Test-retest reliability and minimal detectable change of the Cognitive Abilities Screening Instrument in patients with dementia

Background: The Cognitive Abilities Screening Instrument (CASI) is widely used to assess global cognitive function in patients with dementia. It contains nine cognitive domains, namely long-term memory, short-term memory, attention, mental manipulation, orientation, abstraction and judgment, language, visual construction, and list-generating fluency. However, test-retest reliability and minimal detectable change (MDC) of the CASI are largely unknown in patients with dementia, which limits its utility and the explanation of a score change. Purpose: The purpose of this study was to examine test-retest reliability and calculate MDC of the CASI in patients with dementia. Methods: Fifty-two patients with dementia completed the CASI twice with a two-week interval. The frequencies of the scores in the Clinical Dementia Rating (0.5, 1, and ≥ 2) were 38.5, 36.5, and 25.0, respectively. Test-retest reliability was examined using intraclass correlation coefficient (ICC) for the total score and nine domains of the CASI. The MDC was calculated based on standard error of measurement. Results: The ICC value of the CASI total score was 0.97 while the ICC value for the nine domains were 0.65–0.92. The MDC values (MDC%) were 11.6 (12.9%), 2.8 (23.2%), 4.5 (41.2%), 3.4 (42.1%), 4.9 (49.2%), 5.3 (29.2%), 3.4 (28.8%), 2.2 (22.3%), 3.2 (32.1%), and 3.1 (30.7%) for CASI total score, long-term memory, short-term memory, attention, mental manipulation, orientation, abstraction and judgment, language, visual construction, and list-generating fluency, respectively. Conclusion: Our results revealed that the CASI has sufficient test-retest reliability. The MDC values are useful in determining a real change (i.e., improvement or deterioration) between two assessments of an individual patient. However, four domains (i.e., short-term memory, attention, mental manipulation, and list-generating fluency) demonstrated lower ICC values and substantial random measurement errors. Clinicians and researchers should be cautious while using these four domains to explain score changes between repeated assessments of patients with dementia.