If you work anywhere near drug safety, you have heard the term ICSR. It sits at the centre of every pharmacovigilance operation, yet the end-to-end workflow behind it is rarely explained in one place. This guide does exactly that: what an ICSR is, the steps a case moves through from first report to regulatory submission, and how modern, AI-assisted platforms are changing the economics of the work without removing human judgement.
What is an ICSR?
ICSR stands for Individual Case Safety Report. It is the structured record of a single adverse event involving one patient and one or more medicinal products. Each ICSR captures four minimum criteria for a valid case: an identifiable patient, an identifiable reporter, a suspect product, and an adverse event or reaction. Without all four, a report is generally not a valid ICSR and cannot be processed as one.
Regulators worldwide — the FDA, EMA, MHRA and others — require marketing authorisation holders and their partners to collect, evaluate and report these cases within strict timelines. The data format for electronic exchange is the ICH E2B standard, and case processing is the operational engine that turns a raw report into a validated, codeable, submittable case.
The ICSR case processing workflow, step by step
Although every organisation has its own SOPs, the core lifecycle is consistent across the industry. A case flows through the following stages, each with its own quality gate.
1. Intake and case receipt
Cases arrive from many channels: spontaneous reports by email, patient or HCP web forms, call centres, literature, clinical trials, partners, and increasingly social and messaging channels. The first job is to capture the source document, log the receipt date (which starts the regulatory clock), and confirm the four validity criteria. Modern platforms such as PVgenix AI Case Intake consolidate every channel into one queue and use AI to extract case data from unstructured PDFs, emails and narratives.
2. Triage and prioritisation
Not every case carries the same urgency. Triage flags seriousness, deaths, and life-threatening events so that the most time-critical reports are worked first. Risk-based triage and queue prioritisation ensure expedited cases never wait behind routine ones.
3. Data entry and medical coding
The verified information is entered into structured fields, and the clinical terms are coded — adverse events to MedDRA and products to the WHO Drug Dictionary. Auto-coding speeds this up, but a coder confirms the result. (Note that MedDRA and WHO-DD are licensed separately by the organisation that uses them.)
4. Narrative writing
The narrative is the human-readable story of the case: patient history, the event, treatment, and outcome. AI-assisted narrative generation can draft this from the coded fields, but a qualified reviewer edits and approves it before the case proceeds.
5. Case assessment
- Seriousness and severity against ICH criteria
- Causality — using the WHO-UMC system and/or a company algorithm, capturing company and reporter causality separately
- Expectedness / listedness against the reference safety information (CCDS, SmPC, USPI) to determine whether the case is expedited
These are regulated judgements. Software can surface the data and apply configured rules, but the determination is made and signed off by qualified pharmacovigilance personnel.
6. Quality review and medical review
A quality reviewer checks the case for accuracy and completeness; a medical reviewer confirms the clinical assessment. Duplicate detection runs throughout to prevent the same event being reported twice, and follow-up information is tracked and versioned as it arrives.
7. Regulatory submission
Finally, the case is converted to the required format — E2B(R3) or E2B(R2) XML, or a CIOMS I / MedWatch 3500A form — and submitted to the relevant authorities within the deadline (typically 7 or 15 days for expedited cases). A regulatory clock and rules engine determine which authority receives which case, in which format, by when. Acknowledgements are then reconciled against what was sent.
Where AI and automation fit
Case processing has historically been labour-intensive, and case volumes keep rising. AI now accelerates the repetitive parts: extracting data from unstructured sources, suggesting codes, drafting narratives, and scoring its own confidence so that low-confidence items are routed for human review. For qualifying, low-risk case types, configurable straight-through processing can move a case through with mandatory review gates still in place.
Human-in-the-loop by design
AI assists intake and data entry. Qualified personnel review, edit, and approve AI-generated content before a case proceeds; regulated judgments remain with the client's PV personnel.
What to look for in ICSR processing software
- Multi-channel intake that consolidates email, web, messaging, literature, and partner feeds into one queue
- AI-assisted extraction and narrative generation with confidence scoring and human review gates
- MedDRA and WHO-DD coding with auto-coding and manual override
- A regulatory clock and rules engine for accurate Day 0 and deadline tracking
- Built-in E2B(R2)/E2B(R3) generation and a submission gateway
- Full audit trail, RBAC, and electronic signatures aligned to 21 CFR Part 11
Crucially, the right platform should reduce manual effort without taking regulated decisions out of human hands. That balance — automation for speed, people for judgement — is what makes a modern PV system both efficient and inspection-ready.
Frequently asked questions
ICSR stands for Individual Case Safety Report — the structured record of a single adverse event for one patient and one or more suspect products.
A valid ICSR needs an identifiable patient, an identifiable reporter, a suspect product, and an adverse event or reaction. Missing any of the four means it is not yet a valid case.
No. In a well-designed platform, AI accelerates extraction, coding, and narrative drafting, but qualified personnel review and approve the work, and regulated judgments such as causality and seriousness remain with PV professionals.