write-sbar-brief

name	write-sbar-brief
description	Structures a one-page decision briefing as SBAR — Situation, Background, Assessment, Recommendation — with a hard length cap per section and a single named action, so a decision-maker gets the whole picture and the ask on one page. Use when someone asks for a short briefing rather than a report: "give me a 1-pager on the outage", "SBAR on this vendor", "brief me before Thursday's board", "decision memo on the migration". Not for a long research write-up with methods and results (use `write-imrad-report`) or a factual lookup with no recommendation (`grounded-answer`).
license	MIT
metadata	{"category":"writing","method":"SBAR (Situation, Background, Assessment, Recommendation)","origin":"Michael Leonard, Suzanne Graham and Doug Bonacum, Kaiser Permanente, 2004","version":"2.0.0"}

Write SBAR Brief

SBAR is a situational briefing model in four fixed moves — Situation, Background, Assessment, Recommendation — described by Leonard, Graham and Bonacum at Kaiser Permanente in 2004 and spread through hospitals by Haig, Sutton and Whittington (2006) as a "shared mental model". In the original wording: situation — what is going on? background — what is the context? assessment — what do I think the problem is? recommendation — what would I do to correct it? The core principle is that the sender must have formed an assessment and a recommendation before opening the channel, so the receiver gets a diagnosis and an ask rather than a narrative to interpret. It prevents the failure in which a briefing delivers facts in the order they were discovered and leaves the reader to work out both what they mean and what to do.

When to invoke

Invoke when:

• A decision-maker asks for a short briefing: "give me a 1-pager on the outage", "SBAR on this vendor", "brief me before Thursday's board".
• Something changed and someone must act on it — an incident, a competitor move, a supplier failure — and the ask fits on one page.
• A handover or escalation passes a live situation to someone who was not watching it.

Do NOT invoke when:

• The deliverable is a research write-up with methods, results and citations — use write-imrad-report.
• The question is factual with one right answer and no action attached — use grounded-answer.
• The argument needs building and ordering, not compressing into a container — use pyramid-principle; SBAR is the container, the pyramid the geometry inside it.
• The question is too large for one page — use decompose-research-question, then brief each part.
• The evidence is too thin for any recommendation — use abstain-or-escalate rather than guessing in the R slot.

Procedure

1 — Name the decision the brief serves

Write, in one sentence, the decision the reader must make after reading. If no decision can be named, this is a status update, not a brief. The named decision fixes what belongs in Background — only what bears on that decision — and what the Recommendation must address.

2 — Situation: what is happening now, in ≤3 sentences

State the present state in plain language: what happened, when, to what, and how it is known. Present tense, no history, no judgement — "troubling", "impressive", "alarming" pre-empt the Assessment and cost the reader the ability to disagree with it. The first sentence must be intelligible with zero prior context.

3 — Background: the context that makes the Situation legible, in ≤4 sentences

Supply only the history bearing on the named decision: prior events, the trajectory, related moves, what has been tried. Prune anything the reader would not act differently for. Introduce no new claims — every fact must be sourced or derivable from the Situation, because Background is where unsupported assertions hide most easily.

4 — Assessment: what it means, in ≤5 sentences

This is the analytical contribution, and what separates a brief from a summary. Name what is really going on, the pivotal factor, the risks specifically — not "there are risks" — and the uncertainty that would change the picture. If it reads like restated facts, no assessment happened; rewrite it. Grade the sources with rate-source-admiralty when the conclusion rests on one.

5 — Recommendation: one action, with owner, timing and confidence

State what to do — one primary action, at most one fallback; decisions have singular actions. Give the owner and the date. Attach a confidence tag with a one-line justification: this tag is a library convention, not part of the original model, and it lets a reader tell a firm recommendation from a hedge. Below "medium — could go either way", route to abstain-or-escalate rather than shipping a guess.

6 — Cut to one page

Total ≈400–500 words. Over that, either the writing is loose or the question wants a report — cut Background first, Assessment last. The caps are the method: a brief that does not fit is not a brief.

Output template

# Brief: {topic} — {date}

## Situation
{≤3 sentences · present state · what, when, how known · no evaluative adjectives}

## Background
{≤4 sentences · only the history bearing on the decision · no new claims}

## Assessment
{≤5 sentences · what it means · pivotal factor · named risks · the uncertainty that would change it}

## Recommendation
- **Action:** {one action}
- **Owner:** {who}
- **By:** {date}
- **Confidence:** {high | medium | low} — {one-line justification: what it rests on, what would change it}
- **Fallback:** {optional single alternative}

All four SBAR sections are mandatory and must appear in this order; the Recommendation must name an action, an owner and a confidence tag. A brief without a recommendation is a status update, and one without a confidence tag cannot be weighed.

Worked example

Leonard, Graham and Bonacum's own clinical example (2004), which shows the four moves at their shortest:

Situation:      "Dr Preston, I'm calling about Mr. Lakewood, who's having trouble breathing."
Background:     "He's a 54 year old man with chronic lung disease who has been sliding
                 downhill, and now he's acutely worse."
Assessment:     "I don't hear any breath sounds in his right chest. I think he has a
                 pneumothorax."
Recommendation: "I need you to see him right now. I think he needs a chest tube."

Four sentences carry a patient, a trajectory, a diagnosis and an ask. The same shape outside medicine, in an illustrative incident brief with invented figures:

Section	Content
Situation	The payments API has returned 502s for 11% of requests since 03:40 UTC on 14 March; 1,240 checkouts failed. Monitoring and two customer reports confirm it.
Background	Release 4.7 shipped a connection-pool change on 13 March. A similar exhaustion caused a 40-minute outage in November. The pool cap has not moved since 2024.
Assessment	The pattern matches pool exhaustion under peak load, not an upstream provider fault — provider status is green and latency flat. The pivotal unknown is whether traffic is genuinely up 30% or a retry storm inflates it; if retries, raising the cap makes it worse.
Recommendation	Action: roll back release 4.7 and re-measure before touching the cap. Owner: platform on-call. By: 16:00 UTC today. Confidence: medium — one metric series; a retry-storm signature in client logs would change it.

Note the discipline: "11% of requests" is Situation; "matches pool exhaustion, not an upstream fault" is Assessment; the retry-storm caveat sits with the recommendation it would overturn.

Verification

• Count the sentences: Situation ≤3, Background ≤4, Assessment ≤5, and the whole brief ≤500 words.
• Re-read the Situation for evaluative adjectives and move every one of them into the Assessment.
• Check each Background sentence is sourced or derivable from the Situation — delete any claim appearing only there.
• Confirm the Assessment names a specific risk and the uncertainty that would change the conclusion.
• Confirm the Recommendation names one primary action with an owner, a date and a justified confidence tag.
• Confirm the decision named in step 1 is the one the Recommendation addresses.

Pair with adjacent skills

• pyramid-principle — the argument geometry inside the Assessment and Recommendation; SBAR is the container.
• write-imrad-report — the long-form counterpart when the reader needs methods and results, not a decision.
• rate-source-admiralty — grade the sources behind Situation and Background before the Assessment leans on them.
• abstain-or-escalate — the route when confidence is too low to recommend anything.
• critique-report — the pre-delivery review of the finished brief.

Anti-patterns

• Do not exceed one page. A two-page brief is a report wearing a brief's headings.
• Do not editorialise in Situation. Neutral facts there are what let the Assessment land as analysis rather than as spin.
• Do not offer five recommendations. A menu returns the decision to the reader, which is the work the brief was meant to do.
• Do not omit the confidence tag or leave it unjustified; an unqualified recommendation cannot be weighed against what the reader already knows.
• Do not move the Recommendation to the top. The order is the method: the reader reaches the ask having already accepted the situation, the context and the diagnosis.
• Do not put new facts in the Assessment. It interprets what Situation and Background have already established.

Reference

• M. Leonard, S. Graham, and D. Bonacum, "The human factor: the critical importance of effective teamwork and communication in providing safe care," Quality and Safety in Health Care, vol. 13, suppl. 1, pp. i85–i90, Oct. 2004, doi: 10.1136/qhc.13.suppl_1.i85 — SBAR as "a situational briefing model", with the four questions and the clinical example quoted above.
• K. M. Haig, S. Sutton, and J. Whittington, "SBAR: a shared mental model for improving communication between clinicians," Joint Commission Journal on Quality and Patient Safety, vol. 32, no. 3, pp. 167–175, Mar. 2006, doi: 10.1016/s1553-7250(06)32022-3 — implementation at OSF St. Joseph Medical Center, and the observed hesitancy in stating the Recommendation.
• Institute for Healthcare Improvement, "SBAR tool: situation, background, assessment, recommendation," 2024 — attributes the technique to M. Leonard, D. Bonacum and S. Graham at Kaiser Permanente of Colorado. [Online]. Available: https://www.ihi.org/resources/tools/sbar-tool-situation-background-assessment-recommendation