AD8 cut-offs for mild cognitive impairment and dementia help judge baseline cognition in delirium

Design

Cross-sectional psychometric linking study (equipercentile and IRT)

Population

264 hospitalised older adults (mean age 79.5) and their proxies

Primary outcome

AD8 scores equivalent to IQCODE cut-offs for MCI and dementia

Effect

AD8 2 = IQCODE 3.2 (MCI); AD8 4 = IQCODE 3.5 (dementia)

When an older inpatient is delirious, direct cognitive testing cannot tell you what their thinking was like before admission. Informant questionnaires can. The 16-item IQCODE has established cut-offs for mild cognitive impairment and dementia; the shorter 8-item AD8 had only one, for any cognitive impairment.

In 264 hospitalised older adults, mean age 79.5, both questionnaires were completed by proxies (JAGS, 24 September 2026). Using two linking methods, equipercentile and item response theory, the investigators built a crosswalk between the scales. An AD8 score of 2 corresponded to an IQCODE of 3.2, the threshold for mild cognitive impairment, and an AD8 of 4 corresponded to an IQCODE of 3.5, the threshold for dementia.

The AD8 takes a family member a few minutes. With two cut-offs, it can now help separate a patient who probably had dementia before the delirium from one who had milder impairment, which shapes discharge planning, capacity assessment and follow-up.

• Ask a family member or carer to complete the AD8 when a delirious older patient cannot be tested directly.
• Consider an AD8 of 2 or more as suggesting prior mild cognitive impairment, and 4 or more as suggesting prior dementia.
• Record the informant score in the notes as the pre-admission cognitive baseline.
• Retest cognition directly once delirium has resolved, before making a formal diagnosis.
• Use the result to plan discharge support and follow-up in memory services.

Why it matters

Knowing whether dementia preceded delirium changes discharge planning, but direct testing is impossible at the time it matters most.

Don't overread it

The cut-offs are derived from one sample of 264 patients and map to IQCODE thresholds; they have not been validated against clinical diagnosis here.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.