A Preliminary Study on the Correlation Factor Analysis of Language Cognitive Assessment System Based on Scale Construction

This paper explores the 12 dimensions of language cognition assessment system using factor analysis. Principal component analysis and correlation analysis of factors within the system were conducted on the language cognition assessment system. Factor analysis and principal component analysis: KMO: 0.934, weak bias correlation, suitable for factor analysis, and the contribution of principal component variance was 79.837%, respectively. The correlations between the 12 factors ranged from 0.611-0.903. The correlation values between the remaining 11 factor assignments and the total score of the system were higher than the correlations between the subscales. The system is constructed with good structural validity, but the correlations between the factors are strong. It is suggested that the indicators be combined or the form of the questions be modified to provide guidance for the modification of the language cognition assessment system.


INTRodUCTIoN
As the aging population increases, the prevalence of dementia is increasing each year, and cognitive impairment can occur to varying degrees after brain injury.Mild cognitive impairment is an intermediate state between normal aging and dementia, so there is a strong need for early cognitive screening and intervention for healthy older adults and post-stroke survivors.Current diagnoses of cognitive impairment are mostly based on neuropsychological scales, one of which is a simple structured scale such as the MMSE and Moca, but related studies have shown that the MMSE has low sensitivity, while the MoCA is designed for conceptual transformation in verbal abstraction and has a large difficulty factor for each subscale in the cognitive domain, which is not adapted to severe cognitive impairment (Pendlebury, Mariz, Bull, Mehta, & Rothwell, 2012).One category is the complex and exhaustive set of neuropsychological tests, such as the Wechsler Adult Intelligence Scale (WAIS), the Halstead-Reitain Neuropsychological Inventory, and the Wechsler Memory Inventory, which are highly demanding and time-consuming for personnel.Based on this, we designed a language cognition assessment system that can quickly screen people with cognitive impairment based on literature screening, scale construction, and Delphi expert scoring of language cognition entries based on the model assignment of question selection rules and answer trajectories for the characteristics of Chinese language and culture, and now we are going to analyze and study its system internal about language and cognition from principal component analysis, correlation of factors within the system and language and cognition The analysis and research on the correlation between language and cognition from principal component analysis, correlation of factors within the system, and correlation between language and cognition and synthesis are conducted to provide guidance for the modification of the language cognitive assessment system.

Research Subjects
A total of 78 participants in this study.(Interns and chaperones as normal subjects 24, 12 females, 12 males, aged 38.71years; stroke and traumatic brain injury language cognitive impairment 54, 21 females, 33males, aged 55.55 years years.) Subjects with language cognitive impairment were obtained from patients hospitalized in the Department of Rehabilitation of Jinan University and the Department of Rehabilitation Medicine of the Second Hospital of Guangzhou Medical University from 03/2018 to 03/2020, and enrolled cases including cerebrovascular accidents and brain injuries were confirmed by CT or MRI and diagnosed by specialists.Normal subjects were obtained from undergraduate interns (20-25 years old) in the Department of Rehabilitation of Jinan University and the Second Affiliated Hospital of Guangzhou Medical University.
The following conditions were excluded: (i) patients in a coma, coma, or persistent vegetative state; (ii) those with a history of psychiatric disorders or congenital mental retardation; (iii) those taking drugs with cognitive effects; (iv) patients with severe visual or hearing impairment; (v) those with alcohol dependence; (vi) those with severe visual or hearing impairment; (vii) those who had to stop the test due to adverse reactions, changes in condition, or concomitant other diseases during the test and could not complete it.

Assessment Tool: Language Cognition Assessment System
According to the literature screening, the top three language and cognitive scales were listed, and the first-level indicators of the scales were constructed using the Delphi method.After the expert questionnaire survey, 10 first-level indicators were finally screened and 12 factors were included, and the language dimensions included: spontaneous expression, retelling, reading, figure name, and listening comprehension; among them, the second-level entries of listening comprehension were: listening whether, listening recognition.Cognitive dimensions include: orientation, memory, calculation, attention, and reasoning.The secondary indicators of memory are divided into: instantaneous memory, delayed recall.The secondary and tertiary indicators were modified according to expert opinions, and the clinical tests were adjusted with corresponding questions and guiding phrases, and the scale was simulated and computerized after no objection (Wang, Chang, & Douglas, 2012).Each factor was divided into five levels according to difficulty (10, 30, 50, 70, and 90 points).
Assessment process: (i) Talk to the subject, perform an initial clinical assessment, and inform him/her of the significance of the assessment before entering the test.‚The assessment was conducted in a quiet assessment room dedicated to the department.The testers were professionally trained rehabilitation physicians and therapists, and the test was conducted using uniform and standardized instructional language, avoiding hints and other questions and answers as much as possible, and when the test was ready, the testers were told to "prepare for the assessment, please answer the questions carefully.Based on the results of the test, the initial screening questions, the difficulty level of the questions, the instructional phrases and the revised presentation of the questions are adjusted.
After the subjects answer the questions, they can get their score and trajectory under the factor.The score is assigned to each factor using the existing scoring model.

.1 KMO test and Bartlett test
Using factor analysis, Bartlett's spherical test was conducted to find the KMO value, whether it was suitable for factor analysis, and to observe the cumulative contribution of each dimension to the total variance.

Correlation Analysis
Correlation analysis was conducted on the 12 factors, and the structural validity of the system was measured based on the correlation analysis of each subscale with the total score (Schulze, Hoffmann, Kroke, & Boeing, 2003), and whether the system had good structural validity was judged based on whether the correlation was significant.Correlation analyses were also conducted for the linguistic dimension, cognitive dimension, and integrated linguistic-cognitive system (Gerhardsson, Oras, Mattsson, Blomqvist, & Rosenblad, 2020;Schulze et al., 2003).

Statistical Analysis
Factor analysis of the data was performed using SPSS 16.0 statistical software.Count data were expressed as percentages and measurement data were expressed as (x±s), and the chi-square test was used for gender comparisons and t-test for age comparisons.

Baseline Characteristics of the Study Subjects
The differences between the normal group and the language cognitive impairment group were not statistically significant when comparing the clinical data of gender at P > 0.05 and were comparable.However, the differences in age were statistically significant at P < 0.05 for both the normal and impaired groups (Wouters et al., 2010), and the mean age was older for the impaired group.See table 1.

KMo Test and Bartlett's Test
Factor analysis and principal component analysis: KMO: 0.934, weak bias correlation, suitable for factor analysis (Hasan et al., 2014) Approximate chi-square distribution value of Bartlett's spherical test is 1212.715,degree of freedom is 66, p=0.000, statistically significant difference, suitable for factor analysis.See Table 2.

Total Variance Interpretation and Common Factor Variance
The cumulative contribution of the principal components to the total variance was 79.837%, as shown in Table 3. the variance of the common factors is shown in Table 4.The factors with eigenvalues > 1 were selected as in Figure 1 to obtain 1 principal component and factor loading matrix.It can be seen that the principal components contain at least 69.2% of the information of each original variable, the information loss of each variable is less, the 12 factors can extract most of the information amount of the study items, and the principal component factor analysis is better (Hasan et al., 2014).

Factor Correlations
Correlation analysis of the data from the 78 clinical samples showed that, with the exception of delayed recall and memory, the remaining 11 dimensional assigned scores had higher correlation values with the total system scores than the correlations between the subscales (Krahn, Horner-Johnson, Hall, Roid, & Peterson, 2014), and the correlations between the scores of language and cognition and the total system were higher than the language and cognition dimension Correlations.This indicates that the correlations between the factor dimensions and the overall language and cognition assessment system are good and that the system has high construct validity (Hasan et al., 2014). Tm

dISCUSSIoN
The results of factor analysis showed that factor analysis and principal component analysis: KMO: 0.934, with weak bias correlation, are suitable for factor analysis (Hasan et al., 2014).According to the results of principal component analysis and gravel plot, there is only one factor that needs to be extracted for the language cognition assessment system: spontaneous expression, but spontaneous expression belongs to one factor of the language dimension, and the previous language subitems (Francesmonneris, Pincus, & First, 2013) include listening comprehension, expression, written comprehension, naming, writing, retelling, structure and visual spatial, and use components.The cognitive subsets (Guo, Cao, Zhou, Zhao, & Hong, 2010) include subsets of cognitive functions such as memory, language, visuospatial, computational, and thinking generalization skills.The factors of the computer-based linguistic cognitive system were determined using the Delphi process (Christiansen, Filippou, & Filippou, 2010) with reference to the application trends of language and cognitive scales at home and abroad, screening the primary indicators (dimensions) and secondary indicators (items) of the computer-based linguistic cognitive assessment system, including 15 items for language and 20 items for cognition, Scirè, Balint, & Terslev, 2021), which identified six linguistic and cognitive dimensions, respectively (Guan, Han-Tian et al., 2020).According to the rules of question selection and the results of expert scoring, the tertiary indicators (topics) were prepared.The dimensions, entries and topics of the assessment system were screened, modified and optimized through questionnaire consultation and clinical test validation by experts.The correlation coefficients of the language cognitive system and the 12 factors found in figure 2 were tested to be high.The correlation coefficient of language and cognition in the correlation analysis of language, cognitive system and integrated language cognitive system was 0.953, and the correlation coefficient of language and cognition and language cognitive system was 0.986, which proved that the entries of language and cognitive system correlated higher with the entries of the overall language cognitive system (table 5).The delayed recall and memory correlations were higher than the correlation values with the overall system scores.This is because the delayed recall and memory questions are exactly the same, only tested 5 minutes apart (Glassmire, Ross, Kinney, & Nitch, 2016).The correlation values between the remaining 11 dimensional assigned scores and the total score of the system were higher than the correlations between the subscales, indicating high construct validity (Krahn et al., 2014).However, the correlations between some of the factors were greater than 0.85, suggesting poor identification of factor entries and suggesting merging or modifying the form of topic presentation.For example, the correlations between spontaneous expression and orientation, transient memory, and delayed recall; orientation and computation, and delayed recall; computation and retelling, reading, and memory; and transient memory and delayed recall were all greater than 0.85, and the following reasons were considered: on the one hand, it is possible that correlations greater than 0.85 were consistent with the linguistic neuropsychological pathway at the time of question presentation; in addition, linguistic cognition is inextricably linked (Yi, Chen, Chang, Wang, & Wu, 2018), transient memory is highly correlated with delayed recall, as the measured are memory topics, and the memory capacity and length are consistent.The computational factor was examined by asking the subjects to listen and say the corresponding answer, and the computational questions covered the expression and memory-related factor components.In the directional questions, the subjects were asked to listen to the test and say or point out the corresponding answers, and the directional questions covered the components related to the retelling, expression, reasoning, and memory factors.Therefore, some of the factor correlations were higher than 0.80.Secondly, the factor correlations were higher, and it is suggested that the indicators be combined accordingly (Hasson, Keeney, & McKenna, 2000) or that the questions be adjusted to the language neuropsychological pathway to reduce the factor correlations appropriately.The above results indicate that the entries of the self-administered system have retention value, are reasonably reliable, and some factors need adjustment and topic optimization.The assessment system was effective in assessing cognitive function for each subscale and total score, suggesting that the Language Cognitive Assessment System is a more effective assessment system for assessing language cognitive dysfunction.Several studies have confirmed the presence of cognitive dysfunction in patients with aphasia and the existence of a valid interrelationship between language and cognitive functioning.Their relationship is mutually reinforcing and interactive, with language impairment being an effective activity and interplay of diminished cognitive processes that support language behavior (Vallilarohter & Kiran, 2012).These cognitive processes include attention, executive functioning, visuospatial ability perception, reasoning, and visual memory.For example, a study conducted by seniow (Seniów, Litwin, & Le?Niak, 2009) and others on cognitive functions in patients with aphasia confirmed that cognitive functions closely related to language, such as visuospatial working memory, attention, and thinking and reasoning, were reduced and impaired to varying degrees in patients with aphasia.We used the Delphi method to evaluate the first-level indicators of the language cognitive system (language cognitive domain) by integrating the language cognitive level screening indicators in the Delphi expert rating process, and used neuropsychological pathways to give questions.Our linguistic-cognitive system can be used as an evaluation system for applications in the linguistic and cognitive domains.
This study we provide new pilots and ideas for computerized assessment in the direction of language and cognitive domains.With the development of computers, virtual reality and humancomputer interaction are increasingly used in the field of rehabilitation.Due to the diversity and richness of the computer interface can increase the motivation and participation of patients and bring more positive and effective rehabilitation results.At the same time, it can replace many of the drawbacks of traditional assessments in the verbal and cognitive domains, such as inconvenience of administration, non-reproducibility, and high staff skill requirements (Faria, Andrade, Soares, & Badia, 2016;Penn, Rose, & Johnson, 2009).Research has shown that cognitive function is dependent on a network of brain regions, and lesions in one region may cause disruptions to the network, leading to multiple cognitive dysfunctions (Faria et al., 2016), that focusing on training a single cognitive function may not be effective, that several cognitive abilities are often required to complete tasks in daily life, and that a linguistic cognitive system can train cognitive functions in multiple dimensions.Studies have shown that rich photic stimulation can can promote cortical remodeling and facilitate the reconstruction of neural network areas of the brain (Garcia, Sabater-Navarro, Gugliemeli, & Casals, 2011).Multiple dimensions of language and cognitive domains can be trained at the same time.In future work we have to analyze and study the reliability and validity, cutoff values, for the language and cognitive system in terms of satisfaction and feasibility (Yu et al., 2014), in order to promote and use it in the clinical setting.

CoNCLUSIoN
The computer-based language and cognition assessment system did not extract factors according to the decomposed principal component extraction analysis and the eigenvalues of the factors, and the factor subsets of the system were selected according to the previous Delphi expert hit evaluation .According to the results of the factor analysis, the correlation among the factors is strong, which provides guidance for the modification of the language cognition assessment system topics.
:Figure caption Js:To calculate Tl:reasoning Yh:Delayed recall Tb:Listen to identify Tf:Whether listening to Zy:Pay attention to Gs:language cognitive assessment system

Figure 2 .
Figure 2. Correlation coefficient matrix of various factors in the language cognitive assessment system:zb: Spontaneous expression Dx:directional Jy:memory Fs:repeat Yd:reading