Analytics Platform

Real-world evidence analytics for oncology patient cohorts, built on top of CTOMOP (OMOP CDM v6.0).

Overview

Select a patient cohort using 20+ clinical criteria, then instantly explore outcomes across response rates, treatment patterns, demographics, disease staging, lab values, and more — all derived from structured real-world data in a PostgreSQL OMOP CDM database.

Features

Cohort Builder

Filter patients by:

• Disease (Multiple Myeloma, Breast Cancer)
• ISS stage, ECOG performance status
• Age range, gender, ethnicity, geographic region
• Cytogenetic markers (del17p, t(4;14), t(14;16), high-risk flag, TP53 disruption)
• Lines of therapy (min/max), 1L/2L/3L+ regimen and outcome
• Refractory status
• MM-specific: CRAB criteria, bone lesions, prior ASCT, plasma cell leukemia
• Lab ranges: hemoglobin, β2-microglobulin, albumin, creatinine
• Diagnosis year range, smoking status

Dashboard Panels

• Response Rates — stacked bar by therapy and outcome (CR/VGPR/PR/MR/SD/PD), switchable across 1L/2L/3L+, with chart/table toggle and ORR%
• Treatment Patterns — 1st-line regimen frequency
• Lines of Therapy — funnel (% reaching each line) + distribution donut
• Patient Demographics — age buckets, gender donut, ethnicity and geographic distribution
• Disease Staging — ISS stage donut, ECOG bar, cytogenetic markers (high-risk highlighted), CRAB criteria, bone lesions, ASCT rate
• Laboratory Values — box plots (median/IQR/range) for 9 key lab values
• Treatment Duration — median months by therapy+outcome (top 15), time-to-first-treatment distribution
• Top Treatment Sequences — most common multi-line therapy paths (e.g. VRd → Daratumumab → Cilta-cel)

Stack

Layer	Technology
Backend	Django 5 + Django REST Framework
Database	PostgreSQL (OMOP CDM v6.0) via CTOMOP
Frontend	React 19 + TypeScript + Vite
Styling	Tailwind CSS v4
Charts	Recharts

Project Structure

analytics/
├── backend/                    # Django + DRF API
│   ├── analytics_project/      # Django project settings and URLs
│   ├── patients/               # Unmanaged PatientInfo model (reads patient_info view)
│   ├── cohorts/                # Cohort filter logic and form settings endpoint
│   └── metrics/                # Analytics computation services
│       └── services/
│           ├── response_rates.py
│           ├── treatment_patterns.py
│           ├── demographics.py
│           ├── staging.py
│           ├── labs.py
│           └── treatment_duration.py
├── frontend/                   # React SPA
│   └── src/
│       ├── components/
│       │   ├── CohortPanel/    # Dark sidebar with collapsible filter sections
│       │   ├── Dashboard/      # Main layout with sticky header
│       │   ├── charts/         # One component per dashboard panel
│       │   └── ui/             # MetricCard wrapper
│       ├── hooks/
│       │   └── useAnalytics.ts # State management + 400ms debounced API calls
│       ├── api/client.ts       # Axios client + filter serialization
│       └── types/index.ts      # Shared TypeScript interfaces
└── seed_mm_patients.py         # Seeds 100 realistic MM patients into CTOMOP

Getting Started

Backend

cd backend
pip install -r requirements.txt
python manage.py runserver

The API is available at http://localhost:8000/api/.

Key endpoints:

• GET /api/form-settings/?disease=Multiple+Myeloma — returns therapy lists, staging options, etc. for the cohort builder
• GET /api/metrics/?disease=Multiple+Myeloma&[filters...] — returns all dashboard metrics for the current cohort

Frontend

cd frontend
npm install
npm run dev

The app is available at http://localhost:5173/. API requests to /api/* are proxied to the Django backend at http://localhost:8000.

Seeding Sample Data

To populate 100 realistic Multiple Myeloma patients with clinically plausible treatment histories and outcome distributions:

python seed_mm_patients.py

Outcome probabilities are drawn from published trial data (GRIFFIN, MAIA, KarMMa, CARTITUDE-1, DREAMM-2, etc.) and risk-adjusted for high-risk cytogenetics.

Deploying to Render

Root Directory: backend

Build Command:

pip install -r requirements.txt && python manage.py collectstatic --noinput

Start Command:

gunicorn analytics_project.wsgi:application --bind 0.0.0.0:$PORT

Environment variables:

Variable	Value
DEBUG	false
SECRET_KEY	(generate a random secret)
CORS_ALLOWED_ORIGINS	https://your-frontend.onrender.com

Database

The backend connects to a CTOMOP PostgreSQL instance and reads from the patient_info denormalized view via an unmanaged Django model. Connection settings live in backend/analytics_project/settings.py.