Pull up a standard stop-loss disclosure form and count the clinical fields.
HM Insurance Group’s form gives you a claimant name, a date of birth, one combined free-text cell headed “Diagnosis/Prognosis/High-Cost Drug”, a beginning date of treatment, paid claims for the last twelve months, and pended claims. Skyward’s form asks for an ICD-10 code drawn from an appendix, a hospital confinement checkbox, a transplant checkbox, the most recent date of service, and expenses incurred this plan year.
That is the clinical record on which a laser gets set.
What the code cannot tell you
Take ICD-10 code Z94.0, kidney transplant status. Three claimants can carry it.
The first received a transplant four years ago, is stable on maintenance immunosuppression, and represents a predictable annual pharmacy cost with a modest tail. The second was transplanted eleven weeks ago and sits in the window where acute rejection, CMV infection, and surgical complication all remain live. The third is listed, on dialysis three times a week, and will generate both the dialysis spend and the transplant episode inside your policy period.
Same code. The risk difference across those three runs into six figures, and on the wrong one, seven.
Spinal muscular atrophy shows the same problem with sharper edges. A claimant already dosed with Zolgensma has consumed a one-time cost that will not recur. A claimant diagnosed but not yet dosed carries $2.1 million or more, and the timing sits with the treatment team rather than with you. The ICD-10 code is identical.
Congenital anomalies behave differently again. Voya’s analysis of 2024 claims put congenital anomalies at the top of its ten largest individual claims, averaging $8.87 million. Sun Life reported congenital anomaly claims averaging $335,000 and rising 70 percent since 2021, with a single claim reaching roughly $12 million. NICU risk is front-loaded and terminates. An open-ended code implies otherwise.
The employer is being asked to predict
HM defines a shock loss claim as a loss the plan “reasonably assumes will result in a significant medical expense in the next 24 months”.
Consider who is answering that question. A benefits manager at a 600-life employer, working from a claims feed supplied by the TPA, is being asked for a two-year clinical forecast on a person she has never met and whose records she cannot lawfully read. She signs an attestation that all known potential large claimants have been disclosed.
She is doing her best. She is also the least qualified person in the transaction to make that call, and the form gives her no way to say so.
The industry already knows what it needs
Look at what carriers demand after binding rather than before it.
Great American’s 50 percent notice form, triggered when a claim reaches half the specific deductible, asks for the primary diagnosis with ICD-10 code, claims paid, pending, and denied, the specific deductible, total self-funded claims paid, the nurse case manager’s contact details, and case management notes including prognosis and estimated additional cost.
Prognosis and estimated additional cost. The industry has defined the artifact it wants. It requires it once the money is already at risk and hopes for it while the risk is still priced.
The pricing environment has removed the margin for error
Mercer put average January 2026 stop-loss renewal increases at 23 percent, up from 18 percent the prior year, with healthy groups around 15 percent. Segal put the figure near 13 percent and observed that claims which once qualified as rare shock losses now recur.
The frequency data supports that. Sun Life’s 2026 report, drawn from more than 70,000 high-dollar claims across 3,300 self-funded employers, found million-dollar claims up 46 percent in frequency between 2022 and 2026. Voya measured claim frequency at 32.5 per 100,000 employees in 2024, against 23.8 previously. Aegis Risk found 49 percent of employers had a claimant exceed $1,000,000 in the past two policy years, and a quarter of those exceeded $1,500,000.
Industry loss ratios sit near 85 percent against a 75 percent target. Mercer reports that carriers have become more selective about when they will quote at all, and that decline-to-quote rates have risen substantially.
Underwriters are being asked to price harder risk with the same information they had when the risk was easier.
What better disclosure would look like
The gap is not that underwriters lack judgment. It is that judgment has nothing to work with beyond a code and a number.
For an oncology claimant, the questions that move a laser are stage and grade, biomarker status, which line of therapy the patient is on, response to the current line, whether intent is curative or palliative, and whether the patient is a candidate for transplant or CAR-T. None of that appears on a disclosure form. All of it appears in the medical record.
For a transplant claimant, the single question is where in the sequence the patient sits. Listed, transplanted and inside the first year, or transplanted and past it.
For a hemophilia or ESRD claimant, the question is regimen and consumption rate, because these are annuity risks that recur every policy year rather than events that resolve.
Where the plan has authorization to review records, a clinical abstraction on the three or four claimants who actually drive the price is not a large piece of work. It is a few hundred pages read by someone who knows what a laser decision turns on, delivered as a structured summary rather than a narrative.
The objection worth taking seriously
Most submissions never bind. Brokers market a case to eight or fifteen markets, and a single carrier’s bind rate per quote issued sits in the low single digits. Paying for clinical review on every submission would mean paying for work that gets discarded most of the time.
That objection is correct, and it argues for selectivity rather than for the status quo. The review is worth doing on the submissions where a laser decision is live, and the exposure is large, which is a small subset. It is also worth doing at renewal on your own bound book, where the information belongs to you, the claimant is already known, and the decision recurs every year.
Underwriters who have run this exercise on their own renewals tend to find at least one claimant whose risk profile was materially different from what the code implied. Sometimes the difference runs in the carrier’s favor.
The uncomfortable version
A laser is a binary risk transfer decision worth between $250,000 and $3 million on a named human being. The industry currently makes it on a diagnosis code, a date, a dollar figure, and two checkboxes.
Every underwriter reading this already knew that. The question is whether it stays acceptable at an 85 percent loss ratio.CUBEXLE prepares clinical abstractions of large-claimant records for stop-loss underwriters and claims teams, structured around the decisions the file has to support. Every finding cites its source page.