The CASRAI Dictionary: A Research Administration Glossary

Observer Bias

Observer bias is a systematic distortion introduced by the person doing the observing, recording or interpreting, rather than by the participants or the sampling. The Catalogue of Bias defines it as a 'systematic difference between a true value and the value actually observed due to observer variation'. It arises wherever measurement requires judgement - reading an image, scoring a scale, coding an interview transcript, grading a lesion, deciding whether an adverse event is related - and it is strongest where the observer knows which group a participant belongs to and has an expectation about what they should find. It is a form of detection or measurement bias and belongs to the recording end of a study, not the sampling end: it is distinct from selection bias, which distorts who enters the study, and from confounding, which distorts the comparison itself. It is also distinct from the observer effect, where the participant changes behaviour because they know they are being watched; in observer bias the participant behaves normally and it is the observer's record of that behaviour that is wrong. Observer bias is controlled at the design and procedure level - by blinding outcome assessors, separating the people who know exposure status from the people who record outcomes, pre-specifying operational definitions and coding rules, using independent duplicate coding, and quantifying agreement between observers - rather than by any statistical adjustment after the fact.

Regression to the Mean

Standardized Infection Ratio (SIR)

The Standardized Infection Ratio (SIR) is an NHSN risk-adjusted summary measure calculated as the number of healthcare-associated infections observed at a facility or unit divided by the number statistically predicted for that same population, using a risk-adjustment model fit against a national NHSN baseline. SIR = 1.0 means observed infections matched the risk-adjusted prediction; SIR > 1.0 means more infections were observed than predicted; SIR < 1.0 means fewer were observed than predicted. NHSN reports a 95% confidence interval alongside the point estimate because small device-day or procedure denominators make the SIR statistically unstable at the unit level.

Beyond-Use Date (BUD)

A beyond-use date (BUD) is the date, or date and hour, established by a compounding pharmacist after which a compounded preparation must not be used, stored, or transported. It is set at the time of compounding and is distinct from a manufacturer's expiration date, which applies only to the original, unopened, commercially manufactured product. Under USP General Chapter 797 (sterile preparations, current since November 1, 2023), the BUD is determined primarily by the CSP Category the preparation was compounded under (Category 1, 2, or 3, reflecting the degree of verified environmental control), together with the storage condition (controlled room temperature, refrigerated, or frozen), the dosage form, and any supporting sterility or stability testing or applicable stability data. Under USP General Chapter 795 (nonsterile preparations, also current since November 1, 2023), the BUD is determined primarily by whether the formulation is aqueous or nonaqueous, the dosage form and packaging, the storage condition, and whether supporting stability data exists — the chapter provides conservative default periods to use when no such data is available. In both chapters, the current chapter text and the USP Compounding Compendium contain the exact, category-specific and formulation-specific tables a compounder should use to assign a BUD to a given preparation; a value copied from a pre-2023 reference will not reflect the current framework.

EU AI Act

Regulation (EU) 2024/1689, the European Union's law regulating AI systems and general-purpose AI (GPAI) models placed on or used in the EU market. It classifies AI systems into four risk tiers (unacceptable, high, limited, minimal) with obligations scaling to risk, applies a separate GPAI-model regime, and phases in on a multi-year timeline set by Article 113 and later revised by the 2026 'Digital Omnibus' amendment.

CSE (Council of Science Editors)

CSE (the Council of Science Editors) is an international individual-membership society for editorial professionals across the sciences, founded in 1957 as the Conference of Biology Editors, incorporated in 1965 as the Council of Biology Editors and renamed Council of Science Editors on 1 January 2000. It publishes The CSE Manual (titled Scientific Style and Format through its eighth edition, 2014; ninth edition, University of Chicago Press, 2024), the source of the three CSE documentation systems -- citation-sequence, citation-name and name-year -- and the Recommendations for Promoting Integrity in Scientific Journal Publications, first issued in 2006 and updated on a rolling basis since May 2018.