Detection of Rater Errors on Cognitive Instruments in a Clinical Trial Setting

Abstract

Objectives: This study examines errors committed by raters in a clinical trial of a memory enhancement compound.

Background: Findings of clinical trials are directly dependent on the quality of the data obtained but there is little literature on rates or nature of rater errors on cognitive instruments in a multi-site setting.

Design: Double-blind placebo-controlled study.

Setting: 21 clinical sites in North America.

Participants: Two hundred seventy-five participants.

Measurements: MMSE, WMS-R Logical Memory I and II, WMS-R Verbal Paired Associates I, WASi Vocabulary, WASi Matrix Reasoning, GDS and MAC-Q.

Results: The WMS-R Logical Memory I and II and WASi Vocabulary tests were found to have the greatest number of scoring errors. Few substantive errors were detected on source document review of the MMSE, GDS, MAC-Q and WMS-R Verbal Paired Associates I. Some additional administration and scoring issues were identified during feedback sessions with the raters.

Conclusions: Cognitive measures used in clinical trials are prone to errors which can be detected with proper monitoring. Some instruments are particularly prone to inter-rater variably and should therefore be targets for focused training and ongoing monitoring. Areas in need of further investigation to help inform and optimize quality of clinical trial data are discussed.