The Bundle Is the Problem

The prevailing view in small group health benefits holds that the bundled insurance product, combining network access, pharmacy benefits, and catastrophic protection into a single monthly premium, exists because these three functions are interdependent. Separate them and you lose the risk pooling that makes coverage affordable for sick people. Separate them and you lose the administrative efficiency that makes the product manageable for a 20-person employer. The bundle is not a design choice. It is a structural requirement.

This article argues the opposite. Health insurance for the 1-to-50 employer market bundles three functions that are not structurally interdependent: a negotiated discount on provider charges, a negotiated discount on prescription drug purchases, and catastrophic financial protection. The first two are purchasing functions. The third is actual insurance. Everything else in the architecture, the plan documents, the claims adjudication infrastructure, the prior authorization apparatus, the broker intermediation, the TPA operational stack, exists to manage the bundling. Not to manage the risk. To manage the fact that a purchasing function and an insurance function have been welded together into a single product and must now be administered as though they were one thing.

The Three Functions

#

When a small employer buys a fully insured or level funded health plan, three distinct economic transactions occur simultaneously under a single contractual wrapper. The first transaction is access to a negotiated provider network. The employer’s covered employees can see physicians and use hospitals who have contracted to accept discounted rates. The discount itself has real value: across 2,379 hospitals examined by researchers using hospital price transparency data, commercial negotiated rates averaged 58 percent of the corresponding chargemaster prices, representing a discount of approximately 42 percent from the hospital’s standard list price (Jiang et al.). Against Medicare rates, private employer plans fared differently: RAND Corporation’s Round 5 Hospital Pricing Transparency Study, analyzing data from more than 4,000 hospitals covering 2020 through 2022, found that private employer plans paid an average of 254 percent of Medicare rates for inpatient facility services and 289 percent for outpatient services. The network discount is real. It is also a purchasing discount, not a risk transfer.

The second transaction is access to negotiated prescription drug pricing. The PBM contracted to the plan negotiates rebates from manufacturers, manages the formulary, and sets the copay tiers that determine what members pay at the pharmacy counter. The PBM function is a purchasing function. It produces discounted prices through volume aggregation and formulary positioning. The risk element of prescription drug coverage, the possibility that a member will be prescribed a $500,000 gene therapy or a $80,000 biologic, is a separate, insurable event. The purchasing infrastructure and the insurance event are bundled together in the pharmacy benefit as though they require the same administrative apparatus.

The third transaction is the only genuine insurance event: protection against catastrophic medical costs that exceed the employer’s or employee’s capacity to absorb. For a level funded employer with 15 employees, the stop loss carrier provides specific coverage at a threshold that commonly runs between $20,000 and $50,000 per member annually, and aggregate coverage that limits the plan’s total annual outlay to a predetermined corridor. The stop loss mechanism is the insurance function. It is currently the back end of the product architecture. This article’s argument is that it should be the front end.

The bundled product exists because, historically, separating these functions was operationally impossible and commercially unprofitable for the intermediaries managing the assembly. Neither condition is still true.

Pharmacy Pricing Is a Purchasing Function

#

The clearest demonstration that the pharmacy discount is not an insurance function is the existence of products that perform the discount function without any insurance relationship. GoodRx operates as a prescription discount card, negotiating access to lower drug prices through pharmacy networks and selling that access to consumers directly, without an insurance contract. Mark Cuban’s Cost Plus Drugs publishes manufacturing cost plus a fixed 15 percent markup and a $5 dispensing fee for generic drugs, making the pricing completely transparent and requiring no insurance intermediary. The FTC’s Second Interim Staff Report, released in January 2025, documented that GoodRx prices were referenced repeatedly as offering lower costs than PBM-negotiated prices for the same drugs at the same pharmacies, without any insurance relationship at all.

The FTC report on the pharmacy benefit management industry quantifies what bundling the purchasing function with the insurance function has cost plan sponsors. Between 2017 and 2022, the three largest PBMs, Caremark (CVS), Express Scripts, and OptumRx, generated more than $7.3 billion in dispensing revenue from specialty generic drugs in excess of their estimated acquisition costs. Spread pricing, the practice of billing plan sponsors more than the PBM reimburses the dispensing pharmacy, generated an estimated $1.4 billion in additional income over the same period. Plan sponsor payments for commercial claims grew at a compound annual growth rate of 21 percent between 2017 and 2021. The FTC found markups on specialty generic drugs of hundreds and, in some cases, thousands of percent, including a markup on tadalafil, a pulmonary hypertension medication, of 7,736 percent for commercial payers in 2022.

These margins exist because the PBM occupies the position it does inside the bundled insurance product. The PBM is the gatekeeper to the pharmaceutical purchasing function, and that gatekeeper position is protected by the bundled architecture. A small employer cannot separate the pharmaceutical purchasing function from the rest of the health benefit without effectively exiting the bundled product. They can use transparent PBMs like Smith Rx or Capital Rx as alternatives to the Big 3, but they cannot easily purchase the insurance component and the pharmacy discount component from separate, unrelated parties under the current product structure. The bundle protects the gatekeeper.

For small employers, the per-member economics of PBM services are particularly poor. The PBM rebate and formulary infrastructure is designed for populations of thousands, which is where the negotiating power with pharmaceutical manufacturers actually lies. A 15-person employer’s claims volume gives the PBM no meaningful bargaining position with any manufacturer. The employer pays for purchasing infrastructure it cannot fully use. What value the employer receives flows not from its own purchasing scale but from being pooled with the carrier’s broader book of business, at an administrative cost that is opaque by design.

What Remains Is Catastrophic Protection

#

Strip the network access function and the pharmaceutical purchasing function from the product. What remains is the actual insurance problem: protection against a medical event whose cost exceeds the employer’s and employee’s capacity to absorb. This protection already exists in level funded plans. The stop loss carrier provides it. The specific attachment point for a level funded small group typically falls between $20,000 and $50,000 per member annually depending on group size and carrier (see LFP-02.04 for attachment point mechanics and ranges), with the aggregate attachment adding a second protection layer.

The argument this article makes is structural: the stop loss function should be the front end of the product architecture, not the back end. An employer designs a benefits program by first answering the insurance question, what catastrophic threshold triggers financial protection, and then assembles the purchasing functions separately to manage routine and intermediate costs. The catastrophic policy activates when costs exceed the attachment. Everything below the attachment is handled through direct purchasing arrangements: a provider discount network accessed at a per-member monthly fee, a pharmacy discount card or transparent PBM arrangement, and a direct primary care membership for routine care.

This is not a high-deductible health plan with a savings option. An HDHP/SO is still a bundled insurance product. It has a plan document, a claims adjudication infrastructure, network contracts, an EHB-compliant benefit design, a prior authorization apparatus, and the full TPA operational stack. The member pays more out of pocket before the insurance kicks in, but the administrative architecture is identical to any other bundled plan. The model this article describes has no plan document for routine care because there is no plan for routine care. There is a discount arrangement and an insurance policy. The two documents are different contracts with different counterparties.

The economics of the unbundled model are structural, not incidental. A fully insured or level funded plan for 25 employees at the 2024 small-group average of $9,131 for single coverage represents approximately $228,000 annually in premium for single-coverage-equivalent. If half those employees carry family coverage at the $25,572 family average, total premium approaches $430,000. That cost funds the network access function, the pharmacy benefit function, the catastrophic function, and the full administrative stack that exists to manage all three as a single product. The unbundled model isolates each cost and eliminates the overhead that exists only to hold the bundle together. What the employer stops paying for is the cost of running three economically distinct functions through one bureaucratic apparatus.

Direct Primary Care as Evidence

#

Direct primary care already separates one function from the bundled product. A DPC practice charges a flat monthly membership fee and provides unlimited primary care access with no per-visit billing, no claims adjudication, and no insurance relationship for the covered services. The DPC physician is not contracted to the employer’s health plan. No prior authorization occurs for primary care services. No explanation of benefits document is generated for an office visit. The administrative layer that exists to manage the bundled product is entirely absent from the DPC transaction.

The evidence on employer DPC adoption is accumulating. Hint Health’s 2025 employer trends analysis found that 58 percent of all DPC memberships in 2024 were employer-sponsored, up from 21 percent in 2017, representing a significant shift in who pays for the DPC relationship. Employer retention is high: 85 percent of employers maintained their DPC relationship after the first year and 70 percent remained at two years, suggesting that the value proposition survives initial adoption. Case studies cited by Hint Health indicate that employer groups with DPC arrangements spent approximately 52 percent less than comparable non-DPC cohorts in the same period. The American Academy of Family Physicians documented 2,688 DPC practices operating across all 50 states as of 2024, with 9 percent of family physicians reporting DPC practice in 2023, up from 3 percent in 2022.

DPC layered onto a catastrophic wrap-around policy is the closest existing approximation of the unbundled architecture. The gap is scope: DPC covers primary care only. Specialist access, hospital care, and pharmacy remain inside the bundled product. The unbundled thesis extends the DPC logic across the full care spectrum. Network access for specialist and facility care becomes a discount arrangement, not an insurance product. Pharmacy becomes a transparent purchasing arrangement, not a formulary administered by a vertically integrated PBM. The catastrophic policy remains as the one genuine insurance function.

The Payvider Extension

#

Payvider organizations, provider systems that assume payer risk, demonstrate a different route to the same structural destination. When a health system contracts directly with employers to cover a defined population for a fixed monthly fee, the network access function and the insurance function collapse into a single entity. The provider is not negotiating discounts with itself. The administrative intermediary disappears. No TPA adjudicates claims for care delivered within the provider’s own system. No network aggregator charges licensing fees for access. No broker recommends the product because the employer and the provider negotiate directly.

Kaiser Permanente has operated on this model for decades in its core markets. Intermountain Health, Geisinger Health, and Aultman Health have developed direct employer contracting products. The payvider model is growing in large employer and Medicare Advantage segments. Its application to the 1-to-50 market is limited by geography: a payvider product requires a regional health system capable of covering primary, specialty, and facility services within a reasonable service area, and most rural and suburban small employers do not have a qualifying health system within reach. Where the infrastructure exists, however, the structural logic is complete. The employer buys care from a provider who is also the payer. The bundled insurance product has no function in that transaction.

Why the Bundle Persists

#

The regulatory barriers to unbundling are real, though more limited than commonly assumed for the 1-to-50 market. ERISA requires a plan document for any employer health plan. ACA essential health benefit requirements do not apply to self-funded plans at any employer size, which eliminates one frequently cited constraint. State insurance mandates do not reach self-funded plans under ERISA preemption, eliminating another. MHPAEA parity requirements apply to self-funded plans only when the sponsoring employer has more than 50 employees. For the 1-to-50 market that is the subject of this series, MHPAEA does not apply to a self-funded plan, removing a third constraint that is routinely misapplied in industry discussions. The stop loss carrier’s insurance certificate remains state-regulated, which constrains specific attachment point floors in states that have enacted the NAIC Stop Loss Model Act or equivalent. That is a real constraint, but it affects the catastrophic layer only, not the purchasing functions this article describes separating from it.

The infrastructure barriers are also real. No platform currently assembles a DPC membership, a reference-based pricing or direct provider network, a transparent pharmacy discount arrangement, and a catastrophic stop loss policy into a single enrollment and member experience. The components exist separately. Integration is an engineering problem, not an invention problem, but it is a real problem. An employer who wants to assemble the unbundled model today faces multiple vendor contracts, multiple ID cards or no ID card, and a member experience that requires active navigation of which arrangement to use for which need.

The economic barriers are where the analysis gets direct. Every entity in the current value chain loses revenue if the bundle breaks. The fully insured carrier loses premium. The TPA loses claims administration fees. The PBM loses spread pricing income and affiliated pharmacy revenue. The network aggregator loses access fees. The broker loses commission on premium. These entities have structural incentives to maintain the architecture that generates their revenue, and they exercise those incentives through lobbying, through carrier contract terms that make unbundling commercially difficult, and through the advisory relationships with employers that the broker accountability apparatus reinforces.

The unbundled model is structurally superior for cost, simplicity, and transparency, measured against the three employer objectives the preface establishes. It is politically and commercially constrained by the entities who benefit from the current structure. The question is not whether the constraint is rational. It is. The question is whether cost pressure, technology maturation, and regulatory evolution erode it.

The market data on small group fully insured enrollment suggests erosion is already underway. Peterson-KFF analysis of commercial market data documents that fully insured small group enrollment fell from approximately 17 million in 2013 to approximately 10 million in 2023. Oliver Wyman’s analysis of NAIC data found a 26 percent decline in fully insured small-group enrollment from 2016 to 2023 alone, a compound annual rate of 6 percent. Among the remaining small group risk pool, the departure of healthier employers has worsened morbidity: multiple insurer rate filings for 2026 cited risk pool deterioration as a driver of proposed small group premium increases averaging 11 percent, according to KFF Health System Tracker analysis of 318 small group insurer filings across all 50 states.

The 2024 KFF survey found that 36 percent of covered workers at small firms were enrolled in level funded plans, up from 34 percent in 2023. Level funded’s growth is the market’s intermediate response to the bundle problem: an architecture that shares some characteristics of the unbundled model, including claims visibility, surplus return, and health-status underwriting, while remaining inside the plan document framework that regulatory compliance requires. Level funded did not solve the bundle problem. It moved the employer partway toward transparency while leaving the PBM, the network aggregator, and the TPA administrative stack largely intact. The full unbundled model goes further, but it requires either ERISA plan document compliance (which reintroduces much of the administrative architecture) or legislative action that creates a tax-advantaged vehicle for the non-insurance components.

The regulatory horizon matters here. ACA essential health benefit requirements do not apply to self-funded plans regardless of employer size, which means a self-funded small employer can already exclude coverage categories that a fully insured small group cannot. ERISA preemption of state insurance law means a self-funded plan does not face state benefit mandates. The stop loss carrier remains subject to state insurance regulation, which constrains product design in states with specific attachment point floors, but the plan itself operates under federal law. The regulatory architecture already permits more unbundling than the market has achieved. The dominant barrier is not regulatory. It is commercial: the absence of an integrated platform that makes the unbundled model as operationally simple for a 20-person employer as a single call to a broker and a bundled premium invoice. That platform does not yet exist as a single integrated product. When it does, the remaining structural argument for the bundle, that it is administratively necessary, loses its foundation.