The letter of instruction decides the report: scope, questions and stated assumptions.

The shortest document in a medico-legal brief is usually the most consequential one. The letter of instruction runs to a few pages against the file's several hundred, and yet it decides almost everything about the report that follows: which questions get answered, which facts are assumed rather than found, which materials the opinion is entitled to draw on, and where the opinion's boundaries sit.

A report that treats the letter as a cover sheet, to be skimmed and set aside, tends to discover its real function later, in a supplementary request or in the witness box. Read properly it is the report's contract, and the disciplines that follow from that reading are the subject of this piece.

The letter of instruction is the report's contract

A letter of instruction does four jobs at once:

• it identifies the matter and the parties;
• it puts the questions the expert is retained to answer;
• it lists the materials supplied;
• it states any assumptions the expert is asked to adopt.

The expert evidence rules are built around exactly this structure. Under the Harmonised Expert Witness Code of Conduct, applied in the Federal Court and across the harmonised state courts, every report must state the material facts and assumptions on which each opinion is based, and the code contemplates the letter of instructions being annexed to the report for precisely that purpose.¹ New South Wales binds experts to the same discipline through the code of conduct in Schedule 7 of the Uniform Civil Procedure Rules, and a report may not be received in evidence unless the expert acknowledges that code and agrees to be bound by it.²

The consequence is that the letter and the report are read together, as offer and performance: the letter defines what was asked, and the report is measured against it line by line. Everything else in this piece follows from that one reading.

Answer every question asked, and only those

The first discipline is coverage, and it cuts in both directions. Avant's guidance to report writers puts the positive half plainly: read the briefing letter thoroughly so you are clear about what it is asking, and write a report that answers the questions asked of you with the reasons for your opinion.³

A skipped question is never neutral. At best it triggers a supplementary request, which reopens the file and resets the clock; at worst it invites the inference that the question was avoided because the honest answer was unhelpful, which is a far worse position than any unhelpful answer would have been. The reliable protection is structural rather than heroic: take the questions verbatim, number the report's opinion sections to mirror them, and make the final check before signing a reconciliation of one list against the other, so that a gap has to be visible before it can ship.

The negative half of the discipline gets less attention and causes at least as much trouble. An opinion nobody asked for sits outside the contract, which means it sits outside the materials assembled to support it and often outside the question the retained expertise was matched to. Under cross-examination it draws the one question with no good answer: who asked you to address that?

Volunteered opinions are also where advocacy tends to leak in, because an expert who goes beyond the questions has, by definition, decided independently what the case needs, and deciding what the case needs is the retaining party's job rather than the witness's. Where something genuinely material falls outside the questions, the honest move is to say so and invite a further instruction rather than annex an unrequested opinion to the report. Scope is what keeps the expert an expert rather than a participant.

Assumptions are adopted on the record, not absorbed

The second discipline is keeping two kinds of fact visibly separate: what the records show, and what the expert has been instructed to assume. The law is exacting about this because the opinion's admissibility depends on it.

In Makita (Australia) Pty Ltd v Sprowles, the New South Wales Court of Appeal set out the canonical statement: so far as an opinion rests on facts the expert observed, they must be identified and proved by the expert, and so far as it rests on assumed or accepted facts, they must be identified and proved in some other way, with the report exposing how the expertise applies to those facts to produce the opinion.⁴ The High Court confirmed the same structure in Dasreef Pty Ltd v Hawchar: an opinion whose factual basis is not established and whose reasoning is not exposed may carry little or no weight at all.⁵

For the report writer, this translates into a form of words worth being pedantic about. An instructed assumption enters the report explicitly, as "I am instructed to assume that the applicant ceased work on 4 March 2023", so that the assumption visibly belongs to the instructing party, and it is the instructing party's task to prove it.

What the writer must never do is let an assumption soak silently into the narrative, so that an instructed fact and a found fact become indistinguishable on the page, because at that point the expert has adopted as their own something they never verified.

The same explicitness handles the awkward case where the file contradicts the instruction: the report can answer on the assumption as instructed while recording, in plain sight, that the records suggest otherwise. Where the questions allow it, the strongest reports answer in the alternative, so that if the assumption holds the opinion is X, and if the records are preferred it is Y. That structure costs a paragraph and removes an entire line of cross-examination, because the opinion no longer stands or falls with a fact the expert was never in a position to establish.

When a question cannot be answered, the code says how to say so

Not every question in a letter of instruction can or should be answered, and the rules anticipate this rather than punish it. The harmonised code requires a report to state, where applicable, that a particular question falls outside the expert's area of expertise, and to disclose where an opinion is not a concluded one because the research or the data was insufficient.¹ The Medical Board's code of conduct imposes the same duty from the professional side: good medical practice means making clear the limits of your knowledge, and not giving opinion beyond those limits when providing evidence.⁶

A psychiatrist asked to rate a shoulder, an examiner asked about causation when the pre-injury records were never supplied, a file-review report asked for findings only an examination could ground: each of these is answered properly by naming the limit rather than improvising past it. Declining a question on the record reads as rigour. Improvising an answer reads as rigour right up until the first probing question, and never afterwards.

Scope drift is where reports get taken apart

These disciplines matter because the letter of instruction is the first tool a careful opposing reader picks up. The reconciliation is mechanical, which is exactly what makes it dangerous: questions in the letter against sections in the report, materials listed in the letter against the documents-reviewed table, assumptions stated in the letter against assumptions acknowledged on the face of the opinion.

Every mismatch is an opening:

• a question with no answering section suggests avoidance;
• an answer with no originating question suggests advocacy;
• a document supplied but never listed suggests the file was not fully read;
• an assumption relied on but never declared suggests the expert did not notice where their facts came from.

None of these openings requires the opinion to be wrong. That is the uncomfortable part: a clinically sound report can be dismantled on scope alone, because scope is the part that can be checked without any medical knowledge at all.

The drift continues after the report goes out, which is why the contract has to be maintained rather than merely honoured once. Supplementary questions arrive, new records surface, and the harmonised code requires an expert who changes their opinion on a material matter to provide a supplementary report saying so.¹

Each of these events is, in substance, a variation of the instruction, and it deserves the same treatment as the original letter: the new question answered expressly, the new material added to the documents list, the changed opinion tied to the specific record that changed it. Practices that treat the follow-up round informally, as correspondence rather than as instruction, are the ones whose files stop reconciling, and a file that no longer reconciles cannot be defended by anyone, including its author.

Contract administration, wrapped around the clinical work

Run the letter of instruction like the contract it is. Five checks carry most of the weight, and not one of them requires clinical judgement:

• extract the questions verbatim and number them;
• mirror those numbers in the report's opinion sections;
• state every instructed assumption in terms that mark it as instructed;
• reconcile the enclosure list against the documents-reviewed table before writing, and again before signing;
• close with a pass that checks one thing only, that every question has its answer, every answer has its question, and every assumption is on the record.

That is contract administration wrapped around the clinical work, which is precisely why software can carry it.

Trenthos IME is designed for that layer. It reads the letter of instruction alongside the records, extracts the questions and the stated assumptions, structures the draft so each question has its answering section, and flags a question left unanswered or an assumption not yet declared, so a gap surfaces at drafting rather than in the other side's hands. The examination, the reasoning and the opinions remain the doctor's, and nothing leaves until the doctor has reviewed every line and signed it.

The letter decides the report's shape either way. The only choice is whether that happens deliberately at the start, or expensively at the end.