Aesthetic Risk Score™ Specification

Notice

This document specifies the Aesthetic Risk Score (ARS): a structured, pre-treatment risk-assessment rubric for elective aesthetic procedures.

It publishes the scale, the polarity, the band thresholds, the assessment domains and their relative weightings, the register of clinical guidelines the rubric draws on, and the conformance obligations of an implementation. That is enough for a practice, a reviewer, a carrier, or a researcher to understand what the score measures, how much each domain contributes, and what an implementation must do to claim conformance.

The ARS is a pre-screening structure for a clinical assessment performed by a licensed practitioner. It does not diagnose, authorize, deny, or determine who may be treated, and it never substitutes for the Good Faith Examination.

The ARS has not been prospectively validated against clinical outcomes. See section 11.

Key words MUST, MUST NOT, SHOULD, and MAY are used per RFC 2119.

1. Scope

1.1 In scope

Pre-treatment candidacy assessment for elective cosmetic aesthetic procedures: injectables (neurotoxins, dermal fillers, biostimulators), energy-based treatments (laser, IPL, RF, microneedling), and chemical resurfacing.

1.2 Out of scope

The ARS is not specified for, and MUST NOT be applied to, the assessment of:

• disease diagnosis, staging, or treatment selection;
• prescription therapeutics, including GLP-1 receptor agonists, peptide therapies, intravenous therapy, and supplement protocols;
• any use in which the score is presented to a patient as an authorization, approval, denial, or clearance to receive treatment.

The cosmetic scoping in 1.1 is deliberate and load-bearing. An implementer that extends the ARS to therapeutic indications is not implementing this specification.

1.3 Intended user

The ARS is intended for licensed healthcare practitioners. Section 2 sets out what an implementation must disclose to that practitioner, and what this document discloses publicly. The two are deliberately different, and the distinction is the subject of section 2.

2. Two disclosure obligations

The ARS carries two separate transparency duties. Conflating them is the most common error in reading this specification.

2.1 Point-of-care disclosure: complete, and mandatory

At the point of use, a conforming implementation MUST present to the practitioner, for the individual score in front of them:

1. every contributing factor, named in clinical language;
2. each factor's direction: whether it raises or lowers risk;
3. each factor's magnitude tier (primary, contributing, or minor) as an ordinal classification, not a point value;
4. each factor's governing guideline, per the register in section 7, linked to its source;
5. the confidence interval and the data-quality basis for it (section 9);
6. the clinical inputs the factor was derived from, including the laboratory value or intake response that triggered it.

This obligation is absolute. An implementation that shows a practitioner a score without the factors, their direction, their tier, and their sources is non-conforming. The practitioner must be able to review the basis for the score, reach their own clinical conclusion from the same inputs, and disagree with the score on the specifics. That property is what makes the ARS decision support rather than a determination, and it is the condition on which 1.3 rests.

2.2 Public disclosure: structural, not numeric

This document discloses the scale, polarity, bands, domains, relative domain weightings, clinical guideline register, decay principles, governance, and conformance obligations.

It does not disclose individual factor point values, the baseline anchor, the internal calibration parameters, or the segmentation logic. Those are proprietary to Aesthetic Health Standards and are maintained as trade secrets.

2.2.1 Enforcement

Withholding the calibration is a systems obligation, not a presentation one. A conforming implementation MUST:

1. Never transmit point values to any client. Weights MUST NOT appear in an API response, a database row readable by a client role, a redirect payload, an export, a webhook, or a log accessible outside the scoring service, irrespective of whether any interface renders them. Redaction MUST occur server-side, before transmission. A client that receives the values and declines to draw them has already disclosed them.
2. Resist differential enumeration. Where a caller can vary an input and observe the resulting score, the weights can be recovered by differencing regardless of what is displayed. Implementations MUST constrain this: exact scores are disclosed only to roles with a clinical need, rescoring is rate-limited and audited, and anomalous re-scoring patterns are monitored.
3. Disclose bands, not integers, to third parties. Any surface serving someone other than the patient-of-record or their treating practitioners MUST receive the band (section 4) rather than the integer score. The patient MAY be shown their own integer, following the established practice of consumer credit scoring, provided the enumeration bounds in item 2 are enforced and no per-factor values accompany it (section 8). The exposure to manage is not the number itself but an unbounded oracle beside an editable intake.

An implementation that satisfies 2.2 in its interface but transmits the calibration to the client, or permits unbounded rescoring against varied inputs, is non-conforming.

2.3 Controlled disclosure

Trade-secret protection depends on maintaining reasonable measures of secrecy. Uncontrolled publication destroys it; narrow disclosure under binding confidentiality preserves it. The categories below are therefore deliberately limited, and each is conditioned on a written confidentiality undertaking, a defined purpose, and no right to retain, reproduce, or derive from the material.

Aesthetic Health Standards will disclose the complete rubric, on those terms only, to:

• the ARS Clinical Advisory Board (section 12);
• investigators and biostatisticians conducting the validation program (section 11);
• regulators and accreditation bodies exercising a supervisory function;
• a malpractice carrier conducting underwriting diligence, where the carrier is contractually bound and the disclosure is necessary to that diligence;
• a court, tribunal, or licensing board, and the parties to a proceeding before it, under protective order, where an ARS score is materially at issue.

The final category is not discretionary and cannot be contracted away. Where a score is challenged, the practitioner who relied on it must be able to produce its complete derivation. A trade secret that cannot be examined in the proceeding it affects is not defensible, and compelled disclosure under a protective order does not forfeit secrecy.

Disclosure is not available to implementers, integration partners, resellers, prospective licensees, or any party seeking it to evaluate, replicate, benchmark, or certify against the rubric. Conformance is established under section 13, which does not require access to the calibration.

3. The scale

A conforming implementation MUST clamp the final score to 300-850. It MUST NOT emit a value outside that range, and MUST NOT clamp intermediate sums; clamping is applied once, after all factors are summed.

3.1 The baseline is not the best case

The rubric's baseline sits inside band 2 (section 4). This is deliberate. A patient does not begin as a good candidate; they become one as the record establishes it.

An assessment built only from patient-supplied intake, however unremarkable, is an unexamined patient. The single largest protective credit in the rubric is earned when a licensed provider has performed and signed a Good Faith Examination. A healthy adult with no risk factors at all scores in band 2 on intake alone, and reaches band 1 once that examination is on file.

That is the intended behavior. The rubric does not reward the absence of findings; it rewards the presence of an assessment. Implementers MUST NOT re-tune the baseline or the band 1 threshold to make band 1 reachable from intake alone. That is a substantive change to what the label asserts, not a calibration choice.

3.2 Attainable range

The defined range is 300-850, but not every point in it is reachable at v2.1. The lowest attainable score is 395; scores of 300-394 are not attainable. That interval is reserved as headroom for protective factors a future version may introduce, subject to section 11.

The upper bound saturates readily by comparison, which is why the clamp in section 3 is normative rather than theoretical.

Implementations MUST NOT present the score as a percentile, a population ranking, or a uniformly distributed measure. It is an ordinal risk index over an interval that is currently occupied unevenly.

4. Bands

Four bands, at fixed thresholds. Boundaries are half-open: a score equal to a threshold falls in the higher band.

Band assignment is normative:

band(s) = 1 if s < 450
          2 if s < 580
          3 if s < 700
          4 otherwise

A conforming implementation MUST use exactly one band table. It MUST NOT maintain a second, divergent set of thresholds for any surface, simulator, or export.

Both label columns are normative for their audience. An implementation MUST NOT render "Approved", "Conditional", "Denied", "Eligible", "Ineligible", "Cleared", or any synonym as the band label on any surface. Those words describe an authorization decision the ARS does not make.

5. Assessment domains

Every factor in the rubric belongs to exactly one domain. The percentages give each domain's share of the total risk the rubric can express: its maximum attainable contribution, with mutually exclusive alternatives within a domain counted once.

Shares describe the rubric's capacity, not any individual patient. The domains that determine a particular score vary substantially between patients, and for most patients only a minority of domains contribute at all.

5.1 Two classes of factor

Clinical risk factors derive from a published clinical guideline. Each MUST carry a guideline code resolving to section 7, and that citation MUST be displayed alongside the factor wherever the factor is shown.

Profile factors do not derive from an external guideline. They reflect the completeness and recency of the record itself: an examination on file, a documented age band, prior treatment history. They MUST be labeled as originating from this specification and MUST NOT be presented as guideline-derived.

The distinction is not cosmetic. The review property in 2.1 depends on a practitioner being able to tell which parts of a score rest on published clinical evidence and which rest on this rubric's own judgment.

5.2 Clinical thresholds

The clinical cut-points at which factors change tier are taken from the published guidelines in section 7 and are not proprietary:

The points assigned at each tier are proprietary per 2.2. The direction is not: within every ladder above, worse control or greater deviation contributes more risk, without exception.

5.3 Laboratory supersession

Where a verified laboratory value is on file and within the 180-day scoring window, the implementation MUST apply the laboratory-derived factor and MUST NOT additionally apply the self-reported factor for the same condition. A verified result supersedes a patient's recollection; it does not stack with it.

A verified in-range result reduces, but does not eliminate, a reported condition's contribution. The residual reflects the risk of the diagnosis, not of the result. A patient without the underlying condition contributes nothing from that ladder.

6. Temporal decay

Risk that has passed MUST stop counting. Factors carrying a recency component decay from the recorded event date; where no event date is recorded, the undecayed weight applies.

Four curve families are used: linear to zero at a horizon, step to zero at a threshold, exponential by half-life with a floor, and regulatory washout, which is linear decay across a clearance period fixed by the governing authority rather than by this rubric.

Washout and holding periods taken from published guidance are themselves public and are normative:

Half-lives and horizons for smoking cessation, adverse-reaction history, ultraviolet exposure, and dental work are calibration parameters and are proprietary per 2.2. Their direction and family are not: each decays monotonically toward zero, and none can increase with elapsed time.

7. Guideline register

Every clinical factor resolves to one entry here. Codes are stable across versions.

Citations name the source that establishes the clinical relationship between a factor and procedural risk. They do not establish the weight assigned to that factor. Weights are this specification's own judgment and are the subject of the validation program in section 11. Conflating the two would misrepresent the evidence base, and implementations MUST NOT present a weight as guideline-mandated.

8. Patient-facing constraints

Where an implementation surfaces any part of the ARS to a patient, all of the following are normative:

1. No directive clinical text. Dosing, medication, discontinuation, and scheduling instructions MUST NOT be rendered to a patient. Clinical rationale MUST be mapped to non-directive review topics that name what the provider will review, never what the patient should do.
2. Fail closed. Any clinical string without an approved patient-facing mapping MUST render as a generic provider-review topic. It MUST NOT pass through verbatim.
3. Adjacent disclosure. Every surface making a candidacy claim MUST carry an inline, persistent caveat on the same screen. A once-acknowledged modal does not qualify a badge the patient sees three months later.
4. No improvement gradient. An implementation MUST NOT present the patient with point-valued actions that lower the score, and MUST NOT attach score movement to a purchasable treatment. Score-improvement tooling is practitioner-only.
5. No authorization vocabulary. Per section 4.
6. Own score only, oracle bounded. The patient MAY be shown their own score, integer or band, following the established practice of consumer credit scoring, which discloses the exact score to the consumer while withholding the scorecard. What turns a visible score into a score-optimization instrument is not the number but an unbounded recompute loop beside an editable intake, so this permission is conditional on the enumeration bounds of 2.2.1: rescoring rate-limited and audited, and disclosure of score movement decoupled from any single edit. Surfaces serving anyone other than the patient-of-record and their treating practitioners receive the band only.
7. No numeric factor weights, and none transmitted. Patient surfaces MUST NOT render per-factor point values, and the patient client MUST NOT receive them. See 2.2.1.

9. Confidence

A score is only as good as the record it reads. A conforming implementation MUST compute a data-quality measure on 0-100 and MUST report a confidence interval derived from it.

Data quality rises with: verified laboratory values on file; objective skin assessment; completed appearance-concern screening; recorded dates for time-decaying events; and completeness of core intake, meaning date of birth, height, weight, Fitzpatrick type, medication list, and health-conditions list.

Four confidence tiers are defined (low, medium, high, and very high), with interval width decreasing monotonically as data quality rises. Bounds are clamped to the defined range. The specific point allocations and interval widths are calibration parameters and are proprietary per 2.2.

Implementations SHOULD display the interval alongside the score. A given score drawn from a sparse record is not the same claim as the same score drawn from a complete one, and presenting the two identically overstates both.

Contributions from instrumented skin assessment MUST be scaled by a staleness factor derived from the assessment's age, and that age MUST be recorded as an absolute date.

10. Model and specification versioning

MAJOR.MINOR. A change to any weight, threshold, band boundary, decay parameter, or domain share is at minimum a MINOR increment. Changes to the scale, polarity, or band count are MAJOR.

Every score MUST be persisted with the specification version that produced it, so that historical scores remain interpretable after a revision. An implementation that stamps scores with a version other than the one that computed them is non-conforming.

11. Validation status

The ARS has not been prospectively validated against clinical outcomes. Weights are expert-derived from the guidelines in section 7. No correlation with adverse-event rates has been established or published.

Implementations, marketing, and derived materials MUST NOT state or imply that the ARS is clinically validated, outcome-predictive, or evidence-graded beyond the guideline citations in section 7.

Serious adverse events such as vascular occlusion and blindness are too rare to power a study at this scale, which is why the primary endpoint is a composite. The protocol will be published before enrollment opens.

Investigators receive the complete rubric under 2.3. A validation study of a model its own investigators cannot see would not be publishable, and Aesthetic Health Standards will not assert trade-secret protection against the program that tests it.

12. Governance

This specification is published by Aesthetic Health Standards. Amendments are made by the ARS Clinical Advisory Board.

Board composition (planned): dermatologists, plastic surgeons, core aesthetic physicians, an advanced-practice clinician, an ethicist, and a patient advocate, with published authority over the rubric and full access to it under 2.3.

12.1 Change log

13. Conformance

An implementation conforms to ARS v2.1 if and only if it:

• 1produces an integer in 300-850, clamped once, with lower meaning lower risk (section 3);
• 2uses the band table in section 4, and only that table;
• 3discloses to the practitioner, for every score, the complete factor list with direction, magnitude tier, triggering input, and guideline citation (2.1);
• 3atransmits no point value, anchor, or derivable figure to any client, and constrains differential enumeration (2.2.1);
• 4distinguishes clinical factors from profile factors, and does not present a profile factor as guideline-derived (5.1);
• 5applies laboratory supersession rather than stacking (5.3);
• 6applies monotonic decay, and the regulatory washout periods in section 6 exactly;
• 7computes and reports a data-quality-derived confidence interval (section 9);
• 8persists the specification version that computed each score (section 10);
• 9observes every patient-facing constraint in section 8;
• 10does not gate, authorize, deny, or auto-approve treatment on the basis of the score, and does not represent it as doing so (sections 1 and 4);
• 11makes no validation claim beyond section 11.

A conforming implementation MAY require practitioner acknowledgement at a score threshold. That is a practice-level workflow decision, not a determination by the score, provided the acknowledgement is recorded and the practitioner remains free to proceed.

Every clause above is verifiable from an implementation's own behavior and interfaces without access to the calibration. Conformance is therefore self-assessable, and access to the rubric is neither required nor available for that purpose (2.3).

14. Intellectual property

The structure described in this document, meaning the scale, polarity, band thresholds, domain taxonomy, guideline register, disclosure obligations, and conformance requirements, is published for open implementation. Aesthetic Health Standards asserts no claim over an implementation that adopts this structure and derives its own weights.

The calibration, meaning individual factor weights, the baseline anchor, decay parameters other than the regulatory washout periods in section 6, data-quality allocations, and confidence-interval widths, is proprietary to Aesthetic Health Standards and is maintained as a trade secret. It is disclosed under the conditions in 2.3.

Aesthetic Risk Score™ and ARS™ are trademarks of Aesthetic Health Standards. An implementation MAY state that it implements the Aesthetic Risk Score specification. It MUST NOT use the marks in a manner implying certification, endorsement, or approval by Aesthetic Health Standards. A conformance-certification program, if established, will be published here.