INTRODUCTION

Memory problems are almost an early symptom of dementia ⁽¹. This cognitive variable is also considered to be diminished in Mild Cognitive Impairment (MCI)². MCI is considered an intermediate state between normal cognition and dementia. It is characterized by a deficit of cognitive functions, but especially of memory, it should not have any functional impairment in their activities of daily living or meet the criteria for dementia syndrome; as well as understanding that it is very different from normal aging as this process is progressive ³. Not all people diagnosed with MCI develop dementia, but it is associated with a high probability of having that in the future compared to cognitively healthy people, hence its detection and monitoring is important to prevent or delay its onset.

The identification of cognitive impairment at an early age has become an increasingly important challenge for health professionals, estimating its prevalence between 15% and 20% in people within the 40-age group ⁴. The annual rate at which MCI progresses to dementia is 12%, much higher than the 1 to 2% rate found in the cognitively healthy population ^5).

In this line, cognitive impairment was detected in older adults from different regions of Colombia that were living in community, identifying that 40% of the elderly population included in the study presented a result suggesting cognitive impairment. The most affected age range was between 70 and 79 years old (46.1%), followed by individuals between 80 and 89 years old. The prevalence of cognitive impairment was 57.6% in the elderly with arterial hypertension, 48.7% in diabetes mellitus and 56% in stroke ⁶. On the other hand, in Argentina, in the Rosario metropolitan area, it was found that 3.44% of participants between 50 and 90 years of age had mild cognitive impairment. The reached level of education appears as the main predictor of cognitive impairment, above age and gender. Among those participants with incomplete primary education, the potential prevalence reached 10.84% ⁷.

Likewise, studies carried out in 1381 people within urban Lima and in 552 within rural Lima (Cañete), showed that the prevalence of Alzheimer's disease in the first sector was of 9.3% while in the other one was of 6.5%. Additionally, the research on the prevalence of dementia in older adults in a geriatric clinic in Huancayo (central province of Peru) was estimated at 9.9% and the prevalence of mild cognitive impairment at 11% ⁸.

As human beings, we acquired information about objects and living beings that surround us storing at the same time all this information in a process called semantic knowledge, therefore the memory that sustains and allows the processing of this knowledge is the semantic memory. This amnesic aspect stores, processes and recovers information about meaning of words, objects and concepts, as well as the world and its meaning, being different from episodic memory, which stores, recovers and processes information regarding autobiographical facts, within a temporal space axis ⁹. This differentiation was made by Tulving in 1972 when he proposed two subdivisions of declarative memory: semantic and episodic memory ¹⁰. The semantic memory is a mental thesaurus of organized knowledge about words and verbal symbols, concepts and their meanings, as well as the relationships between them. This is a conceptual knowledge based on facts and free context. In contrast to the episodic memory system, semantic memories are probably less susceptible to involuntary transformation and loss of information. In this context, it is necessary to explain this cognitive process referred to as semantic cognition, understood as a collection of neurocognitive mechanisms that are distributed throughout the brain forming a semantic control network and a hub-and-spoke representational network, interacting with neural bases that include temporoparietal areas and prefrontal cortex related to semantic control ¹¹.

Semantic memory is relatively intact in healthy older adult ¹². One study indicates that semantic memory can differentiate healthy people from those with mild cognitive impairment and from those with Alzheimer's dementia ¹³. Besides that, it is reported that semantic memory deficits are frequently found in dementia and distinct patterns of semantic impairment that characterize dementia subtypes, where life course and cultural experiences significantly influence semantic memory ¹⁴. A meta-analytic study reviewing 22 papers indicated that patients with mild cognitive impairment performed in worse conditions than matched controls, concluding that semantic deficits are a key feature of mild cognitive impairment, implying that semantic tests should be incorporated into routine clinical assessment ¹⁵.

Therefore, it is necessary to have the standardized instruments to evaluate this aspect of declarative memory, so they recommend the use of neuropsychological tests of categorical verbal fluency, naming, conceptualization, categorization, general knowledge questions and word definitions, where the batteries include a combination of tasks, which allows a more complex evaluation than the one offered by isolated tasks, thus it is proposed the Semantic Memory Battery (BAMS) ¹⁶.

In addition, it is important also to have a series of instruments to evaluate semantic memory, reviewing the main tests, starting with the description of those that require verbal responses (naming, fluency, definition of categories and semantic analogies) and then those that need a non-verbal response (pointing, drawing, association, among others), referring to individual tests and batteries ⁹. Among these last two batteries stand out: the Cambridge Semantic Memory Battery ¹⁷ and the Evaluation of Semantic Memory in patients with Dementia of the Alzheimer Type (EMSDA) battery¹⁸.

Hence, there is the necessity to evaluate the semantic memory through culturally appropriate tests, in order to help for having an accurate diagnosis of dementia and facilitate cross-cultural collaborative research, with the need for reliable and valid screening instruments to help to identify affected individuals.

Therefore, the objective of this study was to standardize and evaluate the psychometric aspects of the battery in the Evaluation of Semantic Memory in Alzheimer's Disease (EMSEA), which will provide specialists with valid and reliable instruments.

MATERIAL Y METHOD

RESULTS

In order to evaluate the evidences to identify cases of cognitive impairment, it was examining the discriminative capacity of the EMSEA, showing that all sub-tests and the total score separate cases and non-cases, with significant differences, as shown in Table 1.

Similarly, it is observed in areas of verbal fluency and picture naming that present greater standard deviation in the normal sample, unlike the sub-tests of word hearing/drawing matching, definition of semantic categories, attribute recognition, verification of the truth of statements and semantic analogies where the highest standard deviation is presented in the clinical sample.

Table 1. Comparison of clinical and normal sample in older adults with EMSEA battery. 

^Notes:Normal sample= 30; Clinic sample=30; Mean; Standard deviation

The effect size was also calculated for the comparison of the total of EMSEA battery, resulting in a d = 2.94 which indicates that the magnitude of the effect is large and verifies the differences between the clinical and normal groups.

On the other hand, the diagnostic capacity was evaluated through sensitivity and specificity. A cut-off points of 251 was assumed for the total score, showing a sensitivity of 80% and a specificity of 100%, with an area under the ROC curve of 91.7% shown in Image 1.

Image 1. ROC Curve for the total of EMSEA. 

Regarding the criterion validity in table 2, it is observed positive and significant correlations with the memory impairment test (TAM) (r =.52, p =.00); with the mental state examination (MMSE) (r =.38, p =.00); Lawton and Brody's Activities of Daily Living Index (r =.33, p =.00); Barthel's Functionality Index (r =.22, p =.00). It is achieved a negative correlation using the Pfeiffer Brief Mental Status Questionnaire since the instrument is scored in terms of errors (r =-.49, p=.00). Likewise, the effect sizes are high. These aspects show concurrent validity.

Table 2. Correlation between EMSEA total score and tests of cognitive impairment and daily living functionality. 

^Notes:d =.10 low, .30 mean, .50 high

In addition, a correlation matrix was established between the sub-tests and the total score, being the lowest coefficient with Semantic Analogies and the highest with Verbal Fluency, showing positive and significant coefficients (r =.37 a r =.88) indicating that these areas are part of the same construct (see Table 3).

Table 3. Correlation matrix of the sub-tests and the EMSEA total score. 

^Notes:**. The correlation is significant at the 0.01 level (2-tailed).

^*.The correlation is significant at the 0.05 level (2-tailed).

In order to identify evidences of the structure of the semantic memory construct, an exploratory factor analysis was carried out based on the sub-tests of the EMSEA battery. The KMO index had a value of .79 (IC95% =0.744-0.835), Bartlett's test of sphericity showed a Chi-square result of 343.50, p =.00, these aspects indicate the viability of the factor analysis. The eigenvalues greater than 1 would establish two factors giving an accumulated percentage between both of them of 62%, as shown in Table 4, an aspect that is complemented by the sedimentation graph, which can be seen diagrammed in Image 2.

Table 4. Eigenvalues and cumulative variance of the exploratory factor analysis of the sub-tests of EMSEA. 

^>Notes:Initial Eigenvalues of Total; Initial Eigenvalues of the % of variance; Initial Eigenvalues of the % of cumulative

Image 2. Sedimentation graph for EMSEA sub-tests. 

Regarding the components in table 5, it shows that the Varimax rotation established two factors with factorial weights greater than .30; the first one formed by Definition of semantic categories, Recognition of attributes and Verbal Fluency, and the second one formed by Matching word heard drawing, Semantic analogies, Verification of the veracity of statements and Naming of drawings.

Table 5. Semantic Memory Battery Configuration Matrix - EMSEA by means of Principal Axis Analysis with Varimax Rotation. 

^Notes:Extraction method: principal component analysis

Rotation method: Varimax with Kaiser normalization.

^a.The rotation has converged in 3 iterations.

To confirm the model, it was evaluated the fit indexes observing that all of them meet the criteria, with the Relative Fit Index (RFI) standing out with a value of 0.86, as shown in Table 6.

Table 6. EMSEA battery goodness-of-fit indices. 

^Notes:X²: Chi-square, Degrees of freedom in Chi-square; Normalized adjustment index; Tucker-Lewis index; Comparative goodness-of-fit index; Incremental adjustment index; Relative adjustment index; Root mean squared residual approximation.

The goodness-of-fit indices validate the model shown in Image 2, where it is established that the seven sub-tests that are grouped into two factors, as seen diagrammed in Image 3.

Image 3. SEM model of the EMSEA battery. 

Regarding reliability, the Cronbach's Alpha Coefficient was .75, and the Theta coefficient was .80. It is also observed that all sub-tests contribute to measure the construct, giving positive coefficients between .33 and .69, as shown in table 7.

Table 7. Total correlation of EMSEA sub-tests. 

^Notes:Scale mean if the element has been suppressed; Scale variance if the element has been suppressed; Total correlation of corrected elements; Cronbach's alpha if the item has been deleted.

DISCUSSION

For this reason, the standardization process and the evaluation of its psychometric properties were planned. The results showed that the EMSEA presents the ability to distinguish between a clinical and a normal sample, for the seven sub-tests that compose it. In this regard, the choice of a reliable and valid criterion (sufficient, objective and representative of the behavior of interest) is a critical aspect that determines the goodness of the validation process, so it was taken as a strategy to evaluate whether the test is able to differentiate between a clinical and normal sample, establishing if the test extent is able to differentiate between the two groups ²⁸.

The validity of a test can also be evaluated by analyzing its sensitivity and specificity, an aspect which is called decision making validity, specifically indicating that they are associated with the potential of the test to correctly identify cases when they must be assigned to one of the two categories ²⁹. In this sense, the EMSEA presents a specificity of 93.1% and a sensitivity of 92.9%, with an area under the curve of 98%. The result found in the present work established a sensitivity of 80% and a specificity of 100% with an area under the ROC curve of 91.7%, these differences are probably due to the characteristics of the clinical sample, since the previous one included not only patients with Mild Cognitive Impairment, but also a sample of patients with Alzheimer's Disease who present greater involvement of semantic memory ¹⁹.

Within the context of criterion validity, three tests related to evaluation of cognitive impairment were taken as a reference: Pfeiffer's Brief Mental State Questionnaire (examines several cognitive areas, mainly memory and orientation indicating the level of impairment), Mini Mental State Examination (evaluates temporal orientation, spatial orientation, fixation memory, evocation memory, attention, calculation and language) and the Memory Impairment Test (examines immediate memory, temporal orientation memory, semantic remote memory, free recall memory and recall memory with clues ²³.

The results indicated a positive and significant correlation with each one of the instruments, being evident the highest correlation with the TAM. Also, it was linked the functional ability, represented as a criterion, by the Barthel Index and the Lawton Scale, since “The loss of functionality, being this a component of dementia, constitutes in clinical practice the fundamental difference between the concept of physiological or benign cognitive impairment and dementia itself or pathological cognitive impairment “(p.324). Additionally, positive correlations can be observed, these are logically lower than the impairment tests, indicating the evidences of criterion validity ³⁰.

Another important aspect was to establish the construct validity of the EMSEA, which is the main type of validity being an integrating concept, conceptualized in terms of a scientific process of contrasting hypotheses where empirical and rational judgments are combine and the statistical technique used to a greater extent is factor analysis. Likewise, it is established that construct validity seeks to highlight the degree and relation between the items and test components seeing how it configures the construct to be measured and on which the interpretations will be based, seeking to evaluate the dimensionality of the test, with the aim of determining the number of factors of the instruments.

As a result of the exploratory and confirmatory factor analysis, two factors were found: the first one was composed of Semantic Analogies, Verification of Statements, Word Hearing-Drawing Matching and Picture Naming, while the second factor was composed of Verbal Fluency, Attribute Recognition and Definition of Semantic Categories. The first accounted for 46% of the variance and the second for 16%, exceeding the 60% recommended for social sciences. As explained before, the psychometric properties of the battery of EMSEA are evident, and there is no risk of affecting the internal validity of this research. As established, the EMSEA meets the necessary psychometric conditions to evaluate semantic memory.

CONCLUSIONS

The evaluation of the semantic memory in the Alzheimer's Disease differentiates between cases and non-cases, with the sensitivity of 80% and specificity of 100%, with a ROC curve of 91.7% for a cut-off point of 251 of the total score.

The concurrent criterion validity established significant correlations and high effect sizes.

The factor analysis was organized into two components for the 7 sub-tests with adequate fit indices.

The consistency of the Cronbach's Alpha score was 0.75 and a Theta coefficient of 0.80 where all sub-tests contribute to measure the construct.