Health Pipeline

This is an End-to-end monitoring pipeline for consumer health device data. Ingests time-series readings from blood pressure monitors, pulse oximeters, smart scales, and heart rate trackers, detects clinical concerns, and generates alerts for clinician review. Designed for adults 65+.

Quick Start

pip install -r requirements.txt
python main.py

Sample outputs are in outputfiles if you want to inspect results without running it.

Functions

Generates synthetic patient data (20 patients × 4 devices × 30 days, ~100K readings) with realistic edge cases — sensor faults, Bluetooth duplicates, time gaps, calibration drift — and runs it through a 5-stage pipeline:

1. Data generation — synthetic patients with different clinical profiles
2. Ingestion — validate, tag faults, deduplicate, store to SQLite
3. Processing — daily aggregation, gap detection, rolling statistics
4. Analysis — three detection layers: threshold checks → trend detection → cross-device correlation
5. Alerts — prioritize, deduplicate, output JSON + clinician summary

Design rationale, assumptions, and tradeoffs are in DESIGN.md.

Output

output/alerts.json Structured alerts (machine-readable) output/alert_summary.txt Clinician summary by patient and severity

Structure

├── main.py                           # Entry point
│   config.py                     # Thresholds, parameters
│   models.py                     # Patient, Device, Reading + DB ops
│   data_generator.py             # Synthetic data with edge cases
│   ingestion.py                  # Validation, dedup, storage
│    processing.py                 # Cleaning, aggregation, rolling stats
│   analyzer.py                   # Three-layer analysis
│   alert_engine.py               # Prioritization and output
├── DESIGN.md                         # Assumptions, tradeoffs, decisions
└── requirements.txt

Requirements

Python 3.9+, numpy, pandas.