Screening for dementia: single yes/no question or Likert scale?

Dementia is now a major public health issue, but the optimal method of screening for at-risk individuals is unclear. Government directive, from the Dementia Commissioning for Quality and Innovation (Dementia CQUIN),¹ advocates a single screening question:Have you been more forgetful in the past 12 months to the extent that it has significantly affected your life?’

In an epilepsy clinic, a population of patients with a high frequency of subjective memory complaints, 48% of patients answered this simple memory screening question in the affirmative, suggesting it would have very low specificity and high probability of identifying false positives.²

Since subjective memory complaints (SMCs) may be associated with subsequent cognitive decline, dementia screening might reasonably be focused on this at-risk population. Screening tests for SMCs are available, such as the five-point Likert scale described by Paradise et al³ in which participants are asked ‘In general, how would you rate your memory?’ with a choice of the following five responses: 1 – poor; 2 – fair; 3 – good; 4 – very good; or, 5 – excellent. The scale defines those rating their memory fair or poor (2 or 1) as experiencing SMCs (SMC⁺).³ This Likert scale has been used as a screen for SMCs in diagnostic test accuracy studies of cognitive screening instruments for mild cognitive impairment (MCI), which in some instances is a prodromal stage of dementia.⁴

We administered the SMC Likert scale to 100 consecutive follow-up patients attending a dedicated epilepsy outpatient clinic. Patients were also asked for any examples of their memory problems. Based on Likert scores, groups were categorised as SMC⁺ (n=20) or SMC^– (n=80) and compared (t tests for continuous variables, chi-squared tests for categorical variables; Table 1).

No difference was found in patient age or seizure type (generalised or partial) between SMC⁺ and SMC^– groups. However, there were significant differences in duration of epilepsy and use of antiepileptic drugs (monotherapy versus polytherapy), with longer disease duration and increased frequency of polytherapy in the SMC⁺ group.

All SMC⁺ patients commented on memory issues. No SMC^– patient with Likert scale scores of 4 or 5 had any memory-related comment, although all SMC^– patients with score of 3 did recount examples of memory problems.

This study found a much lower self-rating of memory impairment using the Likert scale than a previous epilepsy clinic-based study that used a single yes/no screening question (20% versus 48%). The single question study found no statistically significant difference in patient age, seizure type, use of anti-epileptic drugs, or duration of epilepsy (trend) between those answering yes or no to the screening question,² whereas the Likert scale did identify significant differences between epilepsy duration and use of polytherapy between SMC⁺ and SMC^– groups, factors likely to be associated with increasing memory impairment.

Using a Likert scale screening question may provide greater diagnostic discrimination than a simple yes/no question, concordant with evidence that longer cognitive screening instruments have greater diagnostic accuracy.⁵ Therefore, we suggest that policy on dementia screening might explore this approach.

Conflicts of interest

The authors have no conflicts of interest to declare.