MODULE 03 · Evidence

Clinical evidence automation

Begin with one evidence question and a traceable literature-to-analysis workflow. Automate repetitive work while preserving clinical, statistical and ethics decisions for qualified reviewers.

THE FIRST USEFUL PRODUCT

Start with a bounded problem.

Who it is for
Academic research groups, evidence teams and nonprofit trial sponsors.
First product
A cited evidence map, structured extraction sheet and reproducible analysis for one tightly scoped question.
Next decision
The question and review protocol are accepted by a clinical reviewer and statistician.
Product boundary

Automating evidence operations does not clinically validate a treatment. Trial conduct, consent, safety monitoring and regulatory decisions require accountable organisations and experts.

PRODUCT 0 → 1.0

The Phase 0–5 strategy.

Each phase has a concrete deliverable and an acceptance gate. These are proposed development plans, not completed scientific or product validations.

PHASE 0 / 0% MILESTONECurrent planning

Define the evidence question

Choose a question that can be answered with accessible evidence.

Work to do

  • Specify population, intervention, comparator and outcomes with a reviewer.
  • Define eligible evidence and the limits of available data.
  • Set a baseline for review time and extraction quality.

Acceptance gate

The question and review protocol are accepted by a clinical reviewer and statistician.

Publish: Question specification and evaluation rubric.

PHASE 1 / 20% MILESTONEPlanned

Create the source pipeline

Make every extracted claim traceable to its origin.

Work to do

  • Import lawful literature metadata and trial-registry records.
  • Store identifiers, retrieval dates and source versions.
  • Build a manually reviewed extraction reference set.

Acceptance gate

Each reference claim links to supporting text; inaccessible content is marked rather than invented.

Publish: Evidence schema and annotated reference set.

PHASE 2 / 40% MILESTONEPlanned

Build the review prototype

Connect screening, extraction and analysis into one inspectable workflow.

Work to do

  • Add reviewer queues and disagreement resolution.
  • Generate protocol drafts with explicit assumptions and gaps.
  • Run an analysis on public or synthetic data with fixed inputs.

Acceptance gate

A reviewer can trace each report statement to evidence and reproduce the analysis.

Publish: Evidence workspace prototype and example report.

PHASE 3 / 60% MILESTONEPlanned

Evaluate accuracy and bias

Measure performance on evidence the system has not already seen.

Work to do

  • Compare extraction and screening with independent human review.
  • Measure unsupported citations, omissions and disagreement rates.
  • Test study overlap, missing outcomes and sensitivity to selection choices.

Acceptance gate

Predefined quality thresholds pass on held-out questions; failures and uncertainty remain visible.

Publish: Evaluation report and reproducibility bundle.

PHASE 4 / 80% MILESTONEPlanned

Pilot with a research team

Test the workflow in a supervised evidence project.

Work to do

  • Assign named clinical and statistical reviewers.
  • Compare total reviewer effort and corrections against the baseline.
  • Document what still requires manual judgment.

Acceptance gate

The team completes the agreed evidence task with an auditable review record.

Publish: Partner pilot report and reviewer feedback.

PHASE 5 / 100% MILESTONEPlanned

Release the evidence workspace

Support repeatable evidence operations within a defined scope.

Work to do

  • Publish versioned exports, access roles and audit-log behaviour.
  • Document verification, retention and incident handling.
  • Release only the evidence workflows that passed evaluation.

Acceptance gate

Independent replication and reviewer acceptance pass; trial operations remain separately gated.

Publish: Product 1.0 evidence workspace.

Progress stays at 0% until the starting scope is accepted and subsequent milestone evidence is published. Phase 1–5 targets are 20%, 40%, 60%, 80% and 100%. These percentages track development, not treatment effectiveness.

WHAT THIS DEPENDS ON

Build with the right foundations.

  • Datasets & models library: lawful sources and frozen evidence snapshots.
  • Clinical reviewer and statistician: question design, extraction and analysis review.
  • Research sponsor and ethics expertise before any prospective study.

Partner roles above are requirements. No partner participation or endorsement is claimed.

FUTURE INTERACTIVE PRODUCT

Evidence project workspace

Open an evidence-only pilot after citation and extraction checks pass. Patient data and trial operations require a separate approved implementation.

  1. Register the question and search strategy.
  2. Screen and extract with reviewer disagreements visible.
  3. Lock the analysis plan and run a reproducible pipeline.
  4. Review the evidence report and its audit trail.
View the portal specification
PUBLIC CHANGELOG

What changed. What is still planned.

Clinical evidence automation

Phase 0–5 strategy and portal plan published

Published the module strategy, concrete phase deliverables, acceptance criteria and future portal workflow. This is a planning update; product completion remains 0%.

Entries are published with site updates. Planned work is labelled separately from completed product milestones.