ABSTRACT
This study compared the difference between original and "corrected" neuropsychological test scores at baseline and following 1 year of experience in 17 non-psychology trained examiners. Test protocols were reviewed for errors in instruction, administration, recording, and scoring. Fewer than 3% of the test scaled scores showed a correction of greater than 1 SD. At baseline, individual test scores that changed T-score classification occurred on Digit Symbol, Trails B, and Logical Memory I and II. At one year, significant classification changes remained for Logical Memory I and II (25% and 15%). Scoring of subjective tests remains problematic and centralized re-scoring is recommended.
Subject(s)
Cognition/physiology , Memory/physiology , Neuropsychological Tests/standards , Psychomotor Performance/physiology , Aged , Analysis of Variance , Humans , Male , Middle Aged , Neuropsychological Tests/statistics & numerical data , Psychometrics/methodsABSTRACT
Research study coordinators from 17 sites participating in a cardiac surgery study were trained to administer and score a brief neuropsychological test battery. Results were sent to the study's centralized laboratory for review and feedback. The average examiner errors on the first six protocols were compared with the average errors on the last six protocols over 12 months for each site. Overall, errors for the first six protocols were 4.42, and errors for the last six protocols were 1.83, representing a significant overall decline. Errors for instruction, administration, and recording showed a significant decrease over time. Despite ongoing feedback to examiners, scoring errors did not decline significantly overall; this suggests that a review of all protocols is necessary to achieve reliable scoring. However, when examiners' number of protocols completed was compared with number of scoring errors per protocol, there was a trend for examiners who had completed more protocols to show more improvement in scoring.