IFU coverage audit — integration-requirements verification
This page is a scratch-pad audit for the internal team. It is not shipped to BSI and is not part of the audit-visible technical file. The audit-visible artefacts produced from it are (i) the IFU delta landed in user-interface.mdx, (ii) the CEP cross-reference sentence landed in R-TF-015-001, and (iii) the CER symmetric update landed in R-TF-015-003.
Scope of the audit
The CEP integrator paragraph (R-TF-015-001, lines 514-516) mandates that every integrating system MUST implement four UI surfaces for the device to deliver the Pillar 3 clinical benefit validated in the pivotal investigations. The paragraph states those requirements are "specified in the IFU integration requirements" and "tracked as risk controls in R-TF-013-002", without naming an IFU section or specific risk-control entries.
This audit verifies, surface-by-surface, whether the IFU's user-interface.mdx (Installation Manual → User Interface — the ITP-facing volume) actually satisfies the cross-reference. Four verification columns are applied:
| Column | What it checks |
|---|---|
| Named | Is the surface named in the rendered IFU prose (not just imported as a JSX component)? |
| MUST | Is the presentation requirement stated with integrator-MUST language (not discretionary)? |
| Visibility / ordering / distinguishability | Are the properties on which Pillar-3 clinical-performance claims depend stated (ranked order, visual distinguishability, always-visible)? |
| RC traceability | Does the IFU cite the R-TF-013-002 entry or entries that cover the surface? |
Reference state of disk (2026-04-20, read-only sweep)
- IFU:
apps/eu-ifu-mdr/versioned_docs/version-1.1.0.0/installation-manual/user-interface.mdx - IFU (binary indicators enumeration):
apps/eu-ifu-mdr/versioned_docs/version-1.1.0.0/installation-manual/endpoint-specification.mdx, lines 97-108 - CEP:
apps/qms/docs/legit-health-plus-version-1-1-0-0/product-verification-and-validation/clinical/Evaluation/R-TF-015-001-Clinical-Evaluation-Plan.mdx, lines 514-516 - CER:
apps/qms/docs/legit-health-plus-version-1-1-0-0/product-verification-and-validation/clinical/Evaluation/R-TF-015-003-Clinical-Evaluation-Report.mdx, lines 474-487 - Risk file:
apps/qms/docs/legit-health-plus-version-1-1-0-0/risk-management/R-TF-013-002.json(62 entries, 3,592 lines)
Surface 1 — Top-5 prioritised differential view
Location in IFU: user-interface.mdx, sub-section "Top-5 Predictions", lines 181-211.
| Column | Current state | Gap |
|---|---|---|
| Named | YES — "Top-5 Predictions" heading and rendered <Top5 /> block. The prose names "the top-5 predictions". | None. |
| MUST | NO — "provide a quick overview of the most likely diagnoses, helping HCPs make informed decisions" is descriptive, not mandating. | Gap 1.1: the integrator-MUST language is absent. |
| Visibility / ordering / distinguishability | NO — the ranked-order requirement is not stated. The JSON snippet enumerates hypotheses but does not say the integrator MUST render them in descending-probability order. | Gap 1.2: ranked-order requirement missing. |
| RC traceability | NO — no citation of R-TF-013-002 in this sub-section. | Gap 1.3: no RC-ID cited. |
Existing RC entries that cover this surface (to be cited):
R-BDR— "Misinterpretation of data returned by the device". Mitigation wording: "We define the minimum user interface requirements to show the medical device outputs to HCPs" and "Minimum user interface requirements are provided in the IFU". This is the umbrella UI-requirement RC.R-A96— "Incompatibility in classification systems" (ICD-11 mismatch). Covers the Top-5 list because Top-5 items are ICD-11 categories.
Pillar-3 dependency: the Top-5 panel is the Pillar-3 clinical-performance surface. All Top-1 / Top-3 / Top-5 accuracy claims in R-TF-015-003 (and the pivotal investigations BI_2024, PH_2024, SAN_2024, MAN_2025, IDEI_2023) assume ranked-order display. If the integrator renders the candidates out of order, Pillar-3 accuracy is not what was validated.
Surface 2 — Malignancy-prioritisation gauge
Location in IFU: user-interface.mdx, sub-section "Malignancy", lines 286-323.
| Column | Current state | Gap |
|---|---|---|
| Named | PARTIAL — "Malignancy" heading names the surface, but the prose does not use the regulatory label "malignancy-prioritisation gauge" used in the CEP / CER integrator paragraphs. | Gap 2.1: regulatory label not surfaced. |
| MUST | NO — "Malignancy risk information aids dermatologists in assessing lesion risks" is descriptive. | Gap 2.2: integrator-MUST language absent. |
| Visibility / ordering / distinguishability | NO — no statement that the gauge MUST be visually distinguishable from decorative UI, and no statement that it MUST re-surface malignancy risk even when a malignant ICD-11 category is not top-ranked. | Gap 2.3: visual-distinguishability requirement missing. |
| RC traceability | NO — no citation of R-TF-013-002. | Gap 2.4: no RC-ID cited. |
Existing RC entries that cover this surface:
R-HBD— "Misrepresentation of magnitude returned by the device" (the gauge is a magnitude cue; mis-styling it as decorative text collapses the magnitude signal).R-BDR— umbrella UI-requirement RC (as above).R-DAG— Clinical-Performance / AI-ML category RC covering wrong-output-presented-to-HCP failure modes.
Surface 3 — Referral recommendation
Location in IFU: user-interface.mdx, sub-section "Referral", lines 247-284.
| Column | Current state | Gap |
|---|---|---|
| Named | YES — "Referral" heading and the rendered <Referral /> block. | None. |
| MUST | NO — "Referral information helps HCPs determine the urgency of a case" is descriptive. Thresholds are hedged as "suggested". | Gap 3.1: integrator-MUST language absent; the "suggested" framing weakens the claim. |
| Visibility / ordering / distinguishability | NO — no statement that the referral field MUST be surfaced alongside (not behind an extra click / tab / expand). | Gap 3.2: co-visibility-with-Top-5 requirement missing. |
| RC traceability | NO — no citation of R-TF-013-002. | Gap 3.3: no RC-ID cited. |
Existing RC entries that cover this surface:
R-BDR— umbrella UI-requirement RC.R-75H— Clinical-Performance / AI-ML RC covering wrong-output-to-HCP pathways.
Surface 4 — Six binary malignancy-surfacing safety indicators
Location in IFU: not present in user-interface.mdx. The six indicators are enumerated as an API-contract description in endpoint-specification.mdx, lines 97-108:
The six binary indicators are:
- Malignant …
- Pre-malignant …
- Associated with malignancy …
- Pigmented lesion …
- Urgent referral — within 48 hours …
- High-priority referral — within 2 weeks …
The enumeration and derivation are in place; what is absent is a UI-surface mandate that the integrator MUST display all six in the user interface.
| Column | Current state | Gap |
|---|---|---|
| Named | YES (in Endpoint Specification, not in User Interface). | Gap 4.1: no named block in the User Interface section; follow-up pointer is absent. |
| MUST | NO — the Endpoint Specification block describes the indicators; it does not mandate UI surfacing. | Gap 4.2: UI-surfacing MUST absent. |
| Visibility / ordering / distinguishability | NO — no statement that all six MUST be always-visible, not hidden behind expansion. | Gap 4.3: always-visible requirement missing. |
| RC traceability | NO — no citation of R-TF-013-002 for the UI-surfacing requirement. | Gap 4.4: no RC-ID cited. |
Existing RC entries that cover this surface:
R-BDR— umbrella UI-requirement RC.R-HBD— magnitude-misrepresentation RC (applies to the binary indicator states being correctly surfaced as always-visible cues).R-SKK— Clinical-Performance / AI-ML RC covering patient-visible erroneous-result pathways.
Summary — nothing missing, but nothing sharp enough
- All four surfaces exist in the IFU in some form. Three (Top-5, Malignancy, Referral) have dedicated UI sub-sections that render JSX previews of the presentation; one (the six binary indicators) is described as API contract in Endpoint Specification but not called out as a mandated UI surface.
- No surface carries integrator-MUST language in the User Interface section (the only MUST clauses in
user-interface.mdxare the GSPR 23.2 Label block at line 470). - No surface carries a visibility / ordering / distinguishability clause stating what the integrator must preserve for the Pillar-3 claims to hold.
- No surface is traced to a
R-TF-013-002entry. The CEP's "tracked as risk controls inR-TF-013-002" claim currently resolves only at the category level (Clinical Performance and AI/ML; System Integration and Interoperability) and not at the per-surface level.
These four gaps (one per surface, plus a fifth "traceability" gap cutting across all four) are what the IFU delta closes.
Risk-control-to-surface map (to be cited in the IFU delta)
| Surface | Primary RC IDs to cite |
|---|---|
| Top-5 prioritised differential view | R-BDR, R-A96 |
| Malignancy-prioritisation gauge | R-HBD, R-BDR, R-DAG |
| Referral recommendation | R-BDR, R-75H |
| Six binary malignancy-surfacing safety indicators | R-BDR, R-HBD, R-SKK |
R-BDR recurs on all four surfaces because it is the umbrella UI-requirement RC — it already states that the minimum user-interface requirements are defined in the IFU. The other RC IDs are surface-specific: R-A96 is ICD-11-specific (Top-5), R-HBD is magnitude-specific (Malignancy gauge, binary indicators), R-DAG / R-75H / R-SKK are AI-output / clinical-performance-category RCs.
No edits to R-TF-013-002.json are needed (user decision Q3); the IFU delta cites the IDs above. If a future BSI round surfaces a request for surface-specific risk-control entries (one per surface), those would be authored in a separate task.
Proceed-to-delta preconditions
All four surfaces have a clear home in user-interface.mdx (three existing sub-sections + one new compact sub-section). The delta is additive — no structural changes, no deletions, no rewrites. This audit justifies proceeding to ifu-delta.md.