Revising the ADAS-cog for a more accurate assessment of cognitive impairment

Background: To examine whether it is appropriate to sum the cognitive part of the Alzheimer Disease Assessment Scale (ADAS-cog) items to assess cognitive impairment. This assumes items to have (1) equal measurement precision and hierarchically ordered categories. Methods: Rasch analysis on the basis of pooled data from 3 Randomized Controlled Trials was used to examine these assumptions and to estimate each patient's level of impairment. Analyses were replicated in an independent sample. Results: The original ADAS-cog scoring did not fit the Rasch Model and did not reliably distinguish between impairment levels. Patients with equal test scores had different impairment levels. Similarly, patients with different test scores could have the same impairment level. Revising the ADAS-cog by (1) weighting the items by their measurement precision and (2) collapsing nonhierarchical item categories resulted in good fit and a valid one to one correspondence between sum scores and estimated impairment levels. This revealed that equal differences in ADAS-cog scores did not reflect equal differences in impairment level along the test's score range. Conclusions: It is appropriate to summate the ADAS-cog items provided that the items are weighted and have their categories hierarchically ordered.