Overview
How many patients and operations exist at every step, from a record to an analysable operation, and what has to happen before a first study can exist.
Research workspace
Learn from every journey.
The knee-replacement data patients have agreed to share, with its sources and its gaps visible. Open since 2 October 2026; the dataset is empty until the first patients enroll, and the workspace says so.
What it holds.
Data catalog
Every variable collected: its definition, whether it is a patient’s word or a document’s word, how it is exported, and how many we actually have.
Cohorts
Search and compare groups, trace each fact to a patient report, a practice report or a reviewed document, and save a definition with a version number.
Source review
Read the operative note or implant card and record what it says, field by field: agrees, corrected with evidence, not enough documentation, or information requested.
Studies
The question, population, endpoints, plan and oversight references, tied to a saved cohort.
Releases
A frozen, hashed copy of a study’s dataset with a manifest, dictionary, CSV and JSON files and a starter script that verifies the hash and prints the first results.
Who may enter.
Three roles, granted by the Operations administrator with a written reason and an audit record. A patient, surgeon or Guide account grants nothing by itself.
Research analyst
Catalog, completeness, cohorts, aggregate reports and release files.
Source reviewer
Source documents and field-by-field review decisions.
Research investigator
Everything an analyst may, plus studies and dataset releases.
On what terms.
- Only records with current research permission are read. Withdrawal stops every later read and is shown against any release that contained the record.
- The data are restricted and coded, not anonymous. No de-identification determination has been made.
- Means and rates are withheld under five distinct responding patients.
- A patient’s report, a practice’s report and a document’s word stay separate; disagreements stay visible.
- Every search, read, release and download writes an audit row before anything is returned. Search terms are hashed.
- Comparisons are descriptive. Nothing here establishes that a surgeon, implant or technique caused a difference.