AugMed Platform Documentation¶
What is AugMed?¶
AugMed is a clinical case review platform for research on AI-augmented medical decision-making. It presents clinicians with de-identified patient cases drawn from electronic health record data (structured in the OMOP Common Data Model) and collects their diagnostic assessments, risk ratings, and clinical recommendations.
Each participant sees a carefully controlled subset of clinical information — determined by the researcher's experimental design — which may or may not include AI-generated predictions. This controlled information disclosure is the core experimental mechanism: by varying which features each participant sees, researchers can measure how AI assistance, clinical history details, and other factors influence clinical judgment.
AugMed is domain-agnostic: any clinical scenario that can be represented in OMOP CDM format can be studied on the platform. See Adapting AugMed for Your Study for how to configure the platform for your clinical domain.
Who This Documentation Is For¶
This documentation is written for:
- Researchers who want to design and run experiments using the platform, export data, and analyze results. No programming background is assumed for the researcher-facing guides.
- Administrators who manage deployment, database configuration, and system maintenance.
- Developers who need to understand the API, database schema, or extend the platform.
Quick Links¶
Deploy¶
- One-Click Deploy (Railway) — Deploy in under 5 minutes, no technical setup required
- Self-Hosted Deploy (Docker Compose) — Run on your own machine or server
- Development Setup — Set up a local development environment
Getting Started¶
- Platform Overview — Architecture, how experiments work, key concepts
- Quick Start Guide — Set up a local development instance
- Terminology Glossary — Definitions of all key terms
Researcher Guide¶
- Creating Experiments — Design arms, build config files, upload assignments
- Managing Participants — Bulk user creation, activation, monitoring
- Monitoring Progress — Track completion rates and timing
- Exporting Data — Run the export script, understand the CSV output
- Analyzing Results — R and Python examples for common analyses
Admin Guide¶
- Deployment — AWS infrastructure, CI/CD, environment variables
- Database — Schema overview, key tables, relationships
- Configuration — System config, answer configs, display config format
- Troubleshooting — Common issues, logs, health checks
Adaptive Experiments (RL)¶
- Overview — How adaptive experimentation works with Thompson Sampling
- Setting Up an Adaptive Experiment — Step-by-step guide to creating and running RL experiments
- Interpreting Results — Understanding arm weights, rewards, and statistical analysis
Reference¶
- API Reference — All REST endpoints with request/response formats
- Data Dictionary — Every table and column with types and descriptions
- Config CSV Format — Detailed specification for display config uploads
- OMOP Mapping — How OMOP concept IDs map to displayed clinical features
Guides¶
- Adapting AugMed for Your Study — Step-by-step guide for configuring AugMed for a new clinical domain
Examples¶
- CRC Screening Study — Complete worked example from the AIM-AHEAD colorectal cancer screening study
Live Site¶
The production instance of AugMed is available at: https://augmed1.dhep.org/
Repositories¶
- Backend API (this repository): Flask + PostgreSQL
- Frontend: augmed-app-v2 (React + TypeScript)
- RL Service: augmed-rl (FastAPI, Thompson Sampling)
- Infrastructure: augmed-infra (Terraform, AWS)