Data cuts and the de-identified export

Freeze the data for an analysis with its fingerprint; a copy for the statistician without codes, free text and real dates.

A data cut freezes all the study's data at one moment: for an interim analysis, an annual registry report or a publication. The data in the study goes on being entered and corrected, but the cut never changes. A year later you download the same cut and get the same numbers, and its SHA-256 fingerprint proves that nobody touched it.

How to make a cut

Open Data cuts in the study menu (or the link on the Export page).
Name the cut, for example “Interim analysis 1” or “Registry report 2026”, and say why if you like: “statistical analysis plan, section 9.2”.
Click Freeze the data. The cut holds every site and every form except the forms for the unblinded team. Its number, time, eCRF version and SHA-256 go to the audit trail.

Cuts are made by the project administrator and data managers with access to all sites. A cut cannot be changed or deleted: when other data is needed, make the next cut.

Download a cut

Every cut has two buttons: As it is (for the project administrator and data manager) and De-identified (for them and for the statistician). Both in Excel or CSV, in the layout of the usual export: a sheet per form, the dictionary of variables, the events sheet. The last Excel sheet and README.txt in the CSV name the cut, its time and its SHA-256. Every download shows in the export history and the audit trail.

The Builder button opens the export builder on the cut: the forms and variables you need, a row per patient, SPSS, Stata, SAS or R. The statistician gets the cut de-identified there too.

The Check button computes the SHA-256 of the stored copy again and compares it with the one recorded at the cut.

The de-identified download

Pseudonyms. The patient code becomes S0001, S0002… in an order unrelated to enrolment. Pseudonyms are the same in every cut of the study, so the statistician can join cuts.
Shifted dates. All dates and times of one patient move by the same number of days, 1 to 30 forwards or backwards. Intervals (enrolment to visit, duration of an adverse event) stay exact. An incomplete date (UK.03.2019) keeps its year only.
Age. The birth year becomes the age at enrolment; 90 and older is written as 90.
No free text. Short and long text fields, the note on the patient's status, the reasons of visits that did not take place and all queries go. The file itself lists the removed variables.

The key for pseudonyms and shifts stays in the study's database and never goes into a file.

The Statistician role

Invite the statistician to the team with the Statistician role. They see the study overview and the cuts page and download de-identified cuts only. They see no patients, forms, queries or the usual export.

Consents. When the study keeps informed consents, the cut holds the Consents sheet too: the version, the date, the scope and any withdrawal of every patient. In the de-identified copy its dates move like the patient's other dates.

A withdrawal after the cut. When a patient withdraws the use of their data after a cut, every later download of that cut leaves them out, in Excel, CSV and the builder alike. The stored cut and its SHA-256 do not change, and the Check button still confirms them. The cuts list shows how many patients were left out, the “About the file” sheet and README.txt say it in the file, and the audit trail records it with each download. Files downloaded before the withdrawal are not called back: whether the data already passed on must be destroyed follows your organisation's procedure.

Pseudonyms. A patient gets their pseudonym at the first de-identified download and keeps it in every later cut, whoever joins the study in between, so the statistician can join cuts.

About the law

De-identified data lowers the risk of recognising a patient, but under Russian law (152-FZ) it is still personal data. Share it only under an agreement with confidentiality terms. If a form has a field that could point to a patient (a rare diagnosis at a small site, an exact workplace), discuss it with the statistician before sharing.