2005 HSR&D National Meeting Abstract
3075 — Screening for Dementia in Primary Care: Validity of Brief Screeners
Williams JJ (DVAMC)
Holsinger T (DVAMC)
Coffman C (DVAMC)
To evaluate brief screening instruments for dementia in a primary care setting.
A random sample of consenting Veterans >=65 years old with a primary care appointment were screened with the mini mental state examination (MMSE) and the Mini Cog (MC). Criterion standard evaluations were performed by a geriatric psychiatrist, blind to screening results, in all veterans screening positive and a random sample of those screening negative. Criterion standard evaluations consisted of a neuropsychiatric battery, a semi-structured medical history, a medication and medical record review, and a structured informant interview. Based on the criterion evaluation, DSM-IV criteria were used to classify patients as demented, mildly cognitively impaired or normal. To evaluate physician recognition rates, the veteran’s primary care provider (PCP), blind to screening results, assigned the patient to one of the diagnostic categories. Screening instruments and PCP diagnosis were compared to the criterion standard interview, adjusting for the sampling strategy.
Of the 172 eligible veterans approached, 143 consented and were screened; 46 completed the criterion standard evaluation. The study sample is 97% male, 75% white and mean age 73.7 with standard deviation 5.7. Prevalence rates are: dementia (20%, 95% CI 14% to 29%), and mild cognitive impairment (35%, 95% CI 27% to 45%). The MMSE at a cutpoint of 23/24 has an unadjusted sensitivity for dementia of 42% (95% CI 15% to72%) and specificity of 100% (95% CI 89% to100%); adjusted sensitivity was 20% (95%CI 5% to 57%). Adjusted specificity remains 100%. The MC at a cutpoint of 2/3 has an unadjusted sensitivity of 58% (95% CI 28% to 85%) and specificity of 85% (95% CI 69% to 95%). The adjusted sensitivity for the MC is 30% (95% CI 10% to 65%]; adjusted specificity was 96% (95% CI 85% to 100%). The response of the PCP to questions on the patient’s dementia status has an unadjusted sensitivity of 42% (95%CI 15% to 72%) and specificity of 94% (95% CI 80% to 99%).
The prevalence of dementia and mild cognitive impairment is substantial in older veterans. Brief screening instruments are highly specific but insensitive.
As the proportion of older veterans increases, cognitive impairment will become an increasingly important clinical problem in the VA. Improved screening strategies are needed to accurately identify veterans with dementia.