Evidence base
This section makes the current focused corpus browsable without weakening its evidence boundaries. The canonical analytical source remains the included-case dataset; the catalogue and map are deterministic views generated from it.
- Browse all included reports for the complete report catalogue and controlled facet counts.
- Open the case–mechanism/evidence map to compare non-exclusive sociotechnical codes across reports and primary workflow groups.
- Review the source register for provenance and reliability notes.
- Review the claims register for claim-level traceability.
- Review exclusions and rechecks before treating the corpus as closed.
What inclusion means
An included row records a PFD report containing a relevant electronic-record, digital-workflow, interoperability, information-flow or health-IT concern under the project's focused protocol. Inclusion does not mean that:
- the digital issue caused or contributed to the death;
- a named product has a general defect;
- the concern is prevalent across the NHS or Defence;
- a response was received or an action was implemented; or
- an em dash in the mechanism map is an evidence gap.
Each generated report page links to its registered primary Judiciary source and preserves the dataset's confidence, qualification and response-evidence status. Downstream Defence links are read only from the canonical REQ- register; they remain analytical translations rather than coronial recommendations.
The included-case CSV retains two renamed legacy_* translation columns from the initial research pass so earlier work is not destroyed. They are audit history, are not rendered as requirements and need not be populated for future reports. Controlled Defence statements live only in defence/requirements.csv.
Controlled browsing facets
Every report keeps all of its non-exclusive sociotechnical codes but receives one conservative primary classification in each of three dimensions:
| Facet | Purpose | Controlled values |
|---|---|---|
| Workflow group | The main safety-relevant work represented by the row | Diagnostic results and investigations; identity and population management; medication and prescribing; messaging, tasking and follow-up; record access and continuity; record integrity and provenance; referral, handover and transfer; risk, care planning and safeguarding; screening and longitudinal surveillance; triage, observations and decision support |
| Setting group | A broad care-context entry point | Acute and emergency care; cross-setting and multi-agency care; maternity and paediatric care; mental health care; primary and community care; secure and custodial care |
| Boundary type | The predominant information boundary represented by the row | Across health and non-health sectors; between care settings; hybrid paper and digital record; national-to-local information flow; patient-to-service channel; within a care setting |
These are navigation aids derived conservatively from existing row text. A single primary assignment inevitably compresses multi-setting and multi-mechanism events. The original clinical setting, care transition, system, function, failure, human-factors, workflow and information fields remain visible on every report page.
Evidence-status fields
verification_status combines the presence of a linked primary source with extraction confidence; it is not a score for source authority or independent review.
- High-confidence extraction: the full primary report was checked and the digital mechanism is explicit in the report-specific concern or circumstances, with causal qualifications retained.
- Qualified extraction: the primary report is linked, but the digital mechanism, remedy or causal relevance is indirect, contested or dependent on respondent context and therefore needs an explicit limitation.
All current assignments remain single-reviewer extractions pending the independent check in VAL-011. A primary report remains authoritative for what the coroner stated regardless of the extraction label.
response_status is intentionally narrow:
- Response evidence noted means the row's qualification refers to registered response/respondent evidence or records a recipient's post-report position. The generated report record links every known attachment used for that row.
- No official response located is used only where the existing row explicitly says so.
- Not assessed in this dataset makes no claim about whether a response exists.
No implementation-status field is inferred from a requested-action recipient or from the mere existence of a response.
Mechanism codes
| Code | Meaning |
|---|---|
LOOP |
Action ownership and closed loop |
GOV |
Governance and assurance |
VIS |
Information visibility |
HF |
Human factors and usability |
INT |
Interoperability and continuity |
ALERT |
Decision support and alerting |
PROV |
Data quality, provenance and auditability |
RES |
Resilience and degraded operation |
IDPOP |
Identity, registration and population management |
The codes are non-exclusive and can be used together. Their counts must not be summed as deaths, and they do not establish causal weight.
Regeneration
Run node scripts/generate-evidence-pages.mjs after changing the canonical included-case dataset. Run node scripts/generate-evidence-pages.mjs --check to confirm that the generated catalogue, map and individual report pages are current and that the row/page counts match.