A cited health hazard evaluation in half the hours, every harm on the grid and every gap declared
A cited first draft in half the hours, every gap declared.
See one case, screen by screen ↓It is the assessment of how much harm a product issue could cause, made before a recall decision. FDA’s recall rule, 21 CFR 7.41, lists what it weighs: injuries already reported, conditions that could lead to a hazard, the populations at risk, the seriousness and likelihood of harm, and the consequences. Health Hazard Evaluation drafts each of these as a section of your HHE form, with a citation on every sentence.
How to choose AI solutions for life sciences →
Updated
A complaint trend, a CAPA or a deviation shows a product issue in the field, and before anyone decides on a recall, Medical, Quality and Regulatory have to say how much harm it could cause. The answer is built by hand from the complaint summary, the risk management file, the supplier and laboratory investigations, the clinical literature, field data by lot and the instructions for use — injuries to date, conditions that could lead to a hazard, populations at risk, seriousness, likelihood and consequences, one section of the HHE form at a time.
Every severity and every probability has to be defensible, because all three functions sign and the recall class rests on it. Severity gets re-argued from memory instead of read from the risk file. Likelihood is the weakest section: the field failure rate is the figure most often missing, and under time pressure a missing figure is reasoned around — and that guess can set the class.
The working solution’s dashboard figures for hours per evaluation by phase on its sample cases — 26 hours before the agents, 12 with them. Review time goes up on purpose: people spend their hours on judgement, not on searching.
Six agents gather, extract, place, map, draft and check the evidence; Medical, Quality and Regulatory sign.
Seven complaints of visible particles after reconstitution come in on Corventa 20 mg vials, lots 26H03 and 26H10 — 6,200 vials in the US. A CAPA opens HHE-114 on 1 October, due 12 October. Here is what happens next, screen by screen, in the working solution.
Maya signs the harm assessment on every evaluation. Her list shows each case with its harms and highest severity, sentences cited, declared gaps, the Medical, Quality and Regulatory sign-offs, the due date and the stage. HHE-114, the vial stopper case, is drafting with 6 days to go; HHE-118, a needle shield case, has just arrived at intake with 8 documents.
Each harm sits on a 5 × 5 grid: severity rows copied from the risk management file, probability columns from complaints over units in the field. H1, particles infused with a local or vascular reaction, is Severity 3 – Serious from risk file row HZ-18. No reaction has been reported; there are seven particle complaints over 6,200 vials, and a fragment must be missed at inspection and pass the in-line filter. The proposal moves it one band up from the file’s Improbable, to Remote. H2, a dose discarded, is Minor · Probable: 3 discarded doses over 6,200 vials.
“Copied from the risk management file, not estimated.”
The stopper supplier has not sent the coring investigation, so no one knows whether other lots used the same stopper batch; section 2 states the defect only. The laboratory identified all seven particles as stopper rubber but did not size or count them, so the share small enough to pass the filter is unknown. Both cells stay hatched. Maya requests each one from its owner — Lena Ortiz in supplier quality, Tomas Reyes in the QC laboratory.
“Request sent to Tomas Reyes — the cell stays hatched until the data arrives.”
The supplier’s 8D, 8D-ST-2026-044, shows stopper batch ST-5512 was used only in lots 26H03 and 26H10, with coring traced to needle-guide wear on the crimping line. The retained-sample inspection, RSI-26H-0412, found fragments in 3 of 400 vials (0.75 %), all 180 to 600 µm — well above the 0.2 µm filter pore size. The agents read only what is new, and the placement is confirmed at Serious · Remote.
“Placement confirmed on cited evidence. Ready for Medical signature.”
The HHE form follows 21 CFR 7.41 section by section, from product data and problem definition to mitigating factors. Section 2 now carries the root cause, cited to the 8D. Every sentence has a citation chip that opens the page it came from — the risk file row HZ-18 opens beside the draft. Section 11, the conclusion and recommended class, is left for the signers.
For HHE-118, a needle shield that comes loose on a Veltrimab 150 mg/mL prefilled syringe, Maya runs the agents on the document set: complaint summary, risk file, design verification, the supplier’s 8D, two papers, field data and the instructions for use. Four harms are placed, all four rated higher in the field than in the risk file, and 9 of 11 sections drafted with 32 of 32 sentences cited.
“Draft ready — 4 harms, 20 cited facts, 9 of 11 sections drafted.”
The evidence matrix lists each harm with its hazardous situation, population at risk, severity and what has been reported: three needlesticks, six delayed doses, one report of pain or bleeding, and an injection-site infection that is Serious but not reported to date. Twenty cited facts come from all 8 documents — among them the supplier’s finding that shields from cavity 7 pulled off at 3.2 to 7.9 N, below the 12 to 35 N acceptance criterion.
“Severity is never estimated by an agent — it is read from the risk file row and cited.”
Units distributed for lots 26G02 and 26G09 are missing: the field data extract stops at June production, so the complaint rate across all five affected lots cannot be computed. The gap blocks Quality sign-off and names who can close it — Dana Okafor in supply chain. When the figures arrive, the agents read only the new file and draft section 7.
“The completeness checker declares a gap whenever a template field has no source. It never fills one from general knowledge.”
Maya and Sam sign in either order. Maya confirms the harms, severities, populations at risk and consequences — sections 3 to 6 and 8. Sam confirms product data, problem definition and likelihood, and that each declared gap is closed or accepted. Omar Haddad signs last, after both, confirming the recommended class and the reporting decisions that follow.
“Your medical assessment — section 11 · written by you, never by an agent.”
The same evaluation, seen by Medical, Quality, Regulatory, supplier quality and risk management — what their week looked like, and what it looks like now.
Six agents gather, extract, place, map, draft and check the evidence; Medical, Quality and Regulatory sign.
Assembles the case as one document set — complaint summary, current risk management file, instructions for use, field data by lot, and any supplier or engineering investigation — and lists what is missing.
Records each harm, its hazardous situation, the population at risk, events reported to date and the severity from the cited risk file row — every fact quoting its passage and page.
Places each harm on the 5 × 5 severity by probability grid: computes the rate from cited complaint and distribution figures, maps it to its probability band and shows the risk file rating beside it.
Maps each cited fact to the 11 sections of the HHE form, following 21 CFR 7.41, and reports sections with no facts.
Writes sections 1 to 10 in plain, precise language from the mapped evidence only, with a citation on every sentence.
Confirms a cited source for every template field, produces the citation coverage report and declares each gap with the data needed and who owns it.
Stay accountable, all three sign. The agents propose; a named person decides.
Ask why a harm sits where it does, what is missing or who is waiting — every answer points to its source.
Why is HHE-114 placed at Serious, Remote?
Severity 3 – Serious is copied from risk file row HZ-18, not estimated. Remote: no reaction reported; seven particle complaints over 6,200 vials, and a fragment must be missed at inspection and pass the 0.2 micron in-line filter. The risk file rated it Improbable, so the proposal moves it one band up. Two gaps keep the cell hatched; I will not fill them from general knowledge.
What would move HHE-114 to another cell?
Only new evidence, and only on the probability axis — severity stays as the risk file rates it. If retained vials show fragments small enough to pass the 0.2 micron filter, the proposal moves to Occasional, the Not acceptable zone for a Serious harm. If the stopper batch reached other lots, the population grows and the rate is recomputed. The recall class is written by the signers.
What is the failure rate for the affected lots?
For all five lots I cannot give one, and I will not estimate it. Lots 26F11, 26F18 and 26F25: 28 complaints over 41,800 units, 0.67 per 1,000. Lots 26G02 and 26G09: 9 complaints, units not reported. Before June 2026 the rate was 0.01 per 1,000. The missing distribution figures are the declared gap in section 7.
What recall class should HHE-118 get?
That decision belongs to the signers — the agents do not propose a recall class or a medical conclusion. What I can give you is the evidence they sign against: highest severity 3 – Serious (infection, none reported), three minor needlesticks, and a likelihood that waits for the distribution figures.
13 screens from the working solution, on its sample data. Pick one to see it large.
How the agents draft, who signs and in what order, the template and the clock are settings.
| Setting | Default | Choose from |
|---|---|---|
| Draft only from cited evidence | Always on | Locked on — a sentence without a page citation is not written |
| Declare a gap when a section has no source | Always on | Locked on — the section stays blank and the gap is sent to a person |
| Leave the conclusion and the class to the signers | Always on | Locked on |
| Copy severity from the risk management file | On | On · Off |
| Block signing while a sentence is uncited | On | On · Off |
| Search the literature when a population has no source | Off | On (adds up to 5 papers for review) · Off |
| Signing order | Regulatory last | Regulatory last · Strict order · All at once |
| Evaluation template | HHE form · version 4 (11 sections) | Version 4 (11 sections) · Version 3 (9 sections) |
| Evaluation due | 5 working days | Working days from case open to the last signature |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample cases · “target” = the design goal, measured in the live solution · The hours by phase are the working solution’s dashboard figures on its sample cases · Sources: 21 CFR 7.41, health hazard evaluation and recall classification · ISO 14971:2019, risk management. People, products, lots and documents named on this page are characters and sample data in the working solution.
It is the assessment of how much harm a product issue could cause, made before a recall decision. FDA’s recall rule, 21 CFR 7.41, lists what it weighs: injuries already reported, conditions that could lead to a hazard, the populations at risk, the seriousness and likelihood of harm, and the consequences. Health Hazard Evaluation drafts each of these as a section of your HHE form, with a citation on every sentence.
Severity is copied from the cited row of your risk management file. Probability is computed from complaints over units in the field and mapped to its band — Improbable below 1 in 1,000,000 up to Frequent at 1 in 1,000 or more. The risk file rating is shown beside the proposed cell, and the zones come from your risk policy.
Nothing is guessed. The completeness checker declares a gap, names the data needed and who owns it, and the request goes to that person. A placement that rests on the gap stays hatched, and the section stays blank; when the file arrives the agents read only what is new.
No. Section 11 — the health hazard conclusion and recommended class — is reserved for the signers, and that rule is always on. Asked for a class, the assistant gives the evidence the signers sign against instead.
The complaint or issue description, the risk management file, design verification, supplier and laboratory investigations, clinical literature, field data by lot and the instructions for use or prescribing information — whatever is linked to the case or its product.
No. Where the field evidence rates a harm higher than the risk file, the harm appears under Risk file updates with its basis, and you send it to risk management for review. Risk management decides; the agents never edit the risk file.
Yes. Medical, Quality and Regulatory each sign electronically with the meaning of the signature stated, and all three are needed to issue. Any section can be edited or redrafted from the same cited evidence, signing is blocked while a sentence is uncited, and every action by people and agents is on the activity trail.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your HHE form, your risk policy, your risk management files and a set of past cases — and it goes live once every quality gate has passed. We will walk you through it on one of your own past cases first.
Any batch-record question answered in seconds, cited to the page and step.
1-min demoValidation Document AutomationProtocols drafted in minutes, every value traced to its source.
1-min demoManufacturing Batch AnalyticsBatch analysis that took three months now takes hours, with the query behind every answer.
1-min demoWe’ll run Health Hazard Evaluation on the document set of one of your own past cases.