Pharmacovigilance follow-up packet to safety case record handoff¶

Canonical pattern(s): Document to structured data handoff Source Markdown: instances/compliance/pharmacovigilance-follow-up-packet-to-safety-case-record-handoff.md

Linked pattern(s)¶

document-to-structured-data-handoff

Domain¶

Compliance.

Scenario summary¶

A drug safety operations team receives a follow-up packet for a serious adverse event involving a hospital patient who was re-admitted after using an infusion therapy. The source packet includes a faxed MedWatch form, an email from the treating physician, scanned discharge notes, product lot photographs, and an internal follow-up questionnaire completed by the field safety specialist. Before any regulator gateway submission can be prepared, the intake workflow must transform the packet into a structured safety case record with the required ICSR staging fields for reporter type, suspect product, event terms, onset dates, seriousness criteria, patient attributes, lot references, and expectedness flags while preserving provenance, uncertainty, and unresolved contradictions.

flowchart TD A["Receive follow-up packet MedWatch fax, physician email, discharge notes, lot photos, and follow-up questionnaire"] B["Run OCR and document parsing capture source spans for consequential safety facts"] C["Populate staged ICSR fields reporter, product, event terms, onset, seriousness, patient, lot, and expectedness"] D{"Any minimum-case, channel, or contradiction issue detected or would completion require unsupported inference?"} E["Place case on safety exception hold for medical review, coding review, or data-entry correction"] F{"Do schema-required fields, provenance, and uncertainty markers support a reviewable handoff package?"} G["Create structured safety case record with transformation trace and reference-data versions"] H["Stop at staged case handoff submission readiness stays with downstream human reviewers"] A --> B --> C --> D D --> E E --> B D --> F F --> E F --> G --> H

Target systems / source systems¶

Pharmacovigilance case-management or ICSR staging system with a versioned intake schema for required safety fields
Safety intake mailbox and document repository holding fax images, emails, scanned clinical notes, and follow-up forms
OCR and document-parsing service tuned for handwritten medical forms and semi-structured hospital documents
Product master, MedDRA coding reference, lot-release records, and approved country or seriousness taxonomies
Safety exception queue for medical review, coding review, and data-entry correction before any external transmission workflow begins

Why this instance matters¶

This grounds the transform pattern in a regulated compliance workflow where the key handoff is not a filing action but the creation of a trustworthy structured case record that downstream safety reviewers and submission tooling can rely on. A schema-valid record is not sufficient if it obscures whether an event term came from a physician note, a handwritten form, or an inferred normalization step. The instance shows why explicit field contracts, provenance, and disciplined exception routing are necessary before a pharmacovigilance team can decide whether the case is complete enough for coding, assessment, and later reporting.

Likely architecture choices¶

flowchart LR subgraph intake["Safety intake boundary"] repo["Safety intake mailbox and document repository"] ocr["OCR and document-parsing service"] agent["Tool-using intake agent"] end refs["Product master, MedDRA coding reference, lot-release records, and approved taxonomies"] subgraph staging["Structured case-record boundary"] icsr["Pharmacovigilance case-management or ICSR staging system versioned intake schema and intake contract for direct, normalized, and pending fields"] end subgraph review["Human review boundary"] queue["Safety exception queue"] human["Medical review, coding review, and data-entry correction"] end repo -->|"Follow-up packet documents"| ocr ocr -->|"Parsed text and source spans"| agent refs -->|"Approved normalization and coding references"| agent agent -->|"Populate staged safety case, provenance, uncertainty markers, and transformation trace"| icsr agent -->|"Contradictions, missing minimum-case criteria, or unsupported inference"| queue queue -->|"Route review work"| human human -->|"Corrections and confirmations"| icsr

A tool-using single agent can coordinate OCR, source segmentation, field extraction, MedDRA candidate lookup, schema validation, and packaging of the staged safety case plus transformation trace.
The target handoff should use an explicit intake contract that distinguishes directly extracted values, normalized values, and fields still awaiting human coding or confirmation.
Approved reference data may normalize product names, lot formats, reporter categories, and seriousness criteria, but the workflow should refuse unsupported inference for missing onset dates, patient age, or causality indicators.
Human review remains mandatory for conflicting seriousness evidence, ambiguous event coding, missing minimum case criteria, or document sets that imply follow-up is incomplete.

Governance notes¶

Every consequential case field should carry a source pointer or excerpt reference, including the exact document and span supporting event narrative, suspect product, reporter identity, onset timing, and seriousness classification.
The workflow should route exceptions instead of handing off when the packet fails minimum pharmacovigilance intake criteria, when two documents disagree on core facts such as event date or product lot, or when protected health information is present outside approved intake channels.
Provenance should include the schema version, coding reference versions, normalization actions, and any lossy mapping from free-text clinical language into controlled safety fields.
Human medical review, not the transformation workflow, should decide final event coding, seriousness adjudication, and readiness for regulator submission; the handoff stops at a reviewable staged case record.

Evaluation considerations¶

Percentage of staged safety cases accepted by downstream pharmacovigilance review without manual reconstruction of source facts
Rate of incomplete or contradictory follow-up packets correctly diverted to exception review before coding or submission preparation
Completeness of field-level provenance for serious-event, product, and patient-attribute fields during audit or case reconciliation
Stability of the handoff when fax quality drops, clinical documents arrive out of order, or the intake schema adds a new mandatory field