Putting the Alzheimer's cognitive test to the test I: Traditional psychometric methods

Jeremy Hobart, Stefan Cano, Holly Posner, Ola Selnes, Yaakov Stern, Ronald Thomas, John Zajicek

Research output: Contribution to journalArticlepeer-review


Background: The Alzheimer's Disease Assessment Scale - Cognitive Behavior section (ADAS-Cog) is the most commonly used cognitive test in AD clinical trials. However, there are concerns about its use in early-stage disease. Herein we examine those concerns using traditional psychometric methods. Methods: We analyzed ADAS-Cog data (n = 675) based on six psychometric properties: data completeness; scaling assumptions; targeting; reliability; validity; and responsiveness. Results: At the scale-level, criteria tested for data completeness, scaling assumptions (item total correlations 0.33-0.59), targeting (no floor/ceiling effects), reliability (Cronbach's α = 0.74), and validity (correlation with MMSE = -0.70) were satisfied. Responsiveness (baseline to 12 months; n = 145) was moderate to high (effect size = -0.73). However, 8 of 11 ADAS-Cog components had substantial ceiling effects (range 32%-83%), and decreased responsiveness associated with low to moderate effect sizes (0.14-0.65). Conclusion: In our study, many patients with AD found large portions of the ADAS-Cog too easy. Future research should consider modifying the ADAS-Cog or developing a new test.

Original languageEnglish (US)
Pages (from-to)S10-S20
JournalAlzheimer's and Dementia
Issue number1 SUPPL.
StatePublished - Feb 2013
Externally publishedYes


  • Alzheimer's disease
  • Clinical trials
  • Psychometrics
  • Reliability
  • Validity

ASJC Scopus subject areas

  • Epidemiology
  • Health Policy
  • Developmental Neuroscience
  • Clinical Neurology
  • Geriatrics and Gerontology
  • Psychiatry and Mental health
  • Cellular and Molecular Neuroscience


Dive into the research topics of 'Putting the Alzheimer's cognitive test to the test I: Traditional psychometric methods'. Together they form a unique fingerprint.

Cite this