Abstract
BACKGROUND: Since its introduction by Hodkinson in 1972, Abbreviated Mental Test Score (AMTS) and its English and other language versions have been widely used in research and clinical practice alike. However, whether the various versions of AMTS yield equivalent information has never been tested. METHODS: We performed cross-sectional assessment of inpatients aged 65+ years with seven AMTS versions and the Mini-Mental State Examination (MMSE) after correction for age and education (MMSEc). We used the MMSEc cut-off score of < 24 as comparator and fitted linear regression models from which we obtained the receiver operating characteristics, and further compared the c-statistics obtained for each version of AMTS. We used Spearman's correlation to check the relation between different AMTS versions. RESULTS: The mean (SD) age of 72 (52.8% women) patients was 76.2 (7.6) years. The average time spent on education was 11.3 (3.5) years. The AMTS score across versions varied between 7.4 (2.0) and 8.2 (1.7). The MMSE averaged 24.1 (4.6) and the MMSEc averaged 25.2 (4.1). We found that the c-statistic across AMTS versions with dichotomised MMSEc as comparator ranged from 0.83 to 0.85 and did not significantly differ from the c-statistic of 0.87 for original AMTS (all p > 0.16). We found AMTS versions to be significantly correlated (all r between 0.83 and 0.99, all p < 0.0001). CONCLUSIONS: We found AMTS to be a reliable and useful tool in the screening for possible cognitive impairment. This seems to be true irrespective of whether we use the original test or any of its studied modifications.