flows rather than a better product.

The list is a working inventory, not a taxonomy. For each model it gives the economic engine and what the model demands of the operator, because the demands usually decide whether a model is available to a given company.

Treat each one as a hypothesis about where value can be captured, testable against a specific innovation and a specific chain.

The first pass on any opportunity is descriptive. Who currently intermediates between this innovation and the patient, and why does each of them exist?

Separate the intermediary functions that are legally, clinically or operationally necessary from the ones that persist because of historical workflow, contracting complexity, information asymmetry or legacy technology. The second group is where business-model work lives.

Then follow the money and the control. Where does margin accumulate? Who controls purchasing, prescribing, patient access and reimbursement? Who bears risk, and who receives the greatest economic benefit when the intervention works? The answers are frequently different parties, and the gap between them is often the opportunity.

Is there an existing reimbursement mechanism for the underlying device, service, procedure, supply, monitoring activity or outcome? Fitting inside an existing mechanism is almost always faster than establishing a new one.

Which entity is legally able to bill, and what accreditation, enrollment, documentation, coding, clinical and billing requirements come with that? Those requirements are the real cost of the reimbursement, and they are often what decides whether a manufacturer can capture it at all.

Ask directly whether the intermediary can capture more economic value than the manufacturer. Where the answer is yes, the interesting question is not how to sell the device but whether to occupy the billing position. Recurring reimbursement, where it exists, is what makes a recurring business model possible rather than aspirational.

For every opportunity, identify these roles separately before assuming any of them coincide.

Can an intermediary be removed at all, or does law, licensure or clinical practice require it? Can several intermediary functions be consolidated into one operator? Can AI perform work currently performed by human coordination inside an intermediary?

Should the innovator become the intermediary? The honest version of that question is whether becoming one improves economics or merely adds operational and regulatory burden. Occupying an intermediary position means accepting its accreditation, staffing, working capital and audit exposure.

Where it is worth doing, the reason is usually structural: controlling the intermediary creates proprietary data, workflow control, distribution or patient access that a product sale never produces. And in some chains the incumbent intermediary is a far better partner than an opponent, because it already holds the contracts, the enrollment and the customers.

Business-model choices reach back into the product. Could a different model make the product simpler? Could consumer-grade components replace specialised hardware? Could hardware become a low-cost endpoint for a higher-value recurring service?

If downstream economics are strong enough, the device can be subsidised or given away. That is only true when software, data, consumables or reimbursement are genuinely worth more than the hardware, which is a claim to test rather than assert.

The last question in this group is the sharpest: can the company sell an outcome instead of a product? It is the most attractive position and the one that most often exceeds a young company’s ability to measure and finance.

Each business model carries a different regulatory load. Selling a component is not the same as becoming a manufacturer, and becoming a provider, DME supplier, laboratory, distributor or pharmacy attaches an entirely new body of obligations, inspections and liabilities.

So the question is not only which model earns most, but which model produces the lowest regulatory burden for the value captured, and what specifically changes if the company takes on a regulated role.

Regulatory execution itself can be the advantage. Where AI-native methods materially lower the cost of quality systems, regulatory documentation, verification, validation, reimbursement research, compliance, contracting and post-market work, a company can hold a position that incumbents priced as prohibitively expensive. That is a claim this area intends to test with real filings and real timelines, not to assume.

The framework below is meant to be reused. Later research pages should apply it to a specific technology, patent, company or market opportunity and publish the answers, so that separate studies can be compared.

Native Alpha here is structural, not technical. It asks what the innovator understands that incumbents assume cannot change, and whether that understanding produces a cost structure a competitor cannot match by trying harder.

The most common candidate is an intermediary whose margin depends on expensive human coordination: eligibility checks, prior authorization, documentation, referral chasing, equipment logistics, benefit navigation. If AI collapses that coordination cost, the intermediary’s price is exposed and its position becomes contestable.

The other candidates are reimbursement knowledge that turns out to be worth more than the technology, and workflow or distribution control that defends better than a product ever will. Combining consumer technology, existing reimbursement and AI-native operations can produce a structure incumbents find awkward to copy, because copying it means cannibalizing their own labor-based revenue.

None of this follows from patents, novelty or technical merit. Intellectual property can enable a particular business model and can make a licence saleable, but it does not create commercial value on its own. A patent with no reimbursement path, no buyer and no distribution is an asset with a maintenance bill.

Start from the narrowest business model that could prove demand, and ask whether the model can be tested before the complete technology exists. In most healthcare cases it can: a manual, staffed version of the service tests the money, the buyer and the workflow long before the product is finished.

The evidence question is what scarce resource the customer will commit. Money, patients, clinical workflow, implementation effort, data, reimbursement claims or contractual rights all count. An expression of enthusiasm does not.

Then compare models on shortest path to revenue, least outside capital, recurring economics, and whether the model produces proprietary data, distribution, workflow position or switching costs. Financeability matters too: some models read well and finance badly.

Finally, name the traps. Models that require holding inventory, fronting equipment or waiting on claims carry working-capital demands that kill otherwise sound companies. Models concentrated in a single reimbursement code carry the risk of one coverage decision. Models that involve becoming a regulated entity carry obligations that do not pause when revenue does.

Future studies under this pillar use one format, so that different opportunities can be read side by side and revisited when facts change. Each field is short and sourced where a source exists.

Do not start by asking how to sell the innovation. Start by mapping how value, authority, reimbursement, risk, workflow and money currently move through healthcare. Then determine whether the opportunity is to sell through the existing intermediaries, remove them, combine them, automate them, or become one of them.

Business-model architecture is itself the object of research: the best company is often not the one that invented the device or the software, but the one that found a structurally better way to deliver, finance, reimburse, distribute or operate it.