- Published
- Updated
An inability to blind the operator in a device investigation does not mean every outcome assessment must be unblinded. Identifying clues that reveal allocation during independent image or event review can help manage bias.
Define the assessment
Define which records reach assessors, in what sequence and with which information removed. Redaction for blinding should not remove information needed for clinical interpretation. Preserve linkage between the authorised original and the assessment copy, with traceability of processing performed on images or documents.
Worked example
For example, images from two devices display different manufacturer headings. Removing a heading accidentally crops the measurement scale. The team checks interpretability instead of considering blinding successful. It produces a corrected copy, identifies whether the old version was read and independently assesses the need to repeat affected readings.
Practical checks
- Map filenames, headings and visible features that could reveal the allocated device group.
- Pilot blinded example copies with a reader to assess clinical interpretability.
- Record accidental disclosure, its timing and the readings potentially affected by it.
Decision and limitations
Acceptance depends on preserving necessary information, not merely removing names. Discovering broken blinding later does not automatically require deletion of data. Justify impact assessment, analytical disclosure and any additional controls through the planned assessment approach.
Frequently asked questions
- Is blinded assessment pointless if operators are unblinded?
- No. Where feasible, independent assessment can address a separate source of bias.
- Does removing a heading prove blinding?
- Not alone. Examine other clues and whether the assessment copy remains usable.