A report is challenged three years after it was signed. The variant it called likely pathogenic has since been downgraded, and the family's solicitor wants to know what the signing scientist knew, what alternatives were considered, and whether the decision was reasonable at the time. The laboratory's system shows a name, a date and a status of 'authorised'. The draft the scientist saw cannot be reproduced, because the database records have been updated since, and the evidence open on the scientist's screen that afternoon is not recorded anywhere.
The scientist remembers weighing it carefully. The record shows a click.
What the field knows
Sign-off is where accountability sits. The laboratory standards (Rehm et al., 2013) require that a qualified individual review and approve each report, and accreditation schemes in every jurisdiction audit that the approval happened. What the standards say less about is what the approval consists of. In practice it is a judgement that the evidence assembled supports the conclusions stated, that the alternatives were considered, and that the report answers the clinical question, made by a person who takes responsibility for it.
The evidence base behind a signature moves. Classifications change (Mersch et al., 2018), and the ACMG points to consider on re-evaluation (Deignan et al., 2019) treat a signed report as a statement about a point in time that may need revisiting. A decision made under one state of the evidence can be entirely reasonable and later wrong, and the difference between a reasonable decision and a careless one is visible only if the state of the evidence at the time was kept.
Other safety-critical fields solved the recording problem long ago. Aviation, pharmacy and surgery each record that a check was done, what was checked and what the checker saw, because an investigation after an incident needs to distinguish a system failure from an individual one. Clinical genomics has the same need and, in most laboratories, a thinner record than any of them.
There is also a quieter purpose. A sign-off that records the reasoning is a teaching document. The next scientist to see a similar case can read why a predecessor decided as they did, and a laboratory can see, across its sign-offs, where its judgements drift from each other. Two reviewers who would classify the same variant differently can be found before a family is caught between them.
Where practice breaks
The system stores the fact of approval and discards the object approved. The report document is kept; the evidence pages, the candidate list, the database versions and the colleagues' comments that were in front of the signer are in browser history, email and memory. Three years later the document can be read and the decision cannot be reconstructed.
Sign-off is applied to the wrong thing. It authorises a document, a PDF or a template filled in, rather than a state of the case. If the document was edited after review, or assembled from a case that has since changed, nobody can say which version of what was approved.
The alternatives vanish at the moment of decision. A scientist who considered two hypotheses and chose one signs a report that mentions only the one. The second, and the reason it lost, is gone from the record, so a later reviewer who finds new evidence for it cannot tell whether it was ever weighed.
Finally, sign-off becomes a formality under load. A queue of drafts, a reviewer at the end of the day and a system that asks only for a click produce exactly the thin record the system was built to hold. The signer's care is real and the system cannot see it, which is unfair to the signer when the decision is later questioned.
The clinician who countersigns is in the same position. A clinical sign-off on a laboratory report is a second judgement, that the report answers the question asked for this patient, and it is recorded, if at all, as a letter filed to the notes. When the two judgements disagree, nothing in either system shows that they did.
How Tulparus handles it
Sign-off in the workbench is an action on a specific, immutable version of the case. When a reviewer signs, the system records which case version was approved, the knowledge releases and model versions that version was analysed against, the hypotheses it contained with their standing, the evidence attached to each, and the open items the reviewer accepted as open. The report is rendered from that version, so the document and the decision cannot drift apart.
The reviewer is asked to decide, in words, about the things that need deciding. The hypothesis the report rests on and the ones set aside are each confirmed or returned, with a reason that the reviewer writes or selects. Where the system has flagged a gap, the reviewer records whether it is accepted as a limitation or sent back for more work. A click that approves everything at once is not available. The record takes a few minutes longer to make than a click, and it is the few minutes that an investigation three years later will turn on.
The decision record is readable later as a whole. A reader, whether a colleague at reanalysis or an auditor years on, can open the signed version and see what the signer saw: the same evidence at the same versions, the same alternatives, the same open questions, with the signer's reasons beside them. The state of the evidence at the time is a stored fact. Nobody has to reconstruct it from memory, and nobody can be accused of having done so.
For the laboratory in the opening case, the answer to the solicitor would be the signed case version itself: the classification, the submissions and population data as they stood that day, the alternative hypotheses with the reasons they were set aside, and the reviewer's own words. Whether the decision was reasonable would be a matter of reading it, which is what a decision record is for.
What you can do tomorrow
A decision record can be kept alongside any reporting system, starting this week.
- At sign-off, save a snapshot of what was reviewed: the candidate list, the evidence pages as exported, and the database versions. A dated folder per report is enough to begin.
- Write one sentence per report on the alternatives considered and why they were set aside. It is the part most often needed later and most often missing.
- Sign a version, not a document. If the report is edited after review, review it again and record that you did.
- Separate 'accepted as a limitation' from 'not noticed'. If a gap is known at sign-off, say so in the record, so that a later reader does not mistake a decision for an oversight.
Sources
- Rehm HL, Bale SJ, Bayrak-Toydemir P, et al. ACMG clinical laboratory standards for next-generation sequencing. Genetics in Medicine, 2013.
- Mersch J, Brown N, Pirzadeh-Miller S, et al. Prevalence of variant reclassification following hereditary cancer genetic testing. JAMA, 2018.
- Deignan JL, Chung WK, Kearney HM, Monaghan KG, Rehder CW, Chao EC. Points to consider in the reevaluation and reanalysis of genomic test results: a statement of the American College of Medical Genetics and Genomics (ACMG). Genetics in Medicine, 2019.
- Richards S, Aziz N, Bale S, et al. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genetics in Medicine, 2015.