Variables that allow a reliable classification of older people with different levels of cognitive state

Abstract

To assess the general cognitive state and identify potential cognitive deterioration issues, screening tests such as the Mini-Mental State Examination have been widely utilized. Various studies have aimed to determine the socio-demographic variables (e.g., age, education) and cognitive abilities (memory, language, executive functions) most closely linked to the cognitive state assessed through tests like the MMSE. The primary objectives of this study were as follows: (a) assess the impact of socio-demographic variables, such as age and cognitive reserve, and other cognitive abilities (working memory, comprehension of written sentences) in accurately classifying a sample of older individuals with varying general cognitive statuses; (b) calculate optimal cut-off points for variables with the greatest importance in classification, striking a balance between true positive rate (sensitivity) and false positive rate (1 - specificity). The participants comprised 159 Spanish older adults, aged 60 to 89, categorized into two groups based on their 35-item MMSE scores: those with scores equal to or greater than the 60thile (normal/high scores: N/Hs group) and those with scores equal to or lower than the 25thile (low scores: Ls group). All participants underwent tests evaluating working memory and comprehension of written sentences, including the digit reordering test, the sequential version of the ECCO-senior test, and the written sentence comprehension test of the Batería de Evaluación de los Trastornos Afásicos (BETA; English translation: Battery for the assessment of aphasic disorders). Cognitive reserve estimation was obtained through Rami et al.'s Cognitive Reserve Questionnaire. Binary logistic regression analysis was initially conducted following a hierarchical method to identify significant variables explaining correct classification. Subsequently, ROC curve analyses were performed to determine optimal cut-off points for relevant variables, as well as measures of overall model quality. The final logistic equation incorporates cognitive reserve, digit reordering, and performance on BETA’s sentences focused on the object and on sentences with one proposition not fitting canonical word order in Spanish in the ECCO test. Area under the curve (AUC), ROC and precision/exhaustivity curves, an overall model quality index, and optimal cut-off values were computed for all these significant variables. Results are discussed in the context of the reviewed literature.