BP Log Sentinel

Use when someone logs home blood-pressure readings and asks what they mean, whether their BP is high, if they should see a doctor, or whether the doctor's office reading disagrees with home readings — computes the AHA 7-day home average, categorizes against ACC/AHA 2017 + ESC/ESH 2023 thresholds, detects white-coat and masked-hypertension patterns, flags measurement-technique errors, and renders a physician-ready summary.

Overview

Home BP monitors are everywhere; correct interpretation of the log is not. People make three systematic mistakes: (1) reacting to single readings instead of averages, (2) comparing home numbers to office thresholds (home runs 3-5 mmHg lower, so home 135/85 ≈ office 140/90), and (3) measuring badly — caffeine, full bladder, talking, crossed legs each add 5-10 mmHg, which is an entire diagnostic category.

This skill turns a raw CSV log into a consult-grade analysis: the AHA home-monitoring average (discard day 1, mean of days 2-7, morning + evening), category per ACC/AHA 2017 and ESC/ESH home thresholds, white-coat vs masked-hypertension pattern when a clinic reading is supplied, contaminated-reading flags from free-text notes, crisis-range escalation, and a doctor-ready summary paragraph.

When to Use

• "I've been logging my BP for a week — is this high?"
• "My home readings are 135/85 but at the doctor's it was 150/95 — which is real?"
• "My monitor says 138/88 every morning. Do I have hypertension?"
• "Should I be worried about this one 180/120 reading?"
• Preparing evidence for a doctor's appointment from a home log.
• Any CSV/list of date+time+sys/dia readings needing interpretation.

Don't use for: stroke/heart-attack symptoms, hypertensive crisis with symptoms (→ emergency services immediately), titrating medication doses (physician's job), or ambulatory 24-h monitor interpretation (different protocol).

How It Works — Numbered Steps

1. Get the log into CSV: date,time,systolic,diastolic,pulse (+ optional notes — free text is scanned for technique contaminants like coffee/talking/legs crossed).
2. Run: python3 scripts/bp_sentinel.py analyze readings.csv --clinic 150/95 (--clinic optional, enables white-coat/masked detection; --window 7 truncates the log; --keep-day1 overrides the AHA day-1 discard).
3. Read the HOME AVERAGE line — this, not any single reading, is the diagnostic quantity.
4. Check MEASUREMENT-ERROR FLAGS: contaminated readings still count, but a clean re-log is recommended before believing a borderline category.
5. If CLINIC COMPARISON is present, read the pattern verdict (white-coat / masked / sustained).
6. Escalate per NEXT STEPS: any ≥180/120 with symptoms = emergency care now.

The Numbers That Matter

Quantity	Threshold	Meaning
Home average (day 2-7)	<120/80	Normal
	120-129/<80	Elevated
	≥135/85 (home trigger)	Hypertensive territory (≈ office 140/90)
	≥150/95 (home)	Stage 2 territory
Single reading	≥180/120	Crisis range — with symptoms, emergency

Risk framing: each 20/10 mmHg rise roughly doubles cardiovascular mortality risk — why category boundaries matter and why single readings don't.

Common Pitfalls

1. Reading the monitor's built-in average. Device averages include day 1 (anxiety-inflated) and don't follow the morning/evening protocol; the AHA day-2-7 average is the reference.
2. Applying office thresholds to home numbers. Home BP runs lower; a home 135/85 is already equivalent to office 140/90. The tool applies home triggers.
3. Wrist/cuff-position errors. Wrist monitors must be at heart level; an unsupported arm below heart level adds up to 10 mmHg.
4. Judging by the highest reading. Trends and averages diagnose; spikes (unless ≥180/120 with symptoms) are usually technique or noise.
5. Ignoring white-coat/masked patterns. Clinic-only or home-only evidence alone misses both; the pattern needs both numbers, which is why --clinic exists.
6. Stopping the log after diagnosis starts. Home monitoring also guides treatment — keep logging monthly weeks.
7. Wrong cuff size. A too-small cuff overreads by 5-15 mmHg — measure arm circumference once and buy accordingly.

Example

text

User: "Doctor's office said 128/82, but my home cuff shows ~135/86 every
evening for a week. One night it hit 184/122."

$ python3 scripts/bp_sentinel.py analyze week.csv --clinic 128/82
HOME AVERAGE: 139/89 mmHg  (12 readings over 6 days, day 1 discarded)
CATEGORY: Stage 1 hypertension (home average)
CLINIC COMPARISON: clinic 128/82 vs home 139/89
  MASKED HYPERTENSION: home high, clinic normal
  -> carries risk close to sustained hypertension; this log is exactly
     the evidence your physician needs.
!! 1 reading ≥180/120 (crisis range): 2026-10-02 20:14 — with symptoms,
   emergency care NOW.

Answer: this is a classic masked pattern — the office reading under-detected
it. Print the report and book an appointment; treat the 184/122 night as an
escalation signal if it recurs with symptoms.

Safety Note

Information tool, not a diagnosis; diagnosis requires a clinician confirming with proper technique. Thresholds: ACC/AHA 2017 guideline, ESC/ESH 2023 home-monitoring guidance. Symptoms (chest pain, breathlessness, neuro deficits, vision change) + high reading = emergency services.

References

• references/measurement-protocol.md — the 10-rule AHA measurement procedure, why each rule shifts numbers, and the correct 7-day logging schedule.
• references/interpretation.md — threshold tables (office vs home vs ambulatory), white-coat and masked hypertension epidemiology and prognosis, and how physicians use home logs.