Employee Benefits Captives Surge as $1M Claims Jump 30% and Stop-Loss Premiums Spike

Willis Towers Watson's 2025 Benefits Trends Survey puts more than 40% of employers using or actively considering captive arrangements for employee benefits, with the recent growth concentrated among employers under 500 lives, a segment that barely appeared in captive formation data five years ago.

The driver is not the captive structure itself. It is that the claim distribution stop-loss pricing was built on has changed shape, and mid-size employers are absorbing the repricing.

Key Takeaways

• Million-dollar claims per million covered employees rose 29% in a single year and 61% over four years, per Sun Life's 2026 report across more than 70,000 high-dollar claims at 3,300 self-funded employers.
• 49% of plan sponsors had at least one claimant above $1 million in the most recent two policy years, up from 23% in the prior IFEBP survey cycle.
• Stop-loss premiums rose 9.7% on average across Segal's 221-plan dataset for plans keeping comparable coverage, and 7.3% even where employers accepted benefit changes.
• Stop-loss PEPM falls from $229.40 at a $100,000 specific deductible to $68.90 at $250,000, which is the arithmetic pushing attachment points upward.
• Captive Resources members earned dividends in 98% of accident years, with 71% of years distributing more than 15% of contributed loss funds.

The Claim Distribution Changed Shape

Segal's 2025 dataset across 221 health plans put the average stop-loss premium increase at 9.7% for plans holding comparable coverage, and 7.3% for those that accepted benefit changes and raised deductibles. The IFEBP's 19th Medical Stop-Loss Premium Survey, covering 1,268 policies and more than 1.2 million covered employees, found single-year increases of 8.8% at a $100,000 specific deductible rising to 10.1% at $500,000, on compound growth of 9.9% to 12.1% over multi-year periods.

Underneath the pricing is a claims environment that has moved rather than drifted. Sun Life's 2026 High-Cost Claims Report, drawn from more than 70,000 high-dollar claims across 3,300 self-funded employers, shows million-dollar claims per million covered employees up 29% in one year and 61% over four. Claims above $3 million rose 47% year over year, ten individual cases passed $5 million, and the largest single claim reached $12.7 million.

The IFEBP survey corroborates it from the sponsor side: 49% reported a claimant above $1 million in the most recent two policy years, against 23% in the prior cycle. Cancer drives 92% of catastrophic claims, with blood cancer treatment episodes averaging $5.45 million per claimant.

That is a regime change in the tail rather than a trend in the mean, and it is what makes the intermediate risk layer a captive provides worth the structural complexity to an employer who previously defaulted to fully insured.

The Attachment Point Is the Whole Decision

Where the specific deductible sits determines how risk splits between the employer's retained layer and the captive pool, and it is moving fast. The modal specific deductible among mid-size captive members ran $125,000 to $150,000 three years ago. Programs now increasingly set it at $200,000 to $250,000, some larger members higher.

Specific Deductible	Avg. Stop-Loss PEPM	Annual Cost (300 EEs)	Retained Risk Profile
$100,000	$229.40	$825,840	Low variance; most large claims transferred
$250,000	$68.90	$248,040	Moderate; retains mid-severity cancer, surgery
$500,000	$50.96	$183,456	Higher; significant catastrophic exposure retained
$1,000,000	$17.69	$63,684	Very high; full gene therapy and cancer episode exposure

The savings are real. For a 300-employee group, moving from a $150,000 to a $250,000 specific deductible saves roughly $90 to $120 per employee per month, $324,000 to $432,000 a year. What the employer buys with that is the corridor between $150,000 and $250,000 per claimant.

The expected cost of that corridor is small. Assume the group expects 0.8 claims a year above $150,000 and 0.4 above $250,000 on current frequency. The corridor is then triggered about 0.4 times a year, at an average retained amount near $60,000, for an expected annual retained cost of roughly $24,000 against $324,000 to $432,000 of premium saved.

The expectation is not the exposure. Two or three corridor claims in one year erases several years of savings, and the frequency figures those 0.8 and 0.4 estimates come from are drawn from a period whose tail has just moved 29% in twelve months. Selecting the attachment on the expected value alone prices the corridor as if the distribution behind it were stable. It is the variance that has to be modelled, and it is the variance the recent data has changed.

Pooling is what makes the middle layer work at all. The coefficient of variation of per-employee claims falls with one over the square root of the covered population, so a 200-employee group carries roughly 3.5 times the relative volatility of a 2,500-employee group. A captive of 100 employers averaging 250 employees pools 25,000 lives and produces aggregate volatility comparable to a single large employer, which is what lets the mid-severity claims between the specific deductible and the catastrophic layer be priced actuarially rather than treated as shocks.

What the Pool Cannot Price

Two things sit outside what pooling solves, and both are getting larger.

The first is cell and gene therapy. Single-treatment costs run from $2.2 million for Casgevy to $4.25 million for Lenmeldy, and more than 60 additional therapies are expected to reach FDA approval by 2030. That combination raises catastrophic event frequency across the pool while each individual event stays far too rare for plan-level credibility. A frequency model for it cannot be fitted to claims history, because the history predates the treatments; it has to be built from disease prevalence and treatment eligibility, which is a different estimation problem with different error properties. Captive programmes are responding with dedicated gene therapy sub-layers and carve-out reinsurance, effectively a fourth tier.

The second is selection inside the pool. When a member's experience deteriorates through a demographic shift, an acquisition bringing a higher-risk population, or ordinary bad luck, the captive has to weigh fairness to the other members against that member's expectation of rate stability. Programmes handle it with corridor credibility adjustments that raise individual experience weighting gradually rather than applying a rate shock, and the whole mechanism depends on a credibility floor that sits near 50 employees. Below it, the member is effectively community-rated within the pool regardless of its own experience, which is the same stop-loss credibility problem the structure was meant to escape.

The dividend record is where these two meet. Captive Resources reports members earning dividends in 98% of accident years, with 71% of years distributing more than 15% of contributed loss funds and 58% more than 20%. That record was built on the claims distribution that existed before million-dollar frequency moved 29% in a year, against per-employee health costs projected above $18,500 in 2026. A dividend history is a description of the loss years that produced it, not a property of the structure.