The words on a denial letter

The Department of Managed Health Care assigns the case to reviewers with no stake in the plan. It costs the enrollee nothing. If the reviewer does not sustain the denial, the plan is required to provide the service — the review is binding on the plan, not advisory.

Every decision on this site is one of these. The Department has published them since 2001 because Health & Safety Code § 1374.33(g) requires it to, and § 1374.33(h)(1)(K) requires each one to carry a summary of the reasoning — which is the only reason a dataset like this can exist at all.

The word describes what happened to the denial, not a judgement about the plan, the enrollee or the treatment. A reviewer overturns when the record in front of them supports the service under the standard they are applying.

Drawn on this site in its own colour, at the same weight as an overturn, and never as the leftover space in a bar. 20,245 decisions since 2001 were upheld; a reader who represents plans is looking at their side of the record, and it is a fact rather than a remainder.

The reviewer applies the plan’s definition of medical necessity to the clinical record — documented severity, what has already been tried, whether the level of care matches the presentation. These are decided against the patient in front of the reviewer, which is why the same treatment can be overturned in one case and upheld in another without either decision being inconsistent.

The question shifts from the patient to the evidence base, and the reviewer weighs published literature, regulatory status and specialty-society guidance. It is the branch where a named authority — a compendium, a society guideline, an FDA label — most often carries the decision, which is why what reviewers cite is a page of its own. Some of these are decided by a panel of three reviewers, and a split panel publishes each opinion.

The Department decides whether a case qualifies. It is worth asking about: expedited cases resolve in a fraction of the standard time, at no extra cost, and the outcome split differs between the two tracks because the mix of cases differs. Ask the DMHC Help Center — it is free, and it is the authority on whether a specific case qualifies.

Independent review sits after this step, which is a large part of why the population this dataset describes is so selected. Deadlines and exceptions change; the DMHC Help Center will tell you where a specific case stands, at no charge, and will file the independent review for you.

Prior authorisation is its most familiar form. Nothing in this dataset describes utilization review itself: these are decisions about denials that were already issued and then contested to the end. The published record does not reliably state the plan’s reasoning, only the ground on which the reviewer resolved the dispute — a distinction the methodology takes seriously, because coding one as the other would be inventing the plan’s side of the argument.

Roughly half of Americans with employer coverage are in one, and the insurance company’s name on the card is usually just the administrator. These appeals go to a process run for the Department of Health and Human Services which publishes neither decisions nor outcome statistics.

This is the single largest limit on everything here, and it is not a limit of this dataset — it is a limit of what anyone publishes. Human resources or the plan documents will say which kind a plan is. See coverage.

External review is the end of a long funnel: fewer than 1% of denials are appealed internally at all, and only a fraction of those go further. Every figure on this site is conditional on a denial having survived that funnel.

That makes the population selected and it makes it informative. It does not make it a census, and a number from here should always be read with the condition attached — “among denials taken to independent review”, not “of denials”.

Below about thirty cases a percentage moves several points when a single decision changes, and it reads as precision it does not have. Where you see a dash instead of a rate, the class exists and the count is real — the explorer will still show it — and no rate is claimed. This is a rule applied before the numbers were seen, not a judgement about which ones looked unreliable afterwards.

Health & Safety Code § 1374.33(g) commands the removal of names before these decisions are published, so a per-plan comparison cannot be built from this source by anyone. Nothing on this site restores an identity, links a decision to any other record, or quotes a case narrative — the narratives stay in the build database and never reach the deployment. If you are looking for your own case here, it is not findable, and that is the intended design of the published file.

Every page here is stamped 2026-09-04. Showing our capture date instead would imply the data is current as of then; it is current as of whenever the Department last published, which is a different and sometimes much older date.