OBJECTIVETo evaluate distribution and influence factors of logic memory (LM) modified in assessing and scoring method in normal population and Alzheimer's disease (AD) patients, and definite the cut-off point of the modified scale.

METHODSTotally 183 AD patients, including 118 mild and 65 moderate in degree, 1,417 controls, including 1,283 normal individuals and 134 individuals suffered from other diseases, were recruited in this study. Modified LM was conducted.

RESULTSEducational level (F=354.36, STB=0.46, P=0.0001) was the most obvious factor in demographic data to influence total score in normal control group by a fitting of multiple regression models. The total score increased with the rising of educational level in normal controls (P=0.0001) and other diseases controls (P=0.0001), but not in AD cases (P=0.1365). The total scores were significantly different among normal controls (20.2 +/- 0.2), other diseases controls (17.5 +/- 0.5), mild AD patients (9.6 +/- 0.5) and moderate AD patients (7.1 +/- 0.7) (P=0.0001, P=0.0059), after adjusted educational level, age, sex and rural/urban status by multiple analysis covariance. The sensitivity of cut-off points using modified methods to diagnose AD reasonably increased to 71.98%, while the specificity was 94.11%. According to the sum of long-delayed recall and long-delayed recognition, the sensitivity increased with the rising of educational levels. For education levels at illiteracy, elementary school, junior middle school, senior middle school and above senior middle school, the cut-off points for total score of modified method were 6.5, 9.5, 10.8, 13 and 15.8, respectively, and for sum of long-delayed recall and long-delayed recognition the cut-off points were 5, 6, 8, 9, 10.

CONCLUSIONSWhen modified LM used as a neuropsychological assessment, it is with high specificity, high accuracy and reasonable sensitivity. It is suitable for the diagnosis of AD in early stages, especially for individuals with high educational levels.