Related Experiment Video
Updated: Aug 2, 2026
Assessment and Communication for People with Disorders of Consciousness
Published on: August 1, 2017
The sensitivity and specificity of three common statistical rules for the classification of post-operative cognitive
M S Lewis1, P Maruff, B S Silbert
1Centre for Anaesthesia and Cognitive Function, Department of Anaesthesia, St. Vincent's Hospital, and School of Psychological Science, La Trobe University, Melbourne, Vic., Australia. m.lewis@latrobe.edu.au
Background:
The application of statistical rules to determine post-operative cognitive dysfunction (POCD) has varied, and partially explains the wide range of reported incidences of POCD in the literature. The current study assessed the sensitivity and specificity of three commonly used statistical rules in a sample of coronary artery bypass graft (CABG) patients and healthy non-surgical controls.
Methods:
Two hundred and four CABG patients [mean age, 68.8 years; standard deviation (SD), 7.0 years] completed neuropsychological assessment pre-operatively (baseline) and 1 week and 3 months post-operatively. Ninety age- and gender-matched non-surgical controls (mean age, 67.8 years; SD, 7.9 years) completed the same tasks at the same time points. POCD was determined in each group using three rules: the 1SD decline on two or more tasks; the 20% decline on 20% of tasks rule; and a modified reliable change index.
Results:
The modified reliable change index demonstrated the greatest combination of sensitivity and specificity. The 20% decline on 20% of tasks rule detected the largest incidence of impairment in the CABG group, but showed large incidences of false positive classifications in the control group. The 1SD rule detected the lowest incidence of POCD in the CABG group, but detected a larger incidence of impairment in the control group.
Conclusions:
The use of the modified reliable change index is recommended, given the sensitivity to change it displayed and the low rates of false positive classification in the control sample. The use of control groups in future research is also recommended.
