"When expert panels span continents and time zones, the methodology for reaching consensus must be as rigorous as the guidelines it produces."
| Attribute | Value |
|---|---|
| Status | Incubating |
| Maturity | Active Development |
| License | Apache-2.0 |
| Part of | Evidence Commons |
| Mission Pillar | Pillar 6 (Governance) + Pillar 7 (Education) |
Digital Delphi is a platform for running modified Delphi consensus studies digitally, designed for clinical guideline development where in-person panels are logistically impractical. It provides multi-round survey orchestration, structured panelist management, anonymous feedback synthesis, and convergence tracking. The initial use case is ABIGIS (African Brain Injury Guidelines Intelligence System) -- a project to develop Africa's first TBI clinical guidelines through structured expert consensus.
The parent platform contains an implementation across 7 database tables (all with row-level security), 4 edge functions (1,141 LOC), and 7 UI artifacts (2,579 LOC). This implementation is functional but untested: there are zero tests, known schema column mismatches (vote_value vs rating), and RLS gaps that need hardening. It is not production-ready. This repository is intended to hold the extracted, standalone Delphi methodology engine suitable for any clinical consensus process. The ABIGIS protocol has not been pre-registered on OSF.
| Component | Description | Platform Status |
|---|---|---|
| Study Manager | Study creation, round configuration, panelist enrollment | Implemented (DB tables) |
| Vote Collection | Multi-round response submission with rationale capture | Implemented (delphi-submit-response, edge function) |
| Round Controller | Round opening/closing, convergence detection | Implemented (delphi-close-round, edge function) |
| Status Reporter | Per-study and per-round status aggregation | Implemented (delphi-status, edge function) |
| Application Flow | Panelist self-nomination and approval workflow | Implemented (submit-digital-delphi, edge function) |
| UI Pages | 4 pages + 3 components, routes wired in App.tsx | Implemented (2,579 LOC) |
| Agreement Metrics | Fleiss kappa, ICC, convergence thresholds | Designed, not implemented |
Database tables: delphi_studies, rounds, items, panelists, votes, vote_rationale, applications -- all with RLS policies.
What exists in the parent platform:
- 7 database tables with row-level security (schema deployed)
- 4 Supabase edge functions (1,141 LOC total)
- 7 UI artifacts -- 4 pages and 3 components (2,579 LOC total)
- Basic Delphi workflow: create study, add panelists, collect votes, close rounds
What does not exist yet:
- Test suite (zero tests across all components)
- Schema fix for
vote_valuevsratingcolumn mismatch - RLS policy hardening (known gaps in current policies)
- Agreement metric calculations (Fleiss kappa, ICC)
- Convergence detection algorithms
- OSF pre-registration for the ABIGIS protocol
- Standalone extraction from the platform codebase
- Fix schema column mismatches (
vote_value/rating) in platform - Write integration tests for the 4 edge functions against the 7 tables
- Harden RLS policies to close identified gaps
- Extract the Delphi workflow engine as a standalone module
- Implement agreement metrics (Fleiss kappa, ICC) and convergence detection
- Define a platform-agnostic API specification for Delphi studies
- Publish v0.1.0 with core engine, metrics, and reference UI
graph LR
A[Expert Panelists] --> B[Digital Delphi]
B --> C[ABIGIS<br/>TBI Guidelines]
B --> D[AIDA<br/>computable guidelines]
style B fill:#2A9D8F,stroke:#1E3A8A,color:#fff
Digital Delphi is part of the Lab-in-a-Box product family (guideline production) and feeds into the AIDA Infrastructure product group (computable guidelines for ministries of health). The ABIGIS use case connects TBI clinical guideline development in Africa with the broader BRIDGE-TBI evidence infrastructure.
Canonical source: EvidenceOS/evidenceos-health-platform (edge functions + DB tables + UI)
This project is not yet accepting contributions. Schema fixes and test coverage are prerequisites for contribution readiness. See CONTRIBUTING.md for future plans.
Apache-2.0 — see LICENSE for details.