Turn a Clinical-Guideline Update into a Staff-Education One-Pager

The task

Whenever a society (AHA/ACC, IDSA, ADA, USPSTF) drops a guideline update, somebody on the clinical informatics or quality team has to translate it into a one-page handout the floor staff will actually read before huddle. This piece walks through doing that with an LLM, keeping it PHI-free and citation-anchored to the source document.

Before AI

The usual path: skim the 80-page PDF, copy a few tables into Word, argue with the formatting, send a draft to a senior clinician for accuracy check, revise, then re-do the layout so it fits on one side. Most teams budget half a day. Realistically it's 2-3 hours of focused work per update, and the result is often three pages because nobody had time to tighten it. Reference: the 2025 AHA/ACC hypertension guideline overview is a typical example of what lands in your inbox.

The workflow

Step 1 — Extract the changes that actually affect bedside practice. Paste the guideline summary (or the "What's New" section, or the executive summary) as the input. The first prompt strips out everything that isn't a change to clinician behavior.

You are a clinical informatics analyst preparing a staff-education one-pager. The text below is a clinical guideline summary or update.

Your job in this step: extract ONLY the items that change what a bedside clinician (RN, NP, PA, MD, pharmacist) should DO differently compared to the prior guideline. Ignore epidemiology, pathophysiology background, and policy/payer commentary unless it directly changes an order, threshold, or workflow.

Output a structured list with these columns:
- Change (one sentence, plain English)
- Old practice (what we did before, if stated or clearly implied; otherwise "not specified")
- New practice (what to do now)
- Affected role(s)
- Strength of recommendation / evidence level (verbatim from source if given)

Do not invent recommendations. If something is ambiguous, mark it [VERIFY WITH SOURCE]. Do not include any patient identifiers — if the source text contains any, redact them.

Here is the guideline text:

GUIDELINE UPDATE BULLETIN — Internal draft, Cardiology Service Line
Source: 2025 AHA/ACC Guideline for the Prevention, Detection, Evaluation, and Management of High Blood Pressure in Adults (synthesized excerpt for staff training)

Key changes from the 2017 guideline:

1. BP categories retained: Normal (<120/<80), Elevated (120-129/<80), Stage 1 HTN (130-139/80-89), Stage 2 HTN (≥140/≥90). No change to thresholds.

2. NEW: Out-of-office BP measurement (home BP monitoring or 24-hour ABPM) is now recommended (Class 1) to confirm diagnosis of hypertension before starting pharmacotherapy in most adults. Previously this was a Class 2a recommendation.

3. NEW: For adults with confirmed Stage 1 HTN and estimated 10-year ASCVD risk ≥7.5% (lowered from ≥10% in 2017), initiate pharmacotherapy in addition to lifestyle changes. Use the PREVENT risk calculator (replaces Pooled Cohort Equations).

4. NEW: BP target for most adults on treatment is <130/<80 mmHg. For adults ≥75 years who are ambulatory and community-dwelling, the same <130/<80 target now applies (previously individualized).

5. NEW: Screening for primary aldosteronism is recommended (Class 1) in all patients with newly diagnosed Stage 2 HTN, resistant HTN, HTN with hypokalemia, or HTN with adrenal incidentaloma. Plasma aldosterone-to-renin ratio is the initial test.

6. NEW: In pregnancy, treat BP ≥140/90 mmHg (lowered from ≥160/110). First-line agents: labetalol, nifedipine ER, methyldopa.

7. Lifestyle: DASH diet, sodium <2300 mg/day (ideal <1500 mg/day), ≥150 min/week moderate aerobic activity, alcohol ≤2 drinks/day men / ≤1 women, weight loss if BMI ≥25.

8. Resistant HTN: Confirm adherence and rule out white-coat effect with ABPM before escalating. Add spironolactone as preferred 4th agent (Class 1).

Implementation note from Dr. A. Reyes, Cardiology Medical Director: "Roll out to floor staff before Q3. EHR order sets will be updated by informatics — see ticket INF-4412."

Step 2 — Draft the one-pager. Now turn the extracted changes into a layout that fits on one printed side. The constraint is real: anything over ~350 words on the page won't get read at shift change.

Using the structured changes you extracted in the previous step, draft a single-sided staff-education one-pager. Constraints:

- Total length: under 350 words of body text (excluding title and headers).
- Audience: bedside RNs, NPs, PAs, and floor pharmacists. Not specialists.
- Tone: direct, second-person ("you"), no marketing language, no "exciting new updates."
- Structure:
  1. Title line + effective date placeholder [EFFECTIVE: __________]
  2. "What changed" — 3 to 6 bullets, each starting with the action verb (Screen, Treat, Confirm, Target, etc.). One line each.
  3. "What stays the same" — 2 to 4 bullets, so staff don't over-correct.
  4. "At the bedside" — a 3-row table: Scenario | What to do now | EHR / order-set note. Use the placeholder [ORDER SET: __________] where a specific order set name would go.
  5. "Escalate / consult" — one short paragraph on when to call the attending or specialty service.
  6. Footer: source citation, version, and a [REVIEWED BY: __________] line.

Do not include patient examples with names, MRNs, dates of birth, or any other identifiers. Use only role labels ("a 72-year-old ambulatory patient" is fine; "Mrs. Smith" is not).

Output the one-pager in plain markdown.

Step 3 — Clinical accuracy + HIPAA self-check. This is the step most teams skip and regret. Have the model audit its own draft before a human clinician sees it.

Now act as a reviewer with two hats: (a) a board-certified clinician checking the one-pager against the original guideline text from step 1, and (b) a HIPAA privacy officer.

Produce a review in three sections:

CLINICAL ACCURACY
- For each clinical claim or numeric threshold in the one-pager, mark it as ✅ matches source, ⚠️ paraphrased (verify wording), or ❌ not supported by source text provided.
- Flag any recommendation where strength/class of evidence was dropped during simplification.
- Flag any place where simplification may have removed a clinically important caveat (e.g., contraindications, age cutoffs, pregnancy considerations).

PHI / HIPAA CHECK
- Confirm there are no direct identifiers (names, MRNs, DOB, addresses, phone, dates of service tied to an individual, photos, device IDs).
- Confirm no quasi-identifiers that could re-identify in a small unit (rare diagnosis + age + role).
- If the one-pager will be posted in a public area, note any content that should move to an intranet-only version instead.

REVISIONS
- Provide the corrected one-pager in full, with changes applied. Keep it under 350 words of body text. Preserve all [BRACKETED PLACEHOLDERS] for the human reviewer to fill in.

Gotchas

• The model will confidently paraphrase guideline language in ways that shift meaning. Class 1 vs Class 2a recommendations, "consider" vs "recommend," and ASCVD risk thresholds are exactly where it slips. The step-3 audit catches most of it, but a human clinician sign-off is non-negotiable before this goes to staff. Per the ACC summary of the 2025 update, evidence-class language matters and shouldn't be flattened.
• Don't paste the full PDF. Paste the executive summary or the "What's New" section. Full guidelines exceed useful context, dilute the signal, and waste the audit pass.
• Never paste real patient examples, case logs, or audit reports as "examples" of how the guideline applies. Even with names redacted, dates + diagnosis + unit can re-identify. Use the role-label convention in step 2.
• EHR order-set names belong in placeholders, not generated. The model will happily invent an order-set name that doesn't exist in your Epic/Cerner build. The [ORDER SET: __________] placeholder forces informatics to fill it from the actual ticket.
• This workflow does not replace the guideline itself. It produces a teaser for floor staff that points back to the source. Make sure the footer citation survives the final edit.

Time saved

Roughly 2 hours per guideline update, compared to the half-day a manual draft-review-revise cycle typically eats. The savings come from steps 1 and 2 — extraction and first-draft layout — which is where most of the staring-at-Word time lives. Step 3 (clinical accuracy + HIPAA self-check) doesn't save time so much as it raises the floor of what reaches your physician reviewer, which usually means one fewer revision round. Conservative estimate; your mileage will vary by guideline length and how opinionated your medical director is about formatting.

What we got when we ran it

We executed every prompt above, in order, against the sample input (model: claude-sonnet-4-6). Outputs below are real and unedited apart from length truncation.

Step 1 output

# Actionable Practice Changes — 2025 AHA/ACC Hypertension Guideline Update > Source: 2025 AHA/ACC Guideline for Prevention, Detection, Evaluation, and Management of High Blood Pressure in Adults (synthesized excerpt) > Prior reference: 2017 AHA/ACC Hypertension Guideline > Prepared for: Bedside clinician one-pager | Internal education use only --- ## Summary Table of Changes | # | Change (one sentence, plain English) | Old Practice | New Practice | Affected Role(s) | Strength / Evidence Level | |---|--------------------------------------|--------------|--------------|------------------|--------------------------| | 1 | Confirm hypertension diagnosis with out-of-office BP measurement before starting drug therapy in most adults. | Out-of-office BP confirmation was recommended but at a lower priority (Class 2a). | Obtain home BP monitoring or 24-hour ABPM to confirm diagnosis prior to initiating pharmacotherapy. | MD, NP, PA | Class 1 (verbatim from source) | | 2 | Use a lower ASCVD risk threshold (≥7.5%) to decide when to start medication in Stage 1 HTN, and switch to the PREVENT calculator to estimate that risk. | Pharmacotherapy initiated at 10-year ASCVD risk ≥10%; risk estimated using Pooled Cohort Equations. | Initiate pharmacotherapy + lifestyle changes in Stage 1 HTN (130–139/80–89) when 10-year ASCVD risk ≥7.5%, calculated via PREVENT calculator. | MD, NP, PA | Not specified in excerpt — [VERIFY WITH SOURCE] | | 3 | Apply a uniform BP target of <130/80 mmHg to ambulatory, community-dwelling adults ≥75 years; individualized targets are no longer the default for this group. | BP targets for adults ≥75 years were individualized. | Target <130/80

…output truncated at 250 words.

Step 2 output

# Hypertension Management: What Changed in 2025 ### Based on the 2025 AHA/ACC Hypertension Guideline Update [EFFECTIVE: __________] --- ## What Changed - Confirm hypertension with home BP monitoring or 24-hour ABPM before starting any antihypertensive medication in most adults — do not initiate pharmacotherapy on office readings alone. - Screen all patients with newly diagnosed Stage 2 HTN, resistant HTN, HTN + hypokalemia, or HTN + adrenal incidentaloma for primary aldosteronism — order a plasma aldosterone-to-renin ratio as the first test. - Calculate 10-year ASCVD risk using the PREVENT calculator (replaces Pooled Cohort Equations); the threshold to start medication in Stage 1 HTN drops to ≥7.5% (was ≥10%). - Target <130/80 mmHg for ambulatory, community-dwelling patients ≥75 years — individualized targets are no longer the default for this group. - Treat pregnant patients at BP ≥140/90 mmHg (lowered from ≥160/110); use only labetalol, nifedipine ER, or methyldopa. - Add spironolactone as the preferred 4th agent in true resistant HTN — but first confirm adherence and rule out white-coat effect with ABPM. --- ## What Stays the Same - BP categories are unchanged: Normal <120/80, Elevated 120–129/<80, Stage 1 130–139/80–89, Stage 2 ≥140/≥90. - Lifestyle counseling remains first-line alongside or before medication: DASH diet, sodium restriction, ≥150 min/week aerobic activity, alcohol limits, weight loss if BMI ≥25. - Stage 1 HTN without elevated ASCVD risk is still managed with lifestyle changes before adding medication. - Labetalol, nifedipine ER, and methyldopa remain the only recommended antihypertensive agents in pregnancy. --- ##

…output truncated at 250 words.

Step 3 output

REVIEW REPORT --- ### CLINICAL ACCURACY **"What Changed" bullets:** | Claim | Status | Notes | |---|---|---| | Confirm HTN with home BP or 24-hr ABPM before starting pharmacotherapy in most adults | ✅ matches source | Class 1 preserved. | | "do not initiate pharmacotherapy on office readings alone" | ⚠️ paraphrased (verify wording) | Source says "in most adults" — the one-pager carries this qualifier inline but the imperative phrasing is stronger than the source. Confirm no exceptions exist in the full guideline (e.g., hypertensive urgency, very high-risk presentations). | | Screen Stage 2 HTN + resistant HTN + HTN/hypokalemia + HTN/adrenal incidentaloma for primary aldosteronism via aldosterone-to-renin ratio | ✅ matches source | Class 1 preserved. | | PREVENT calculator replaces Pooled Cohort Equations | ✅ matches source | | | ASCVD risk threshold drops to ≥7.5% (was ≥10%) for Stage 1 HTN pharmacotherapy | ✅ matches source | ⚠️ **Evidence class was dropped during simplification** — source excerpt marks this change as "not specified / [VERIFY WITH SOURCE]." One-pager presents it without any caveat. Recommend adding a verify flag. | | Target <130/80 for ambulatory, community-dwelling patients ≥75 years | ✅ matches source | ⚠️ **Clinically important caveat removed:** Source specifies "ambulatory and community-dwelling." One-pager includes this in the table scenario but omits it from the "What Changed" bullet itself. Non-ambulatory or institutionalized older adults may not share this target — the bullet needs the qualifier. | | ⚠️ **Evidence class dropped:** <130/80 target change

…output truncated at 250 words.

---

This content is for informational purposes only and is not medical advice. AI tools used with patient data must meet your organization's HIPAA and privacy requirements.