Reliability of the Brazilian version of the Physical Activity Checklist Interview in children

OBJECTIVE: To assess the reliability of the Lista de Atividades Físicas (Brazilian version of the Physical Activity Checklist Interview) in children. METHODS: The study is part of a cross-cultural adaptation of the Physical Activity Checklist Interview, conducted with 83 school children aged between seven and ten years, enrolled between the 2nd and 5th grades of primary education in the city of São Paulo, Southeastern Brazil, in 2008. The questionnaire was responded by children through individual interviews. It is comprised of a list of 21 moderate to vigorous physical activities performed on the previous day, it is divided into periods (before, during and after school) and it has a section for interview assessment. This questionnaire enables the quantifi cation of time spent in physical and sedentary activities and the total and weighed metabolic costs. Reliability was assessed by comparing two interviews conducted with a mean interval of three hours. For the interview assessment, data from the fi rst interview and those from an external evaluator were compared. Bland-Altman’s proposal, the intraclass correlation coeffi cient and Lin’s concordance correlation coeffi cient were used to assess reliability. RESULTS: The intraclass correlation coeffi cient lower limits for the outcomes analyzed varied from 0.84 to 0.96. Precision and agreement varied between 0.83 and 0.97 and between 0.99 and 1, respectively. The line estimated from the pairs of values obtained in both interviews indicates high data precision. The interview item showing the poorest result was the ability to estimate time (fair in 27.7% of interviews). Interview assessment items showed intraclass correlation coeffi cients between 0.60 and 0.70, except for level of cooperation (0.46). CONCLUSIONS: The Brazilian version of the Physical Activity Checklist Interview shows high reliability to assess physical and sedentary activity on the previous day in children. DESCRIPTORS: Child. Motor Activity. Evaluation. Questionnaires. Reproducibility of Results.


INTRODUCTION
The association between physical activity practice and health outcomes in all ages is well-known, especially in child obesity prevention and treatment. 12,18ysical activity assessment is a complex process because it is a construct with more than one domain -leisure, household, transportation and occupational -and its quantifi cation depends on the type, intensity, duration and frequency of activity. 4,15The Physical Activity Checklist Interview (PACI) 23 is a questionnaire used to measure physical activities in children.According to the cross-cultural adaptation procedures proposed by Herdman et al 10,11 (1997, 1998) and operationalization procedures proposed by Reichenheim & Moraes 19 (2007), Cruciani et al 6 (2011) suggested a Brazilian version known as "Lista de Atividades Físicas" (LAF).The stages of analysis of conceptual, item and semantic equivalences were conducted to achieve this purpose.
The present study aimed to assess LAF reliability in children, as part of the analysis of its psychometric properties.

METHODS
This study was conducted with primary education students from the 2 nd to the 5 th grades, in the city of São Paulo, Southeastern Brazil, in 2008.Total sample size was 83 students, selected according to a simple random sample of the total (N=240), aiming to estimate population intraclass correlation coeffi cient (ρ I ) with a one-sided 95% confi dence interval, ρ o = 0.85, 10% precision and two replicates. 24F was administered in an individual interview on a school day and it enabled the assessment of sedentary and physical activities on the previous day.This instrument includes an initial part with general guidance and instructions for the interview (Attachment).
An analog clock and geometrical fi gures representing 5, 15 and 30 minutes were used to help a child estimate physical activity time.Examples of routine activities were provided to ask the children which lasted more or less than fi ve minutes.
The questionnaire was divided into sections A, B and C. Section A included data such as child's name, grade, start and fi nish time of interview.Section B was aimed at reporting the engagement time (in minutes) in 21 types of moderate to vigorous activities (for fi ve or more minutes of activity), before, during and after school, in addition to the time of sedentary activities (watching television/videos or DVDs and video/ computer game playing before and after school).Section C was aimed at assessing the interview in terms of attentiveness of subject, subject's ability to recall activities and estimate time, level of cooperation, believability of the interview and overall rating of interview, and it should be completed by the interviewer after the end of the interview.
LAF quantifi es the following: time (in minutes) of sedentary activities (SA); time (min) of moderate to vigorous physical activities (PA); total metabolic cost of related activities (min x Metabolic Equivalents (MET) of each physical activity -TMC) and total weighed metabolic cost (min x MET value x adjustment for intensity rate -WMC).
The intensity rate for each physical activity was obtained with the following question: "When you <rode a bicycle> for fi ve minutes or more yesterday <before school>, did you feel the body cues 'breathing hard' or 'feeling tired' none, some or most of the time?".If the intensity of physical activity varied between light to moderate (MET of up to 5.9), this was multiplied by 1.1 (sometimes) or 1.25 (usually).Vigorous physical activities (MET ≥ 6) were multiplied by 0.75 (never) or 1.25 (usually). 23MET values were obtained from the Compendium of Physical Activities by Ainsworth et al 1 (1993).
LAF was applied twice by two interviewers, with a mean interval of three hours between applications (before and after school interval).Sections A, B and C were completed by interviewers; and section C by an external evaluator as well.Interviewers and external evaluators participated in a training course that included an explanation of the research objectives, the LAF study and discussions about the "Interview Instructions".Data obtained from the questionnaires were doubleentered and data entry validity was checked.a The Stata software, version 10.0, was used for statistical analysis.b The values of both interviews (section B) were compared to assess reliability and data from the fi rst interview were compared with those of the external evaluator to assess this interview (section C).Intraclass correlation coeffi cient (ICC) and respective 95% confidence interval (CI95%), 8 Bland & Altman's 2 proposal (1999) and Lin's concordance correlation coeffi cient 13,14 (1989, 2000) were used in the analysis.

RESULTS
Of all 83 participating children, 42 were boys and 41 were girls, with a mean age of 9.3 years (minimum value = 7.1; maximum value = 10.9; standard deviation (sd) = 1.0 years).Due to the occurrence of outliers, identifi ed by the excessive differences in the results between the fi rst and second interviews, three schoolchildren were excluded from the analysis of physical activity and nine from the analysis of sedentary activities, and one who had problems in both activities: basketball before school (1 st interview (I) = 30 min; 2 nd I = 5 min); soccer before school (1 st I = 90 min; 2 nd I = 60 min); volleyball before school (1 st I = 60 min; 2 nd I = 7 min); games before school (1 st I = 0 min; 2 nd I = 90 min); outdoor play before school (1 st I = 30 min; 2 nd I = 0 min); jumping rope before school (1 st I = 30 min; 2 nd I = 10 min); outdoor chores before school (1 st I = 30 min; 2 nd I = 0 min); outdoor play during school (1 st I = 30 min; 2 nd I = 0 min).There were no differences between sexes for the study variables (Table 1).In both sexes, mean time of moderate to vigorous physical activity (PA) was 88 min (95%CI: 75;101) and that of sedentary activity (SA) was 104 min (95%CI: 80;127).Mean value and 95%CI of TMC was 528.5 MET (445.2;611.7)and that of WMC was 503.1 MET (424.2;582.0).By comparing PA, SA, TMC and WMC values between both interviews, the lower limit of the ICC was found to vary between 0.84 and 0.96.Precision (r) and accuracy regarding the perfect agreement (C b ) values varied from 0.826 to 0.972 and from 0.985 to 1, respectively (Table 2).
Figure 1 shows that the line originated from the data reveals estimates of inclination and intercept similar to those of the line of 45º, indicating high agreement (precision) of the data originated from the LAF. Figure 2 indicates independence between the magnitude and dispersion of values of PA measures of both interviews.
Mean differences of PA, TMC, WMC and SA were 4.48 min, 17.8 MET, 25.8 MET, and 1.35 min, respectively.Based on agreement limits, the fi rst application was found to underestimate PA in 95% of times in up to 50 minutes; TMC, up to 294.6 MET; WMC, up to 298.4 MET; and SA, up to 47.4 min (Figure 2, Table 2).
Mean duration of interview was 27 minutes (sd = 7 min).In the overall rating of interview of Section C (Table 3), 13.3% of interviews were classifi ed as fair and 86.7% as good, very good or excellent.In 95.2% and 96.4% of cases, "attentiveness of subject" and "level of cooperation" were classifi ed as good, very good or excellent, respectively.Subject's ability to recall activities performed on the previous day and the believability of the interview showed similar results, 14.5% of which were categorized as fair.Subject's ability to estimate time showed a higher percentage of cases classifi ed as fair (27.7%), whereas the remaining cases were good, very good or excellent.
ICC values varied between 0.60 and 0.70, except for "level of cooperation" (ICC = 0.46) (Table 3).Confi dence intervals for all items were wide, indicating low precision in the process of estimation of parameters.

DISCUSSION
The exact estimates of ICC for PA, TMC, WMC and SA were lower than those in the original study: 23 0.64; 0.65; 0.65 and 0.75 respectively, when compared to the lower limits of estimates per interval in the present study (0.84 to 0.96).Such differences can be justifi ed by the different methodologies used.In the original study, PACI values were obtained from individual interviews and the second application was self-administered by the entire class.In this study, researchers sought to create the same experimental conditions in both applications.
The estimates of reliability in the present study are higher than the majority of results from previous day questionnaires for children, which indicate ICC values of 0.17, 27 0.59, 25 0.60 22 and 0.94. 20In the study conducted by Ridley et al 20 (2006), this instrument was used in 32 Australian children aged between nine and 13.5 years, with a mean interval of four hours between applications, whereas, for other instruments, these applications were conducted on separate days.According to Patterson 17 (2000), the interval of time between applications in a reliability study should be such that the assessed days overlap, thus preventing variability due to the weekday.
The period of measurement covered by the instrument is the object of studies and previous day questionnaires 26 are those most frequently used to assess physical activity in children.According to Cale 3 (1994), instruments that require one to remember activities performed in periods such as the previous week, month or year may be inappropriate for children.Thus, a positive aspect of the LAF is that it reduces the chance of memory bias.
The small mean differences found in the values obtained from both applications for PA,   20 (2006) argued that, although ICC was high, their results reveal a wide variation in the agreement limits, suggesting that the instrument can be more reliable for group estimates, rather than individual ones.
By comparing the interview assessment results (Section C) of the PACI 23 with those of the present study, the percentages of interviews assessed as very good or excellent were similar (respectively): attentiveness of subject -67% vs 60.3%; subject's ability to recall activities -54% vs 56.7%; subject's ability to estimate time -41% vs 36.1%;level of cooperation -83% vs 77.1%; believability of the interview -53% vs 57.8%; and overall rating of interview with subject -56% vs 57.8%.In both studies, the results indicate that children have a diffi culty in estimating time of activities performed on the previous day.
However, LAF's Section C items showed low reliability values.It is recommended that such section be used to identify children with low scores and thus conduct another interview.
Ainsworth et al's Compendium 1 (1993) was used to calculate scores of physical activities (weighted and not weighted), following the original study, which enables the comparison of reliability values between studies to be made.This compendium 1 was proposed for the adult population and, although Ridley et al 21 (2008) had published one that was specifi c for children and adolescents, there are not suffi cient studies that adopt such compendium to emphasize its use.
LAF's format is based on the school period (before, during and after school), thus not enabling the assessment of physical activities performed on weekends.The school environment is described in the literature as one that promotes a higher physical activity level, because children have the opportunity to participate in playful activities, games and lessons that promote physical activity, including physical education. 7This could be one of the reasons for children to seem to be more active during the week, rather than on weekends. 7,9,16According to the LAF, the effect of quantifying PA only during the week interferes with weekly estimates.However, in association or correlation studies, it is possible that there is no need of a weekly quantifi cation, but rather of identifying more or less active children.It is believed that the analysis of PA during school days can produce suffi cient information to qualify population groups in terms of the amount of physical activity performed.
Reporting physical activity depends on the respondent's cognitive maturity, 15 as observed when 11 children were excluded, suggesting that the information provided before the interview would not be suffi cient to change this situation.As the majority of exclusions refer to sedentary activities (n = 9), it is supposed that the diffi culty in quantifi cation is associated with the low perception of identifi cation of moments when sedentary activities begin and end.Thus, it is necessary that additional studies be performed to explore strategies that reduce the diffi culty in assessing this type of activity.
LAF can be applied to children who go to school in the mornings, although the after-school period will be longer than that in the original format.From the point of view of physical activity assessment, this fact does not interfere with the results, because physical activity indicators are originally proposed for the day as a whole, rather than its periods.LAF has important characteristics for children's physical activity questionnaires: 3,5 administration through interviews, previous day assessment, questionnaire responded by the child itself, questionnaire with a list of physical activities and day divided into periods, and instructions to guide interviewers.
LAF shows high reliability to measure physical activity in children aged seven to ten years.Because it is a cross-cultural adaptation of the instrument validated in the culture of its country of origin, the analysis of its validity in the target culture is recommended.Comercial de TV? (Criança responde mais ou menos de 5 minutos?)

a
Dean AG, Arner TG, Sunki GG, Friedman R, Lantinga M, Sangam S, et al.Epi Info, a database and statistics program for public health professionals.Atlanta: Centers for Disease Control and Prevention; 2002 b StataCorp.Stata Statistical Software: Release 7.0.College Station: Stata Corporation; 2001.

Figure 2 .
Figure 2. Joint distribution of physical activity values in both interviews.São Paulo, Southeastern Brazil, 2008.

Table 1 .
Mean values of daily physical activity and sedentary activity, according to sex and total.São Paulo, Southeastern Brazil, 2008.

Table 3 .
Distribution of students according to assessment concepts and interview aspects and intraclass correlation coeffi cient with 95% confi dence intervals for LAF's Section C items.São Paulo, Southeastern Brazil, 2008.(n = 83) It is possible that the LAF is more appropriate to characterize physical activity of population groups than individual ones.Similar values of agreement limits, described by Ridley et al 20 (2006), support this hypothesis, which showed values of -53.4 and 51.2 minutes according to Bland & Altman's proposal 2 (1999).Ridley et al ICC: Intraclass correlation coeffi cientFigure 1. Joint distribution of physical activity values in both interviews.São Paulo, Southeastern Brazil, 2008.