Insurers call rosacea cosmetic. Your own record is what changes that.

The two-paragraph letter

The denial letter is shorter than the intake form that produced it. Somewhere in it sits a phrase like "considered cosmetic in nature and not a covered benefit," and that one line closes the file before any clinical detail gets read.

By the time that letter arrives, most people have been managing this for years. Burning that starts a few minutes before anything shows on the face. Eyes that feel like sand by mid-afternoon and blur enough to make evening driving unpleasant. Thickening around the nose that changed how breathing works. Evenings declined, because a warm room is a guaranteed flare.

None of that was in the file the reviewer opened. It was in the exam room, in passing, in the two minutes before the prescription got written. So it never became evidence.

That's the gap. Not the science. The paperwork.

The cosmetic default

Nobody at the plan sat down with the dermatology literature and ruled against it. What happens is duller and harder to fight: rosacea arrives at intake carrying a default, and the submitted record never displaces it.

Call it the cosmetic default. It's the assumption that a condition whose headline feature is facial redness is a complexion issue with a medical vocabulary attached. The default is applied first and tested only if something in the file makes testing necessary. Most rosacea files contain a diagnosis code, a prescription, and nothing else. Nothing else is exactly what confirms the default.

Which means the lever isn't persuasion. It's the record.

Arguing the science is the wrong appeal

The standard advice in rosacea forums is to come back with studies. Print the comorbidity papers, cite the cardiovascular associations, show that this is a real inflammatory condition and not vanity.

We understand the instinct and we think it mostly fails, for a boring reason. A medical-necessity reviewer is applying plan criteria to one member's chart. Population-level comorbidity evidence establishes that rosacea patients as a group carry elevated risk. It says nothing about whether this member has an impairment the plan is obligated to treat. The reviewer isn't being asked to grade the literature.

The second habit fails for a similar reason. People send a photo, because the photo feels self-evidently damning. One image of a flushed face, dated nowhere, taken under kitchen light, reads to a reviewer exactly like the thing they already assumed it was.

What moves a file is longitudinal, functional, and specific to one person.

What the clinical classification actually says

Rosacea's own field moved away from the framing insurers still use. The 2017 National Rosacea Society Expert Committee update (Gallo et al., Journal of the American Academy of Dermatology, 2018) and the global ROSCO consensus (Tan et al., 2017) replaced the old four-subtype model with phenotypes: individual features, assessed and treated on their own. Two of those features are difficult to call cosmetic with a straight face.

Phymatous change. Fibrous tissue hypertrophy, most often at the nose. It's progressive, it doesn't reverse with topical therapy once established, and at volume it narrows the nasal airway. Airway obstruction is a functional finding, not an appearance finding.

Ocular involvement. Blepharitis, meibomian gland dysfunction, chronic dryness, and in a minority of cases corneal complications that threaten vision. Ocular rosacea can precede any skin finding at all, and it's under-recognised because dermatology and ophthalmology sit in different notes. Prevalence estimates vary widely across studies (Meta-analysis of 11 studies and 124,093 individuals found diagnosed ocular rosacea prevalence of 10.3% among cutaneous rosacea patients ([source](https://www.canadianjournalofophthalmology.ca/article/S0008-4182(25)00156-5/fulltext))).

On comorbidity, Egeberg and colleagues' Danish nationwide cohort work established associations between rosacea and several systemic and neurological conditions. We're deliberately not re-litigating that evidence here; a separate piece on this desk covers the cardiovascular overlap in detail. For appeals purposes it's background, not the argument.

One more thing the classification implies. Erythema is the feature everyone photographs, and it is the feature hardest to see on brown and black skin. Adamson & Smith (JAMA Dermatology, 2018) flagged that image-based dermatology assessment is built on datasets that under-represent darker skin tones, and Daneshjou et al. (2022) assembled the Diverse Dermatology Images set in direct response. The practical version for a patient with Fitzpatrick IV to VI skin: a record built on visible redness alone will under-represent your disease. Sensory phenotype (burning, stinging, tightness, heat) and ocular symptoms carry the weight instead, and they have to be written down because they don't photograph.

What ninety days changes

Consider the shape of a file that works, in the abstract, since every plan's criteria differ.

A denial arrives in March. Instead of writing back that week, the patient starts logging: a photo in the same spot each morning under the same light, plus three lines of text on the days something happens. Eyes gritty from 2pm, lubricating drops four times. Woke at 3am from facial burning, second night this week. Skipped the client dinner, restaurant was too warm.

By June there are roughly ninety dated entries. The photo series shows papules appearing and clearing on a cycle, and it shows the nasal contour in April against the nasal contour in June. The text shows twenty-two nights of disrupted sleep and eleven declined work or social obligations, each with a date. The dermatologist reads the export in the appointment and writes a letter of medical necessity that cites specifics rather than adjectives.

That file is doing something the March file couldn't. It's demonstrating impairment in domains the plan already recognises as medical: sleep, occupational function, ocular symptoms, progressive tissue change.

One photo of a flushed face reads as a complexion. Ninety dated photos with symptom notes read as a disease course.

The five domains a record has to cover

If the cosmetic default is the problem, the counter is a record organised around function rather than appearance. These are the domains worth capturing, and the translation from what a reviewer assumes to what the file has to show.

We'd add two process notes. Keep dates on everything, including the things that didn't happen (the appointment you cancelled, the shift you covered from home). And ask your dermatologist which of these domains their letter of medical necessity will lean on, because the record should be built to feed that letter, not filed in parallel to it.

We're not giving legal or benefits advice, and we can't tell you what your plan will accept. Plans differ, state external-review rules differ, and a patient advocate or your clinic's billing team will know your specific criteria far better than any article does.

Why we built the log this way

Skinframe exists because that record is tedious to keep and nearly impossible to reconstruct from memory in March for a denial you received in February.

The design decisions follow from the same evidence. Features get logged individually rather than rolled into one severity number, because the 2017 phenotype model treats them individually and because a composite score collapses exactly the detail an appeal needs. Photos are timestamped and grouped into series, so the useful comparison (April against June) is one tap instead of a scroll through a camera roll. Sensory symptoms, ocular symptoms, sleep disruption and avoided activities are first-class entries rather than a free-text notes field, because those are the domains that read as medical. Everything stays on the device, which matters more than usual when the data is a health record you may hand to an insurer on your own terms.

We don't run skin analysis on the photos and we're not going to. The literature on image-based dermatology assessment and skin tone (Adamson & Smith 2018; Daneshjou et al. 2022) is reason enough. The photo is your evidence. It isn't our interpretation of you.

What we're watching

Two things would change this picture. The first is whether payer policy language starts tracking the phenotype model, which would make ocular and phymatous involvement legible on their own terms instead of as sub-notes under a redness diagnosis. The second is external-review data: state insurance departments publish independent-review outcomes, and in principle those records could show how dermatology denials fare on appeal, which would be a better signal than any anecdote. Nobody seems to have pulled that data apart by diagnosis yet, at least not in anything public, so for now the outcomes stay anecdotal even though the reporting infrastructure to fix that already exists.

In the meantime, the practical move is unglamorous. Start the record before you need it, and take it to your dermatologist. They're the one who writes the letter, and the letter is stronger when it can cite your dates instead of your adjectives.