> ## Content Index
> Fetch the complete content index at: https://www.thefringetheory.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# In case of emergency, break glass
- URL: https://www.thefringetheory.com/ichra-choice-arrangement-employers/
- Published: 2026-09-25T16:01:53.000Z
- Updated: 2026-10-08T22:52:29.000Z
- Description: An ICHRA, now called a CHOICE Arrangement, swaps a group plan for a defined contribution employees spend on individual coverage. It can help with catastrophic claims, a claims pattern that will not revert and a scattered workforce. The costs are age rating, thin rural markets and narrower networks.
- Author: Derek Winn
- Tags: Theory

Every building you have ever worked in has a red box on the wall with glass across the front and a lever behind it. It is there for the day the ordinary exits stop working, and the glass is there because you are supposed to think for a second before you pull. 

This is my analogy for ICHRA/CHOICE, at least for now. Early adopters of CHOICE plans are largely motivated by issues with plan performance, underwriting rules, and more. Long-term I see a different landscape, which I'll come back to revisit, but for now I'll focus on the current state.

The rationale is simple: it is a departure from the traditional group health plan, and while there may be comfort in the traditional plan, ICHRA/CHOICE can help solve for the cost of that comfort.

The box got a new label this month. On September 3, CMS Administrator Dr. Oz and SBA Administrator Kelly Loeffler stood up at Hancock Health in central Indiana and announced that Individual Coverage HRAs would now be called CHOICE Arrangements, Custom Health Option and Individual Care Expense. They took it to Atlanta the following week, where Oz called it an "underused option that combines many of the tax benefits of traditional group plans with the flexibility of purchasing coverage independently" \[1\]. Nothing about the structure changed. The 2019 rule still governs it \[2\], and the bill that would have written CHOICE into the tax code has not moved in the Senate \[3\]. What did change is that the federal government is now actively selling it, down to a cost estimator spreadsheet on the CMS site.

Given that, it is worth being clear about what an ICHRA actually solves. My read is that it solves a specific set of underwriting problems that a group plan may have few answers for, and that it does so by moving risk rather than reducing it. That makes it an emergency exit, or a stark alternative to the traditional group plan.

Most plan sponsors already know an ICHRA is a defined contribution that employees spend on individual coverage. What this adds is the three problems it actually solves, and the mechanism that lets it solve them. It rests on the 2019 rule, the federal rating rules, Kaiser's employer survey and published gene therapy prices.

- Which of these three problems do you actually have, and for how many years running?
- If you broke the glass, what would change for your people, and what would stay the same?

## Three problems a group plan struggles to solve

The problems are three fold, and they overlap.

**The catastrophic claimant with no remedy.** Let's say you run a 300-life self-funded plan spending about $3 million a year. One dependent is diagnosed with hemophilia B, and the treatment on the table is a one-time gene therapy with a list price of $3.5 million \[4\]. The sickle cell therapies approved in 2023 list at $2.2 million and $3.1 million \[5\]. One claim is now larger than your entire plan.

Stop-loss catches most of it this year. The problem is next year. [A known high-cost claimant](https://www.thefringetheory.com/what-a-deductible-increase-saves/) at renewal generally means a laser, a higher specific deductible on that one person, a higher premium, or all three, and the ongoing cost of the condition stays with you. If you are fully insured in the large group market, the same claim shows up in your experience and your renewal. There is no lever inside the group plan that makes that person's health stop being your rating factor, and the nondiscrimination rules are built so that there is not.

This is mostly a mid-size and large employer problem. Fully insured small group coverage is already rated without regard to health status under the same federal rules as the individual market \[6\]. Level-funded plans are a different story: 37 percent of covered workers at firms with 10 to 199 workers are in one (Kaiser, 2025) \[7\], and they are underwritten at renewal like any other self-funded plan.

**The high claims year that turns into a pattern.** In Kaiser's 2025 survey, 67 percent of covered workers were in self-funded plans, and at firms of 200 or more it was 80 percent \[7\]. A self-funded plan is supposed to have bad years, that is the deal. What it is not built for is three bad years in a row with no sign of reverting. At that point the plan is effectively funding a chronic cost trend out of operating budget, and the tools left tend to come down to cost shifting and benefit cuts, or to a care management program that takes years to show up in the renewal.

**The distributed workforce.** A group plan is one contract, one situs, one network and one plan lineup, and it assumes your people live roughly where your office is. When they are spread across a dozen states, that assumption breaks. The network that works at headquarters may be thin in rural Montana, and the plan design that fits the DC market is overpriced in Ohio. A small or mid-size employer often ends up buying a national carrier's network largely because nobody else can cover every zip code. An ICHRA lets each employee buy coverage priced and networked for where they actually live.

## How the glass works

Here's the mechanism, and it is what changes your exposure to all three at once.

Under an ICHRA you fund a contribution instead of claims. That converts an open-ended medical claims liability into a defined one. The arrangement is generally still an ERISA group health plan, with the employer mandate and the compliance work still attached, but the claims are no longer yours.

The employee takes that contribution to the individual market, which is not experience rated. Under 45 CFR 147.102, individual premiums may vary by family composition, rating area, age and tobacco use, and not by health status or claims history \[6\]. Once the whole plan moves, the dependent on the $3.5 million therapy is a member of a state's individual market risk pool, priced like every other person their age in their rating area. Their health stops being a rating factor for you. The three bad years stop compounding into a fourth. The employee in rural Montana buys a Montana plan.

There is also a member side to this, and it is easy to undersell. Under a group plan the employee gets whatever the employer picked. Under an ICHRA each employee chooses the insurer, the network and the plan design that works best for them, a lean plan for the 28-year-old who rarely sees a doctor, or a richer one with the right specialists for the family that needs them. In a decentralized model that choice is the differentiator for your people, as long as the local market offers one, which is where the sharp edges below come in.

An ICHRA is a decision about how you fund coverage for your workforce, not about any one person on it. There is a second order effect hiding in that, particularly for employees 65 and over, and it is the subject of part 2.

## What you stop doing

There is a fourth thing an ICHRA solves, and in a lot of the conversation around CHOICE it is the real reason. It is the job.

Running a group plan is an operating function whether or not anybody calls it one. Somebody renews it every year, and the renewal is a negotiation. Somebody picks the carrier and the plan lineup, reads the benchmarking, sits through the claims review, and manages the stop-loss placement, the TPA, the PBM and the point solutions. Somebody answers the question every spring about why it went up again. Under an ICHRA most of that apparatus collapses into one decision, which is how much you are willing to give, and one number, which is the contribution.

Let's face it, that is a philosophy and not a plan design. It says the employer's job is to fund coverage rather than to run it. Some employers have been operating that way in everything but name for years, raising the deductible and stepping back a bit further each renewal. Others built care management, direct contracts and centers of excellence because they believed the plan could be run better than the market runs it, and for them the contribution gives up the thing they were proud of. Both are legitimate positions. The point is to choose one on purpose rather than back into it because the renewal was bad.

## Why it holds

The employer market is starting to test it. Kaiser found 4 percent of firms offering health benefits, and 9 percent of firms that do not, gave at least some employees money to buy individual coverage in 2025 \[7\]. The HRA Council counted more than 500,000 people covered through an ICHRA at the start of 2026, with applicable large employers more than doubling in a year \[8\]. In the EBRI and Morgan Health survey of about a thousand benefits decision-makers, 36 percent of large employers and 23 percent of small ones said they were very likely to adopt within two years \[9\].

The people who would sell it are warier. TASC's 2026 broker survey found 54 percent had never sold one, up from 47 percent a year earlier, and full support fell from 48 percent to 29 \[10\]. Employee confusion is the reason brokers cite most often. So the exit exists, the government is pointing at it, and most of the market has not walked through it yet.

## What it costs

The exit is not free. Age and affordability are covered here. The network is where most of the downside lives, so it gets the next section to itself.

Start with age. Individual coverage is priced per person, and on the federal default curve that most states use, a 64-year-old pays three times what a 21-year-old pays for the identical plan \[11\]. A few states set their own curves, and New York and Vermont do not rate by age at all. [Your composite rate hides that today](https://www.thefringetheory.com/greening-rate-aging-workforce-cost/). A flat contribution under an ICHRA exposes it, and it is a decision about your older employees whether you meant it that way or not. The 2019 rule generally lets you vary the contribution by age, up to three times the youngest participant's amount \[2\], but you have to choose to.

Then there is affordability, and for an employer large enough to owe the mandate it is a conversation for your broker or advisor to drive, with your counsel, before it is a question for the spreadsheet. The published test generally starts from the lowest cost silver plan for self-only coverage in the employee's rating area, subtracts the monthly contribution, and compares what is left with 9.96 percent of household income for 2026 and 10.22 percent for 2027 \[12\]. That number sets your floor, and the enhanced premium tax credits that used to soften the individual market expired at the end of 2025 \[13\].

## Be cautious of sharp edges

The broken glass exit may have sharp edges, and most of them come down to where your people live. The downsides are four fold.

**The contribution buys something different in every rating area.** States draw their own rating areas, usually out of counties or metro areas, and individual premiums are set separately in each one \[6\]. A flat allowance does not follow those differences, so one that buys a gold plan in one rating area may buy a bronze plan two hours away. For a distributed workforce, the feature that solves the network problem creates a pricing one.

**Some markets have one door.** By Kaiser's count, 165 counties had a single marketplace insurer for 2026, up from 93 the year before, and one county in three had fewer insurers than it did in 2025 \[14\]. My own count from the CMS plan files lands within one county of that \[15\]. Either way, the thin markets are rural. Using the USDA's rural-urban codes, nonmetro counties are 62 percent of all counties and 88 percent of the single-insurer ones. About one remote rural county in ten has a single insurer, against fewer than one metro county in fifty, and the median benchmark premium runs about 16 percent higher in those remote counties, by my count \[15\]\[16\]. For 2027, Kaiser's tracker counts nine carriers leaving the marketplaces against six entering new states, and Cigna is leaving the individual market in all eleven of its states \[17\]. In a single-insurer county, the "choice" in CHOICE is a choice of metal level inside one network. The companion tool maps this county by county.

**The network is built to be local.** On Kaiser's analysis of 2021 network data, 84 percent of marketplace enrollees were in HMO or EPO plans, which generally cover no non-emergency care outside the network, and the average plan included about 40 percent of the physicians practicing nearby \[18\]. As of 2025, 37 percent of Americans lived in a state with no marketplace PPO at all (Robert Wood Johnson Foundation) \[19\]. By my count from the 2026 CMS plan files, 56 percent of counties had no marketplace PPO, and unlike insurer count, that share is about the same in metro and rural counties. Most counties had at least one plan with a deductible under $1,000, but the typical plan's deductible was $2,000 or more in nearly all of them \[15\]. A 2022 study in the American Journal of Managed Care, drawn from hospital transparency files, put exchange negotiated prices near 89 percent of commercial group rates \[20\]. [If that is where the discount comes from](https://www.thefringetheory.com/what-a-network-discount-buys/), the savings and the narrower network are the same fact viewed from two sides. It is also where the practical complaints start. A dependent at college in another state, or an employee who spends half the year on a client site, is often covered for emergencies only once they leave the service area.

**The calendar does not bend.** Individual policies generally run on the calendar year. If your plan year starts in July, your employees' deductibles still reset on January 1, halfway through your plan year. And every fall the market can change underneath you, as carriers exit and networks move, so employees may be shopping again whether they wanted to or not.

None of this rules out breaking the glass, but it does mean looking at a map of your own workforce first. The companion tool, [CHOICE: The individual market where your people live](https://www.thefringetheory.com/ichra-choice-employer-tool/), does that for a list of ZIP codes and ages, and its figures are estimates for a first look, not quotes.

## What to do

1. Name the problem before you name the solution. If it is one claimant, price the laser and the next two renewals first. If it is a three-year pattern, price the exit. If it is geography, count how many rating areas you actually have people in.
2. Map your workforce by rating area and count the insurers and PPOs available in each one. [The companion tool will do the first pass](https://www.thefringetheory.com/ichra-choice-employer-tool/) from a ZIP and age list. If a meaningful share of your people sit in single-insurer counties, that is your answer on network before it is your answer on price.
3. Compute your real per employee per year from three years of paid claims, administration and stop-loss premium, and put it next to what the individual market charges for your census in your rating areas. Your renewal is not the comparison.
4. Decide who would move. The whole company is the simplest version to model, and the rules on offering it to only part of a workforce are technical enough to settle before anyone runs numbers.

## The structural point

Breaking the glass moves your risk into a larger pool. It does not change what the pool is buying, and it does not make the dependent on the gene therapy any less expensive to the system, only to you. For an employer whose claims have gotten away from it, or whose workforce no longer lives in one place, that is still a trade worth pricing. An exit is worth having on the wall. Before anyone pulls it, it is worth knowing where it leads.

*Next in the series: part 2,* [*It's good to have CHOICE(s)*](https://www.thefringetheory.com/ichra-choice-side-effects-employers/)*, on the second order effects and what happens to demographics, claims and retention once the glass is broken.*

## Sources

1. Centers for Medicare & Medicaid Services, CMS, SBA, and Georgia Recognize Innovative CHOICE Arrangements as Major Win for Small Businesses, press release, September 14, 2026; Daily Journal (Franklin, Indiana), Dr. Oz visits New Palestine, September 4, 2026.
2. Health Reimbursement Arrangements and Other Account-Based Group Health Plans, final rule, 84 FR 28888, June 20, 2019; 26 CFR 54.9802-4.
3. H.R. 6703, Lower Health Care Premiums for All Americans Act, passed the House December 17, 2025.
4. CSL Behring, Hemgenix (etranacogene dezaparvovec) US approval and pricing announcement, November 2022.
5. Vertex Pharmaceuticals, Casgevy US approval, December 8, 2023; bluebird bio, Lyfgenia US approval, December 8, 2023.
6. 45 CFR 147.102, fair health insurance premiums, applicable to the individual and small group markets.
7. Kaiser Family Foundation, 2025 Employer Health Benefits Survey, October 22, 2025.
8. HRA Council, 2026 data report on ICHRA and QSEHRA growth, released August 12, 2026.
9. Employee Benefit Research Institute and Morgan Health, employer survey on ICHRAs, chartbook released July 28, 2026.
10. Total Administrative Services Corporation, 2026 Broker ICHRA Survey, released September 10, 2026\. Sample size not disclosed in the release.
11. CMS, Final Guidance Regarding Age Curves and State Reporting, December 16, 2016.
12. Revenue Procedure 2025-25 (2026); Revenue Procedure 2026-26, July 21, 2026 (2027).
13. Kaiser Family Foundation, What We Know So Far About 2026 ACA Marketplace Enrollment, Premiums, and Deductibles, May 19, 2026.
14. Kaiser Family Foundation, How Has Insurer Participation in the ACA Marketplaces Changed in 2026?, June 11, 2026.
15. Centers for Medicare & Medicaid Services, 2026 Exchange Public Use Files (rate, plan attributes, service area) and state-based exchange public use files; my own county-level tabulation, published with the companion tool's data manifest. Counting each licensed issuer separately gives 144 single-insurer counties; rolled up to parent company, as Kaiser counts, 164.
16. U.S. Department of Agriculture, Economic Research Service, Rural-Urban Continuum Codes, 2023\. Metro is codes 1 to 3; remote rural is nonmetro counties not adjacent to a metro area (codes 5, 7 and 9).
17. Kaiser Family Foundation, Tracking Insurer Participation Changes in the ACA Marketplaces in 2027, as of September 15, 2026.
18. Kaiser Family Foundation, How Narrow or Broad Are ACA Marketplace Physician Networks?, August 26, 2024, using 2021 network data.
19. Robert Wood Johnson Foundation, Marketplace Pulse, January 22, 2025.
20. Randall, Duffy, Green et al., American Journal of Managed Care 28(9), September 2022.

Fringe Theory is independent and unaffiliated. Views expressed are my own and do not represent those of my employer. Nothing here is legal, tax, medical, or investment advice. Figures described as modeled or derived are my own arithmetic from the public sources listed, not disclosed data.