What v1.1.0 adds, annotated

Three participant-level metrics, cut down from a longer candidate list by @jwildfire: saf0001 Hy's Law candidate, saf0002 QTcF prolongation, saf0003 serious / related AE. Each scores one row per participant with an ordinal tier through the standard gsm contract — Analyze_Identity() + Flag(), following gsm.kri's pat0015 precedent — so subject rows land in Reporting_Results beside site and country rows and nothing downstream changes. The liver and QT cut-points come from the vendored data contracts of the charts they correspond to, so the metric and the chart cannot drift apart.

1 2 3

hub#138's Data Requirement set a binding gate: actual flagged-participant counts on demo data are reported before any metric is treated as real. The counts below are the release's own evidence — and two of them changed the design.

Measured on DEMO-301's mapped output, 765 enrolled participants. Three findings reading the code would not have surfaced: the QTcF metric is carried by the change criterion (exactly one participant exceeds 500 ms absolute; ten more reach red only through ≥ 60 ms change from baseline); only 400 of 765 participants have a post-baseline QTcF at all — a coverage finding, and why the overview leads with coverage; and the AE metric as first written flagged 560 of 765 participants, so its amber threshold moved up a tier — "serious or related" alone is a property of being in a trial, not a review signal.

The retuned queue. After the gate, saf0003 flags 319 of 765 — 251 amber, 68 red — a queue someone can actually work, with amber starting at a serious event and red at a serious, related event at or above the grade cut. The discontinuation leg has no source column in this demo's AE domain; Input_SafetyAE() records ActionColumnPresent = FALSE on every row rather than scoring as though nobody discontinued — the gap is stated in the output.

SafetyCensus() reduces the mapped domains to what a safety overview leads with: enrollment, randomisation, exposure, person-years, deaths — and data coverage per visit, the figure that decides whether a quiet visit is reassuring or empty.

The census as DEMO-301 renders it. Week 12 labs cover 17 of 765 participants (2%) — the display says so before anyone reads "no week-12 findings" as good news. This is the same coverage fact that explains saf0002's 400-of-765 scored count above.

The report workflows move from inst/workflow/3_reports/ to inst/workflow/4_modules/ — the standard gsm phase directory for meta.Type: Report workflows, the one gsm.kri already uses — and the new 2_metrics/ phase slots in where the ecosystem expects metric generation. 3_reports was the one phase directory in the portfolio that deviated, and it sorted beside gsm.reporting's unrelated 3_reporting.

inst/workflow/
├── 1_mappings/          # unchanged
├── 2_metrics/           # NEW — metric generation, ahead of module assembly
│   ├── saf0001.yaml     #   Hy's Law candidate
│   ├── saf0002.yaml     #   QTcF prolongation
│   └── saf0003.yaml     #   serious / related AE
└── 4_modules/           # was 3_reports/ — the gsm standard for Report workflows
    ├── ae_explorer.yaml … safety_shift_plot.yaml   # 9 modules, unchanged content

One breaking path. Any pipeline that globbed inst/workflow/3_reports/ updates its path — demo-301's did, in the same arc. Workflow content is unchanged.