Nutrición ADAPTATION AND REPRODUCIBILITY OF A QUESTIONNAIRE TO ASSESS PHYSICAL ACTIVITY IN EPIDEMIOLOGICAL STUDIES AMONG YAQUI INDIANS FROM SONORA , MÉXICO

Introduction: Due to the benefits of physical activity (PA) for health and the associated risk of physical inactivity with obesity and non-communicable diseases, the evaluation of PA is of great interest in public health. Objective: Was adapt and evaluate the reproducibility of a questionnaire to assess PA in epidemiological studies among Yaqui Indians from Sonora, México. Methods: The participants were 100 Indians from the 9 Yaquis communities, including men and women aged 20 or older. The adaptation phase of the PA questionnaire (PAQ) was done by interviewing 40 Yaquis about their activities related to leisuretime, occupation and transportation. For the reproducibility analysis, the PAQs applied by duplication to 60 adults were evaluated by the Spearman Correlation Coefficient (ρ) and Intraclass Correlation Coefficient (ICC). The level of agreement was evaluated by the Bland-Altman plot. Results Eighteen leisure time activities, 34 occupational activities, and 5 transportation activities were included in the PAQ. Very strong correlations were found for total leisure-time and occupational activities (ρ =0.92 and ρ=0.88, respectively; p<0.0001), and sedentary behaviors (ρ=0.94; p<0.0001) for the previous 12 months. The Bland-Altman plots indicated high agreement between PAQs. Conclusions: The adapted PAQ had high reproducibility for measuring PA in the Yaqui Indian population.

The advantage of the PAQs, besides their low cost, is that they can be designed and validated for a given population, including ethnic groups, thus providing information of the specific patterns of physical activity of the population (Kriska et al., 1990;Lavandera-Torres & Esparza-Romero, 2017;Macniven et al., 2016). PAQs can lead to inaccuracies because they depend on the memory of the interviewee , but strategies can be generated to reduce this disadvantage. One important aspect that should be considered when PAQs are selected to measure PA is their previous adaptation based on the target population to be studied. Applying a PAQ generated from a particular community to a new population with different cultural characteristics can lead to inaccuracies (Arredondo, Mendelson, Holub, Espinoza, & Marshall, 2012). The short version of the International Questionnaire of Physical Activity, which originally had adequate validation and reproducibility for 12 countries, had poor validity and a modest reproducibility for the evaluation of moderate and vigorous PA in the Mexican population (Medina, Barquera, & Janssen, 2013). Although not discussed by the authors (Medina et al., 2013) it is important to consider the adaptation process in selecting PAQs (Moreira et al., 2013;Oyeyemi et al., 2016;Oyeyemi et al., 2011).
Indigenous populations have less favorable health conditions and higher prevalence of obesity and NCD than non-indigenous populations (Gracey & King 2015). This makes the specific study of PA and its association with obesity and NCD in these ethnic minorities relevant. However, it is necessary to generate and evaluate measurement instruments, such as PAQs, according to the context and culture of these populations (Arredondo et al., 2012;Medina et al., 2013). Some questionnaires have been designed and validated for indigenous populations (Esparza et al., 2010;Kriska et al., 1990;Lavandera-Torres & Esparza-Romero, 2017;Trost, Marshall, Miller, Hurley, & Hunt, 2007). Kriska et al. (1990) developed a PAQ to study PA and its role in type 2 diabetes in U.S. Pima Indians. Similarly Lavandera-Torres & Esparza-Romero (2017), adapted and studied the reproducibility of a PAQ in indigenous Seris of México with good results. Esparza et al. (2010) adapted the questionnaire by Kriska et al. (1990) to investigate the relationship of PA with obesity in Mexican Pima Indians.
The Yaqui Indians are an ethnic group from Sonora, Mexico, originally settled along the Yaqui River, between the municipalities of Cajeme and Guaymas. Its territory includes three areas: mountainous, fishing and farming (Moctezuma-Zamarron, 2007). This indigenous group has presented changes in their lifestyle; for example, their agricultural activity has gone from being self-sufficient to commercial, from manual to mechanized and even to land rent. It is reasonable to speculate that their physical activity has decreased as a result. Furthermore, while studies have reported that the Yaquis present problems of obesity and NCDs (Brito-Zurita et al., 2007;Rodríguez-Morán et al., 2008), no information has been reported about their PA. The purpose of this study was to adapt and evaluate the reproducibility of a PAQ to study PA in adult Yaquis in Sonora, Mexico.

Materials and Method Participants
The study participants were 100 Yaqui Indians, including men and non-pregnant women aged 20 or over from the following Yaqui Indians communities: Loma de Guamúchil, Loma de Bácum, Tórim, Vícam Pueblo, Vícam Switch, Casas Blancas, Pótam, Ráhum and Bélem. Participants were selected if they had stable and/or temporary employment or domestic duties. The number of participants were 40 for the adaptation phase of the PAQ and 60 for the evaluation of its reproducibility. Participants were selected proportionally to the adult population of each community using information from the National Institute of Statistics and Geography (INEGI, 2010). The sample size was based in obtaining significant correlation coefficients greater or equal to 0.4, with α=0.05 and a power of 90% (Norman & Streiner, 2000). The interviews were completed in the homes of the participants during the period of June through August, 2016.
Description of the Physical Activity Questionnaire The applied questionnaire was similar to that designed and validated by Kriska et al. (1990) for U.S. Pimas and adapted by Esparza et al. (2000) for the Pima population of Sonora, México. This PAQ evaluates the PA of the previous 12 months and for the previous week. For purposes of this paper, only the previous 12 months were evaluated. The PAQ consists of two sections: the first includes questions about the leisure-time activities and the second about occupational activities. Frequencies such as months per year, weeks per month, days per week and the number of hours per day for each activity were registered to estimate the average number of hours per week (h/wk) for each activity.

Adaptation
For this phase, the methodology proposed by Lavandera-Torres and Esparza-Romero (2017) was carried out for 40 Yaquis Indians. During each interview, the participants were asked to describe in detail their leisure-time, occupational and transportation activities. The information generated was book recorded, considered and included in the PAQ.
To classify the identified activities, they were analyzed in detail and assigned with metabolic equivalents (METs) based on the Compendium of Physical Activity (Ainsworth et al., 2011). Activities were classified as sedentary behaviors (1 to 1.5 METs) and light (1.6 to 2.9 METs), moderate (3 to 5.9 METs) and vigorous (≥6 METs) activities. A MET represent the oxygen rate of approximately 3.5 ml/kg · min for an average adult who is sitting quietly (Ainsworth et al., 2011).

Reproducibility
The reproducibility of the adapted PAQ was evaluated through the test-retest protocol. The PAQ was applied to the Yaquis adult sample twice with a margin between applications of 1 to 3 weeks.

Anthropometric Measurements
Participants were evaluated for weight, height, and waist circumference according to the methodology of the International Society for the Advancement of Cineanthropometry (ISAK, 2001).

Statistical Analysis
The PA (h/wk) was analyzed as total, light, moderate and vigorous leisure-time and occupational activities and as sedentary behaviors (i.e., watching television and talking/sitting at home with family and neighbors [socializing]). Furthermore, PA was additionally analyzed as specific domains such as sports (soccer, baseball and softball), craftwork (embroidery and making reed curtains), agricultural (sowing by hand, harvesting, and fumigating) and housework (sweeping, mopping, washing clothes by hand, and preparing food).
Bland-Altman plots were used to determine the level agreement between the PAQ1 and PAQ2 regarding total leisure-time, occupational activities and sedentary behaviors (h/wk). In other words, Bland-Altman plots provided information about the magnitude of overestimation (mean bias above the zero line) or underestimation (mean bias below the zero line) of the PAQ1 with respect to PAQ2; in either case, mean bias was analyzed by the paired ttest. Simple linear regression was performed to evaluate homogeneity of the differences (bias) over the range of average values (h/wk) from the two PAQs measurements. The dependent variables were the differences (h/wk) between PAQ1 and PAQ2, and the independent variables were the average (h/wk) of both PAQs. Biases falling outside the Bland-Altman agreement interval were expressed as percentage.
Stata software version 14 was used for the statistical analysis (Stata Corp LP, College Station, Texas, USA; URL: http://www.stata.com). Significance was considered at a p-value of p≤0.05.

Ethics approval
The protocol was approved by the Ethics Committee of the Research Center for Food and Development. Before the interview, each participant received complete information about the project and signed a consent form.

Participants
Initially, there were 66 participants for the study of reproducibility, but only 60 completed the second PAQ (15 men and 45 women). Ages ranged from 20 to 65 years, with an average of 40.0 years for men and 37.8 years for women. Anthropometric characteristics such as weight, height, body mass index and waist circumference are shown in Table 1.
Adaptation Eighteen leisure-time activities were recorded, including sports and exercise related activities such as football, baseball, softball, jogging, and walking as well as sedentary behaviors such as watching television and socializing. Thirty-four occupational activities were reported and related to agriculture, silviculture, construction, craftwork, among others; each of the occupational activities consisted of a series of sub-activities. Housework activities consisted of sweeping, mopping, handwashing, hauling water, making flour tortillas and others. Activities related to transport included walking, biking, and riding horses or traveling in cars or trucks. Table 2 shows a larger number of activities reported by the Yaquis Indians along with the values of METs assigned to them. These activities were listed in the PAQ. In addition, an open question was added at the end of the list to further include activities that interviewees considered as being part of their PA pattern.
With regard to total sedentary behaviors, the ρ value (ρ=0.94; p<0.00001) indicated a very strong correlation. The correlations for watching television and socializing at home were also considered very strong. The ICC values for all sedentary behaviors indicated almost perfect correlations (Table 3). The Bland-Altman plot (Figure 1) shows mean bias, the regression line and the limits of agreement for total leisure-time activities (h/wk). The mean bias was not different from zero (mean bias=0.72, p>0.05). In the same plot, the regression line indicated that differences between measurements (β = 0.041, p=0.77) were homogeneously distributed throughout the average (h/wk) of total leisure activities. The limits of agreement (LOA) were -10.11 (lower LOA) to 8.66 (upper LOA) with 10% of bias outside of LOA. Regarding total occupational activities, the Bland-Altman plot (Figure 2) revealed that mean bias was also similar to zero (mean bias=-0.39, p=0.76).
Further, the plot also shows that differences were homogeneous among the average values in h/wk of total occupational activities (β=0.87, p=0.16). The LOA were 21.3 (lower LOA) to 20.5 (upper LOA), with 5% of the bias outside of LOA.

Figure 2:
Bland-Altman plot for total occupational activities (h/wk) as assessed by the PAQ1 and PAQ2

Discussion
The adaptation phase allowed for a detailed identification of common leisure-time, occupational and transportation activities among the Yaqui Indians to investigate how activities were performed.  The leisure-time activities mainly included watching television, socializing (sedentary behaviors) and some sports activities. The activity of socializing is facilitated because Yaqui households consist of a large, fenced-in land in which different generations of the same family live together (grandparents, parents and offspring). During the afternoon and mid-morning, Yaqui family members typically chat in their homes under the shade of trees or ramadas (a roof, without walls, built with mesquite sticks, reed and other plants). On the other hand, occupational activities were related to agriculture, construction, silviculture, craftwork activities and housework, among others.
The adaptation phase allowed for uncovering the types of PA that adult Yaquis perform in their daily lives as well as how they are carried out. This latter information was useful to further classify sedentary behaviors light, moderate, and vigorous PA. This phase also permitted the identification of the appropriate terms used for the different PAs, which was useful for reducing misunderstandings during the interview and thus improving the accuracy of the PAQ.
Regarding reproducibility, the correlation values obtained for total leisure-time and occupational activities in the previous 12 months were similar to those reported by Kriska et al. (1990) in U.S. Pima Indians 21 to 36 years of age (ρ=0.92 y ρ=0.88, respectively, p<0.05). These correlation values were higher than those reported by Lavandera-Torres and Esparza-Romero (2017)  The reproducibility of sitting time as a marker of sedentary behaviors was evaluated in different studies using the International Physical Activity Questionnaire (Chu & Moy, 2015;Oyeyemi et al., 2011;Vasheghani et al., 2011). The results of these studies were ICC=0.84, ICC=0.73 and ICC=0.56, for Malaysia, Iranian and Nigerian populations, respectively (Chu & Moy, 2015;Oyeyemi et al., 2011;Vasheghani et al., 2011). Sedentary behaviors in this study had higher correlation values.
The Bland-Altman plots showed that PA measured by PAQ1 was similar to that measured by PAQ2. That mean bias was close to zero indicates agreement between both questionnaires. In addition, linear regression analysis showed that the error or differences between PAQs were constant over the range of h/wk of physical activity.
The results of the reproducibility of this study are largely explained by the adaptation process because it allowed for the use of specific terms for the different activities in the community and for the idea of listing them in the questionnaire format. This means that interviewees were able to better understand the questions, and the interviewer could be clear about the meaning of the questions. Another important aspect of this study was the previous training received by the interviewers, part of which consisted of understanding the community and conducting previous PAQs with people of the community.

Conclusions
The adaptation phase allowed for a detailed identification of the most important leisure-time, occupational and transportation activities among the Yaqui Indians to investigate how these activities were performed and to determine the appropriate terms to use for the PAQs. The adapted PAQ had high reproducibility for measuring PA during the previous 12 months, and it may therefore be used to evaluate PA and its association with obesity and NCD in the Yaqui Indian population.