Background: The Hopkins Verbal Learning Test (HVLT) has been validated for detecting dementia in English-speaking populations. However, no studies have examined the Chinese version of the HVLT scale, and appropriate cut-off scores for dementia in the Chinese population remain unclear.
Methods: 631 subjects aged 60 and over were recruited at a memory clinic at Dongzhimen Hospital in Beijing. Of these, 249 were classified as exhibiting normal cognition (NC), 134 were diagnosed with mild cognitive impairment (MCI), 97 were diagnosed with Alzheimer's disease (AD), 14 met the diagnosis for vascular dementia (VaD), and 50 were diagnosed with other types of dementia, including mixed dementia. The discriminative capacity of the HVLT total learning score, recognition score and total score were calculated to determine their sensitivity and specificity for detecting MCI, AD and other dementias, and various cut-off scores.
Results: HVLT scores were affected by age, education and sex. The HVLT total learning score exhibited an optimal balance between sensitivity and specificity using a cut-off score of 15.5 for distinguishing AD and other types of dementia from NC using the ROC curve, with sensitivity of 94.7% for distinguishing AD and all types of dementia, and specificity of 92.5% for detecting AD and 93.4% for detecting all types of dementias. We stratified the AD and MCI groups by age, and calculated the validity in each age group. In the 50-64 years age group, when the cutoff score was 18.5, the sensitivity of 0.955 and specificity of 0.921 were obtained for discriminating the NC and AD groups, and in the 65-80 years group, and optimal sensitivity and specificity values (0.948 and 0.925, respectively) were obtained with a cutoff score of 14.5. When the cutoff score was 21.5 in HVLT total recall, an optimal balance was obtained between sensitivity and specificity (69.1% and 70.7%, respectively) in distinguishing MCI from NC.
Conclusion: A cut-off score of 15.5 in the HVLT total learning score led to high discriminative capacity between the dementia and NC groups. This suggests that the HVLT total learning score can provide a useful tool for discriminating dementia, but not MCI, from NC in clinical and epidemiological practice.