Psychometric properties of the Exercise Benefits/Barriers Scale in Mexican elderly women

ABSTRACT Objective: analyze and assess the psychometric properties of the subscales in the Spanish version of the Exercise Benefits/Barriers Scale in an elderly population in the Northeast of Mexico. Method: methodological study. The sample consisted of 329 elderly associated with one of the five public centers for senior citizens in the metropolitan area of Northeast Mexico. The psychometric properties included the assessment of the Cronbach's alpha coefficient, the Kaiser Meyer Olkin coefficient, the inter-item correlation, exploratory and confirmatory factor analysis. Results: in the principal components analysis, two components were identified based on the 43 items in the scale. The item-total correlation coefficient of the exercise benefits subscale was good. Nevertheless, the coefficient for the exercise barriers subscale revealed inconsistencies. The reliability and validity were acceptable. The confirmatory factor analysis revealed that the elimination of items improved the goodness of fit of the baseline scale, without affecting its validity or reliability. Conclusion: the Exercise Benefits/Barriers subscale presented satisfactory psychometric properties for the Mexican context. A 15-item short version is presented with factorial structure, validity and reliability similar to the complete scale.

their personal background, the belief persists that it could be a waste of time in their daily agenda (13) .
The perceived barriers for the action allude to the negative mental assessments or -imaginary or realindividual impediments that can hinder a commitment to a health conduct. The barriers represents the perceived unwillingness, inconvenience, cost, difficulty or time spent to execute the conduct; they encourage towards the avoidance of the conduct planned. Therefore, when the willness to perform the action is low and the barriers high, executing the conduct will be difficult (6) . The main barriers to exercise identified include the bad climate, the lack of discipline, time, money or company to perform the action (14) . In addition, in adult women of median age, the barriers are health problems, age-related injuries and problems (15) . (16) was designed in English to measure these thoughts by Dr.

The Exercise Benefits/Barriers Scale [EBBE]
Nola J. Pender in the United States of America. It has been translated and validated in elderly populations in Korea (17) and Brazil (18) and, in China, an adapted version was developed and validated to be applied in dialysis patients (19) . The Spanish version was also published by the original authors of the EBBE with acceptable reliability coefficients in Colombia (12) and Mexico (2) . Nevertheless, no published information has been found on the psychometric properties of the Spanish version. These perceptions might differ in function of the group studied. In addition, variations within the same language might affect the validity of the adapted scales (20) . Therefore, it is relevant to analyze the functioning of the scale in an elderly population in the Mexican context.
The objective was to analyze and assess the psychometric properties of the subscales in the Spanish version of the Exercise Benefits/Barriers scale in an elderly population in Northeast Mexico. In addition, in a secondary analysis, the feasibility of a short version will be assessed, intended to make it easier to estimate the strength of these perceptions in this population.

Method
A methodological study was undertaken in a population of 2701 community-based elderly, affiliated with five public centers for senior citizens in the metropolitan area of Northeast Mexico. As the number of men attending these centers is very low, the participants in this study were exclusively women.

Participants
Women between 60 and 80 years of age were included, with intact cognitive skills according to Pfeiffer's Questionnaire, able to read and write, without medical contraindication to exercise and who accepted to participate Enríquez-Reyna MC, Cruz-Castruita RM, Ceballos-Gurrola O, García-Cadena CH, Hernández-Cortés PL, Guevara-Valtier MC.
in the study voluntarily. Women who despite the result of the Pfeiffer Questionnaire demonstrated inability to understand instructions were excluded. The sample was calculated using the formula for finite populations and consisted of 329 participants. Simple stratified sampling was used, based on the list of women who attended each of the strata (public centers for senior citizens).

Instrument
The Spanish version of the EBBE has been published together with the English version ( Figure 1) and was initially translated to Spanish by Juarbe T. It consists of a 43-item quasi Likert scale with four alternative answers. The score "four" corresponds to strong agreement with the assertion, "three" to simple agreement, "two" to simple disagreement and "one" to strong disagreement with the item. The scale includes two subscales: 29 items for the subscale of perceived exercise benefits and 14 for the subscale of perceived exercise barriers. The items of the exercise barriers subscale correspond to assertions 4,6,9,12,14,16,19,21,24,28,33,37, 40 and 42. To assess the 14 statements representing the barriers, the answer range varies between 14 and 56; for the benefits, on the other hand, scores range from 29 to 116. In both subscales, a higher score corresponds to a higher perception concerning exercise practice (16) .

Procedure
The preliminary analysis included two steps: 1) linguistic and culture review by experts and 2) qualitative pilot study in a small sample of elderly with characteristics similar to those of the final sample. The first step was to review, involving three experts in gerontology nursing, the adaptation of the vocabulary and writing to the Spanish of this Mexican context. The guidelines of the International Test Commission for the adaptation of tests were followed: cultural and language differences, technical aspects and methods and interpretation of results (20) . After collecting the information from the experts, the EBBE was applied to a group of 30 elderly to assess the clarity and adequacy of the measure. As a result of these steps, the formulation of 12 items was modified.
The study received approval from the institutional ethics committee and from the authorities of the public centers for senior citizens. Properly trained physical exercise professionals collected the data individually and privately while the participants attended the clubhouse.
The completion of the EBBE took between five and ten minutes.

Data analysis
First, the complete version of the EBBE was analyzed,

Results
The participants' average age was 69 years (SD=5.44), with 6.5 years (SD=2.92) of education. Only 42% confirmed having a partner.

EBBE -complete version
The result of the application of Bartlett's test to the correlation matrix among the EBBE items was significant    For the confirmatory factor analysis, the least squares method and the maximum likelihood method were used, in accordance with the distribution of the data.  (20) . GFI, AGFI and NNFI coefficients superior to .90 indicate good adjustment (22) . The standardized PGFI coefficients range between 0 and 1. As none of both reaches the limit of .90, coefficients closer to .80 are considered adequate (23) . For the RMSEA, coefficients between .05

Psychometric properties per subscale
and .10 are considered acceptable, and ideal coefficients correspond to 0.08 or less; for the RMR, low coefficients are required, with coefficients closer to zero indicating better adjustment (24) .

Secondary analysis to facilitate the application to elderly people
In view of the difficulties to adjust some parameters, the relevance of a factorial solution was analyzed that would be satisfactory for the structural parameters of the model as well as for the validity and internal consistency. In Table 3, the results for the exercise benefits subscale are displayed.  (Table 4).  Enríquez-Reyna MC, Cruz-Castruita RM, Ceballos-Gurrola O, García-Cadena CH, Hernández-Cortés PL, Guevara-Valtier MC.

Discussion
The reliability results of the Mexican version of the EBBE present essential similarities with the parameters published for the original version (16) . The alpha coefficients of the two subscales of the EBBE presented adequate internal consistency coefficients and were similar to the results obtained in the adaptations in Korea and Brazil (17)(18) .
Considering that the reference point to discuss the results of the adaptation of a scale to a linguistic and cultural context are the related studies (20) ; the validity and reliability coefficients found support the use of the EBBE in an elderly population in Northeast Mexico.
The factorial structure and item distribution between the factors of the exercise benefits subscale are in line with the findings for the original version (16) . The high inter-item correlations support the construct validity of this subscale; the discrimination indices can be considered adequate and similar to the findings for the original version.
In contrast, the exercise barriers subscale demonstrated merely acceptable reliability and validity coefficients. This detail was also observed when the original version of the EBBE was applied in an adolescent American population (25) and in the other adaptations published (17)(18) .
The confirmatory factorial analysis reveals the problem; the low inter-item correlation coefficients suggest the need to review the construct (20) . To give an example, item 21 refers to the "husband or partner's lack of support to exercise", the lack of explanatory power of this item in this sample can be due to the small proportion of participants who signaled having a partner. This explanation could also apply to the case of the adolescent population. As the exercise barriers may depend on aspects directly relate to the population context and culture, the construct needs to be analyzed before making decisions based on this subscale. In short, the EBBE demonstrated a two-factor structure, in accordance with the theoretical principles that guided its construction.
The analysis of the factorial structure of the two subscales revealed that, in this sample, the fit indices