Who Level Funded Serves and Who It Fails: The Coverage Map and Its Gaps

LFP-06.SYN | Synthesis | Series 06: The Populations

This series examined ten populations. The synthesis is not a summary of those ten. It is the argument that the pattern of who level funded serves and who it fails maps to five design assumptions embedded in the architecture of the model. Where all five assumptions hold, the model works well. Where any assumption fails, coverage degrades in predictable, population-specific ways. Where multiple assumptions fail simultaneously, coverage degrades compoundingly.

The pattern is not random. It is structural. The failures are not defects in implementation. They are consequences of design choices that were rational for the population the model was built for and are increasingly misaligned with the population that actually works for small employers.

The Five Design Assumptions

#

Level funded plan architecture embeds five assumptions about the member population. Each assumption is operationalized in the mechanics of the product.

The first assumption is stable employment. The plan year runs 12 months. The claims fund is sized for expected claims over that period. The deductible and out-of-pocket maximum accumulate toward annual limits. Stop loss attachment points reset at year end. The entire financial architecture assumes the member is enrolled for the full plan year. A worker employed for four months and then separated does not fit this architecture. The plan’s cost structure per member enrolled assumes duration that high-turnover industries systematically deny.

The second assumption is sufficient group size. Stop loss underwriting requires enough members to absorb actuarial variance. Below 10 enrolled lives, the actuarial credibility of the group collapses. Attachment points at very small group sizes are set so high relative to expected claims that the specific stop loss protection is nominal. The model’s economics break down before the group reaches the size needed to form a credible risk pool.

The third assumption is affordable cost sharing. Plan designs with $2,500 deductibles and $7,500 out-of-pocket maximums assume the member can absorb these costs. The plan functions as intended when the member can afford the cost sharing. The plan functions as catastrophic-only coverage when the member cannot, and research on high-deductible plan effects consistently finds that low-income members reduce both necessary and unnecessary care in response to cost sharing that exceeds their means.

The fourth assumption is network proximity. Leased PPO networks are built for metropolitan areas. The same network that provides 40 participating specialists per category in Dallas provides two or three in a rural county. Provider directories list names that may be disconnected, retired, or no longer accepting patients. The plan document promises network access. No federal regulator tests that promise for ERISA-governed self-funded plans the way CMS tests it for marketplace qualified health plans.

The fifth assumption is health status stability. Stop loss underwriting assumes the group’s health profile will not deviate dramatically from the assumptions embedded in the attachment points and claims fund. A member with hemophilia or a transplant history produces claims that destabilize the actuarial model for a small group, and the stop loss laser is the mechanism by which the carrier transfers that risk back to the employer while leaving the plan coverage obligation in place.

These five assumptions were reasonable for the population level funded was originally designed for: the stable, full-time employee of a small professional services firm, earning moderate income, living in a metro area, without health conditions that would trigger underwriting concerns.

The Assumption Failure Map

#

For each population the series examined, specific assumptions fail.

The 55-to-64 cohort (06.02): the group size assumption fails and the health status assumption is stressed. These entrepreneurs and small business owners have purchasing power and willingness to pay, but their businesses are often 2 to 5 employees, below the threshold where stop loss underwriting works at standard terms. Their age cohort produces health complexity that stop loss carriers price aggressively.

Fractional and portfolio workers (06.03): the employment assumption fails entirely. No single employer exists to sponsor coverage. The BLS Contingent Worker Supplement found that 7.4% of all employed Americans are independent contractors on their main job, with 36% of those contractors aged 55 or older. The high-earning independent professional has purchasing power and demand for group-quality coverage. The architecture has no place for her.

Low-wage workers in level funded industries (06.04): the cost-sharing assumption fails. A $2,575 deductible equals 7.4% of gross income for a worker earning $34,900, which is the median wage for home health and personal care aides nationally. The coverage exists. The member cannot afford to use it for anything short of catastrophe. The Commonwealth Fund’s 2024 Biennial Health Insurance Survey defines this condition as underinsurance. Sixty-six percent of underinsured Americans have employer coverage.

Workers with chronic conditions (06.05): the health status assumption fails operationally through the laser mechanism. The plan must cover the member. The stop loss carrier is not bound by HIPAA’s nondiscrimination requirements and can exclude that member from specific stop loss protection. The employer absorbs the exact catastrophic exposure the stop loss was supposed to prevent.

Mental health in the level funded workforce (06.06): network proximity fails, and plan compliance fails separately. Milliman’s 2019 research found that patients were 5.7 times more likely to use out-of-network providers for behavioral health services than for medical services, reflecting the structural inadequacy of in-network behavioral health provider availability. Additionally, the DOL’s own Reports to Congress on MHPAEA implementation found that none of the NQTL comparative analyses initially submitted to EBSA were sufficient to demonstrate parity compliance. Parity exists in statute. It has not been built into the operational administration of most small self-funded plans.

Rural and limited-English-proficiency members (06.07): network proximity fails for geographic reasons, and the implicit assumption that the member can navigate the system in English fails for a workforce in which 26.4% of home health workers and 23.1% of food service workers speak a language other than English at home.

High-turnover workers (06.08): the stable employment assumption fails. PHI National documents annual turnover rates among home health aides approaching 80%. The BLS JOLTS data shows leisure and hospitality sector total separations running at approximately double the all-private-industry rate. A plan-year framework is not compatible with a workforce that turns over before many members have served out the waiting period.

Undocumented workers in level funded industries (06.09): the employment assumption fails, and the entire coverage architecture fails. Statutory exclusions from ACA marketplace coverage and Medicaid, combined with documentation requirements at enrollment, place these workers entirely outside the coverage system. The Pew Research Center’s 2025 report estimates 10 million unauthorized workers in the U.S. labor force, with construction at 15% unauthorized and agriculture at 14%.

Dependents (06.10): the cost-sharing affordability assumption fails for low-wage employees who cannot afford the family coverage premium increment above their employee-only contribution. The health status assumption is also stressed by adverse selection in dependent enrollment, which concentrates higher-health-complexity spouses and adult children in enrolled dependent populations.

The Compounding Effect

#

The most vulnerable populations are those for whom multiple assumptions fail simultaneously.

Consider a low-wage worker earning $34,900 annually, employed by a home health agency in a rural Texas county, with type 2 diabetes and limited English proficiency. The cost-sharing assumption fails: she cannot afford the deductible. The network proximity assumption fails: the endocrinologist her condition requires is 60 miles away. The language assumption fails: the phone tree, the provider directory, the prior authorization form, and the appeal process are all in English. Her chronic condition creates claims experience that will show up in renewal underwriting and may produce upward pressure on attachment points or aggregate funding requirements.

Each assumption failure would individually reduce coverage value. Together, they produce a coverage experience that is categorically different from what the plan document promises. She has a card. She has nominal access to a network. She has a plan that will pay claims after she meets a deductible she cannot afford, delivered by providers she cannot reach, administered by a system she cannot navigate.

The compounding is not additive. Each failure amplifies the others. The worker who cannot afford cost sharing delays care. The delayed care worsens the chronic condition. The worsened condition increases claims. The increased claims drive renewal rate increases. The employer responds by increasing cost sharing or reducing contribution toward dependent coverage. The coverage for the worker and her family degrades further.

The populations concentrated at the intersection of multiple assumption failures are disproportionately low-income, disproportionately rural, disproportionately non-English-speaking, and disproportionately employed in the industries where level funded adoption is growing fastest. The fit between the model’s design assumptions and the workforce it is now being sold to is deteriorating.

What the Pattern Reveals About the Model

#

The failures are not implementation problems that better TPAs could fully solve. They are design problems that exist because the model was built for a population that is shrinking as a share of small employer workforces, while the industries driving level funded growth employ populations that diverge from those assumptions in specific, measurable ways.

Level funded emerged in the late 1990s as a product for the professional services small business: the 15-person accounting firm, the 25-person law practice, the 40-person engineering company. The workforce in these firms was stable, moderate-income, metropolitan, and healthy enough that stop loss underwriting was routine. The model fit the population.

The industries adopting level funded at scale in 2025 are different. Home health care, landscaping, food service, staffing, hospitality, and light manufacturing employ workforces that are less stable, lower-income, more geographically dispersed, and more likely to include members with chronic conditions, mental health needs, language barriers, and documentation gaps. The model has not changed. The population has.

The question is not whether level funded works. It works for the populations whose characteristics match the design assumptions. The question is whether the populations for whom it works are expanding or contracting, and whether the populations for whom it fails are concentrated in the industries and geographies where level funded growth is occurring. The evidence supports the latter.

The Feed-Forward

#

The populations this series identifies as underserved define the product gap that subsequent series must address.

Series 15 designs a tiered product architecture whose tiers are calibrated to the population segmentation this synthesis establishes. The population that matches all five assumptions can be served by a standard product at a standard price. The populations for whom one or more assumptions fail require either product modification or honest acknowledgment that level funded is not the right fit.

Series 09 and 10 build on the cost driver and cost management analysis that the populations with chronic conditions, mental health needs, and dependent complexity require. The cost interventions Series 10 proposes are most effective when targeted to the populations whose health status or access barriers generate disproportionate claims.

Series MS.04 addresses the fractional worker coverage gap as a strategic question for the TPA market. The population that no current product serves may represent an opportunity for operators willing to build outside conventional ESI structures.

Series 16 designs coverage for the 65-plus population. The 55-to-64 cohort examined in 06.02 will, within a decade, become the population Series 16 addresses.

The synthesis is not an endpoint. It is a map of the populations the model was not built for and what that gap demands from the product design work that follows.