Dermatology retired rosacea subtypes in 2017. Most tracker apps still ask which one you have.

The four-bucket question

Open any rosacea-tracker app available on iOS today and within three onboarding screens you will be asked which type of rosacea you have. The choices are usually four: erythematotelangiectatic (the flushing, persistent redness, visible vessels presentation), papulopustular (the bumps and pustules presentation), phymatous (the skin-thickening, nasal-rhinophyma presentation), and ocular (the eye-irritation presentation).

For a patient at the start of their disease, this question is unanswerable. Most newly-diagnosed rosacea sufferers do not know which of those four buckets they sit in, do not know whether they sit in only one, and do not know that the buckets were never mutually exclusive. For a patient who has had rosacea for years, the question is misleading: the literature now treats the four labels as a description of which features happen to be visible at a single moment, not as a stable subtype the patient belongs to. Rosacea features come and go, sometimes cluster, sometimes don't. Asking a patient to pick one identity for the disease misrepresents the disease.

This is not a minor UX nitpick. The four-subtype model was the formal NRS 2002 framework, and it was specifically replaced in 2017 because the field concluded it was clinically misleading. The current diagnostic standard is a phenotype framework. The trackers in the App Store have not caught up with that change, and the people downloading them are paying the cost in confused onboarding and data fields that do not match the disease they are trying to track.

What the 2017 update actually said

Two near-simultaneous expert-committee statements changed the standard.

The ROSCO panel (Tan et al., Br J Dermatol 2017) was a Delphi-method consensus among 17 dermatologists and 3 ophthalmologists from the global rosacea-research community. Their published recommendation was explicit: discontinue the discrete-subtype model in favor of a phenotype-based diagnostic framework. The ROSCO 2019 update (Schaller et al., Br J Dermatol 2020) reaffirmed that recommendation and extended it into management decisions.

The National Rosacea Society Expert Committee published its own statement the next year (Gallo et al., J Am Acad Dermatol 2018), reaching the same conclusion independently. The NRS had originally authored the 2002 four-subtype document; the 2017 statement was the same body retiring its own prior framework.

The phenotype framework defines two diagnostic phenotypes (persistent centrofacial erythema; phymatous changes), either of which alone is sufficient to diagnose. It lists four major phenotypes (transient flushing; telangiectasia; inflammatory papules and pustules; ocular manifestations), of which any two together are sufficient to diagnose in the absence of a diagnostic phenotype. It lists three secondary phenotypes (burning or stinging; edema; dry or scaly appearance) as supportive features.

The phenotypes are independent. A patient with persistent centrofacial erythema, two telangiectasias under the right cheek, occasional flushing after wine, and intermittent stinging across the central face has four features active at the same time. The 2002 model would have forced that patient into the erythematotelangiectatic bucket while quietly ignoring the burning sensation and the flushing pattern. The 2017 model represents them as four parallel features and tracks each one on its own timeline.

Why the field made the change

The clinical motivation was not theoretical. The four-subtype model produced known classification errors that the field had been documenting for a decade.

The first error was inter-subtype overlap. The published case-series literature consistently showed that a patient classified as erythematotelangiectatic this year would routinely develop papules next year and shift toward the papulopustular bucket without anything changing about the underlying disease. The buckets were not stable; the labels described a moment, not a person.

The second error was the assumption of mutual exclusivity. Rosacea patients frequently present with phenotypes from more than one of the 2002 subtypes simultaneously. The four-bucket model required clinicians to either pick one and ignore the others or create unwieldy hybrid labels that the literature never agreed on.

The third error was patient-facing comprehensibility. Patients reading their own chart notes could not tell whether a subtype label was a description of their current presentation, a prognosis, a stable diagnosis, or a treatment plan. The label was operating as a placeholder for several different clinical concepts that should have been kept separate.

The phenotype framework solves all three. Features are independent so they do not have to overlap or exclude each other. Each feature is tracked on its own scale rather than being collapsed into a global subtype. Patients reading the framework see seven concrete observations rather than four discrete buckets they need to choose between. The literature now expects clinicians and trackers to operate this way.

What the tracker category has built instead

We looked at every rosacea-named app currently listed on iOS in our category audit window (May 2026). Without naming them, here is the category pattern.

Four of the five apps with onboarding flows we could complete asked the user to choose between two and four subtypes, using the 2002 labels or visual approximations of them, on the second or third onboarding screen. None of them surfaced the phenotype framework. None of them allowed the user to skip the subtype question. Two of them gated downstream features on the subtype answer, so a user who picked papulopustular got a different in-app tracking interface than a user who picked erythematotelangiectatic for the same underlying disease.

The fifth app skipped the subtype question entirely and asked the user to rate a single composite severity score on a 0 to 10 slider. This is a different problem (a single global score collapses the phenotypes back into one number, the same complaint dermatology had with the IGA, see our companion piece on rosacea severity scoring) but it shared the underlying mistake of treating the disease as a one-dimensional fact about the patient.

None of the apps we audited matched the published consensus. This is not because the developers were careless; it is because the consensus change happened in the dermatology journals and never propagated into the consumer-app category. The clinical literature and the App Store are not the same information graph. The trackers that the patients use day to day are nine years behind their dermatologists.

What a phenotype-aware tracker looks like

Shipping the 2017 framework in a tracker is not difficult; it just requires throwing out the subtype-picker screen.

The correct shape is independent feature toggles. At onboarding, a patient indicates which features they are currently noticing: persistent centrofacial redness, transient flushing episodes, papules or pustules, visible vessels (telangiectasia), eye irritation, burning or stinging sensations. Each toggle has its own field in the tracker. None of them is required at onboarding; the patient can turn them on as features show up.

The daily-log flow then records each active feature on its own. Redness today (a per-feature severity, adapted from the validated Patient Self-Assessment scale: clear, almost clear, mild, moderate, severe). Papule count or rough estimate. Visible-vessel awareness. Eye comfort. Burning or stinging frequency. The dashboard surfaces each feature's trajectory separately, because that is what the phenotype framework predicts: features move on their own clocks.

There is no "which type" question because the framework no longer assigns the patient to a type. There is no global severity score because the literature explicitly identified the global-score collapse as the failure mode of the IGA. There is no need for the four-bucket vocabulary at all; the phenotype vocabulary is more accurate and patients learn it from the in-product copy.

This is what Skinframe ships. The schema is not a marketing choice; it is the dermatology literature's published answer to the question of how to represent rosacea, implemented in software.

Why this is the test that matters

If you are evaluating which rosacea tracker to use, the cleanest single test is this: does it ask you to pick a subtype in onboarding?

If yes, the app is running on the 2002 framework that dermatology retired in 2017. Whatever else the app does well, the data model is at least nine years out of date, and the trends it surfaces will be shaped by the wrong assumptions about how rosacea behaves. The tracker is asking you to commit to an identity (papulopustular, erythematotelangiectatic) that the literature has stopped using and that the disease itself does not respect.

If no, ask the next question. Does the app record features independently or does it collapse them into a single score? A single-score app has the second failure mode the literature flagged (the IGA collapse problem) and is hiding the phenotype dimensions even if it does not force the subtype question.

If the app records features independently, lets each feature have its own scale, and lets the patient turn features on as they show up over time, the app is on the right side of the 2017 update. That is the minimum bar for treating rosacea the way the field treats it now.

We built Skinframe around that minimum bar because the alternative is asking patients to enter data into a schema dermatology has formally moved past. The 2017 update was published nine years ago. The category has had time.