What is already known on this topic

Dementia prevalence is anticipated to increase with ageing of the population since age is a risk factor for dementia. The diagnosis of dementia in primary care settings has been encouraged in the UK by national directives1 and financial incentives2 in the hope that this will facilitate earlier therapeutic intervention. However, a significant ‘dementia diagnosis gap’ (the difference between expected and observed numbers of cases of dementia diagnosed) remains, perhaps related to underuse of cognitive screening instruments in primary care despite their being specifically designed for assessment of memory complaints in this setting.3 Simple questions that might facilitate early dementia diagnosis in primary care may help to address this diagnostic gap.

Methods of the study

The Caerphilly Prospective Study, a community-based investigation of cardiovascular disease, enrolled 1225 men aged 45–59 years between September 2002 and June 2004 (phase 5 of the study) to investigate the diagnostic utility of simple clinical questions to diagnose dementia. Initial cognitive screening was undertaken using the Cambridge Cognitive Examination (CAMCOG), with a cut-off score <83. Study participants (n=250) comprised 205/300 screen positive (68%) and 45/925 (4.8%) screen negative patients. Study participants were then investigated with both, the index tests (standardised clinical evaluation, neurological examination and items from the Informant Questionnaire for Cognitive Disorders in the Elderly (IQCODE)) and the reference standard (expert consensus diagnosis of dementia based on Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV) criteria). Diagnostic utility for each index item was assessed by calculating standard measures of discrimination, including sensitivity, specificity, positive and negative likelihood ratios (LR+, LR−), diagnostic ORs and area under the receiver operating characteristic curve (AUC), with 95% CIs. Combinations of index items were modelled using logistic regression.

What this paper adds

• Three simple questions were individually found to have high utility for dementia diagnosis: no problems learning how to use new gadgets was useful for ruling out dementia (LR− 0.22, 95% CI 0.11 to 0.43), whereas problems handling personal finances (AUC 0.80, 95% CI 0.72 to 0.88) and with reasoning (LR+ 10.4, 95% CI 5.06 to 21.5) were useful for ruling in dementia.

• Combinations of simple questions had comparable or better diagnostic utility than the mini-mental state examination (MMSE): models comprising problems with personal finance and either planning (AUC 0.92, 95% CI 0.86 to 0.97) or reasoning (AUC 0.86, 95% CI 0.80 to 0.93) had better diagnostic utility than MMSE at the cut-off ≥24/30 (LR− 0.28; LR+ 2.14; AUC 0.72).

Limitations

• The study sample comprised only middle-aged men and was enriched for those with cognitive impairment (scoring <83 on CAMCOG).

• The reference standard was subject to incorporation bias (circularity), a common problem with clinical definitions of dementia.

• Since dementia prevalence increases with age, and is slightly more common in women than in men, the age and gender structure of the patient sample recruited in this study makes the generalisability and transferability of the results uncertain, as do the selection procedures that enriched the sample for patients with cognitive impairment.

What next in research?

The Caerphilly Prospective Study was not a diagnostic test accuracy study, so prospective studies of these simple questions in independent patient cohorts in primary care will need to be designed to corroborate the findings. Whether the same simple questions also help in the diagnosis of mild cognitive impairment versus subjective memory complaints, a more challenging differential diagnosis than dementia versus no dementia, will also need to be examined, since the diagnosis of mild cognitive impairment may be more critical for early intervention and treatment of cognitive decline.

Do these results change your practices, and why?

Possibly, but results need to be confirmed in other studies. Questions about functional activities of daily living (eg, ability to use a telephone and public or private transport, handling personal finances and medications) have already been found to be predictors of dementia in community-based studies.4 The current study adds to the limited evidence base for using simple questions in screening for cognitive impairment.5 The specific questions identified may have greater utility than a previously suggested single screening question,1 which may have very high sensitivity, thus avoiding underdiagnosis of cases but with many false positives.6