health-report-generator

name	health-report-generator
description	Generates weekly and monthly medication adherence reports for patients, caregivers, and physicians. Identifies patterns, trends, and risk flags in a human-readable format.
allowed-tools	Bash Read Write

Health Report Generator Skill

Purpose

This skill transforms raw adherence log data into actionable, readable reports that help caregivers and physicians make better decisions — without needing to dig through raw data.

When This Skill Triggers

• Every Sunday at 9:00 AM (weekly report, auto-scheduled)
• First day of every month (monthly summary)
• On-demand: patient or caregiver asks "Can I get a report?" or "How is [Name] doing?"
• Before a scheduled physician appointment (pre-appointment summary)

Report Types

1. Weekly Patient-Facing Report

Simple, encouraging, and actionable. Sent to patient via their preferred channel.

Format:

📋 MedGuard Weekly Summary — Week of Aug 12–18, 2025
Patient: Margaret Collins

This week you took 42 out of 49 scheduled doses. That's 86%! 🌟
Great job with your morning Metformin — 7 for 7!

📌 One to watch: Your evening Warfarin was missed 3 times this week.
   Would you like me to add a phone call backup reminder for 6 PM?

Next week's tip: Taking Warfarin at the same time every day helps keep your levels stable.

2. Weekly Caregiver Report

More detailed with trend data and specific missed dose information.

Format:

MedGuard Weekly Caregiver Report
Patient: Margaret Collins | Week: Aug 12–18, 2025
Generated: Sun Aug 18, 2025, 9:00 AM

ADHERENCE SUMMARY
─────────────────
Overall adherence this week: 86% (↑4% from last week)

By Medication:
  Metformin 500mg (Morning)   ████████████ 7/7 — 100% ✅
  Metformin 500mg (Evening)   ████████──── 6/7 — 86%
  Warfarin 2.5mg              ████────────  4/7 — 57% ⚠️

MISSED DOSES
────────────
• Warfarin 2.5mg — Missed: Aug 13 (6 PM), Aug 15 (6 PM), Aug 17 (6 PM)
  Pattern: Every other day. Possible confusion with schedule?
  Recommended: Confirm schedule with Dr. Rao at next visit.

ALERTS SENT THIS WEEK
─────────────────────
• Aug 13: Level 1 FYI alert — Warfarin missed — Acknowledged ✅
• Aug 15: Level 2 alert — 2nd Warfarin miss — Acknowledged ✅
• Aug 17: Level 3 alert — Pattern detected — Action pending ⚠️

PATIENT INTERACTIONS
────────────────────
Total interactions: 24 | Initiated by patient: 6
Voice notes logged: 2 ("felt a bit dizzy after Warfarin on Aug 13")

3. Pre-Appointment Physician Summary

Compact, clinical, and ready to share with a provider.

Format:

MedGuard Pre-Appointment Summary
Patient: Margaret Collins | DOB: 1948-03-14
Appointment: Dr. Priya Rao, Aug 22, 2025

30-Day Adherence: 79% (Below target of 80%)
Critical Medication Adherence (Warfarin): 62% — FLAGGED ⚠️

Reported symptoms:
• Aug 13: "felt dizzy" after Warfarin — logged

Missed Dose Pattern:
• Warfarin misses cluster around weekends and evenings
• Metformin adherence is stable and above 90%

Recommended Discussion Points:
1. Review Warfarin dosing schedule — patient may benefit from simplified timing
2. Dizziness symptom after Warfarin — possible interaction or dose review needed
3. Consider blister-pack pharmacy service to reduce confusion

[Download Full 30-Day Log as PDF]  [Export as HL7 FHIR JSON]

Data Processing Steps

1. Read all adherence_*.jsonl files for the patient in the target date range
2. Compute per-medication adherence rates
3. Detect missed dose patterns (time of day, day of week clustering)
4. Cross-reference with any logged symptoms
5. Pull caregiver alert history for the period
6. Generate natural language summary using the templates above
7. Deliver via patient's preferred channel (app, email, SMS)

Pattern Detection Heuristics

• Weekend drop: adherence on Sat/Sun < 70% while weekday avg > 85%
• Evening drop: evening doses missed >30% more often than morning doses
• Alternating miss: missed every other day — possible schedule confusion
• Streak break: 5+ consecutive days of perfect adherence followed by sudden miss — often warrants a gentle check-in

Output Formats

• Patient: Plain-text or WhatsApp-compatible message
• Caregiver: Email with embedded charts, plain-text fallback
• Physician: Structured PDF + optional FHIR-compatible JSON export