Background
Physical activity is an important factor among the determinants of health due to it’s protective factor and preventive role [
1]. More than half of the Hungarian population is overweight and two thirds do not do sports regularly [
2,
3]. Such behaviours among developed European citizens have been associated with chronic metabolic and musculoskeletal disorders such as type two diabetes, hypertension, obesity, and coronary heart disease, as well as psychological impairments and imbalanced mental health status [
4‐
7].
The World Health Organization (WHO) guidelines and recommendations state that to maintain health, adults younger than 65 years old should perform at least 150 min of moderate intensity physical activity or at least 75 min of vigorous intensity physical activity throughout the week [
8‐
10]. In this case physical activity (PA) has been defined as “
any bodily movement produced by skeletal muscles that results in energy expenditure”. The main domains of PA are work, active transportation and leisure time activities. According to intensity, moderate (4 MET) and vigorous activities (8 MET) can be classified and walking activities should be also distinguished (multiplied by 3.3 MET) [
11,
12].
The monitoring techniques are useful to examine the population’s activity and determine lifestyle trends. Self-reported measures such as questionnaires are most commonly used in public health studies because of the low costs, minimal burden, easy implementation, and valuable information. However, completing a self-administered PA questionnaire could be difficult to understand for participants, may induce bias, and thereby may over- or underestimate actual patterns of PA. Therefore, accelerometers are a widely used method to assess concurrent validity of PA questionnaires [
13].
At the end of the twentieth century the International Physical Activity Questionnaire (IPAQ) was developed; the long form with 31 and the short form with 9 items [
14,
15]. The long form has been considered too long and the short version not sufficient to analyse the physical activity patterns of the respondents. To complete and correct these deficiencies, the Global Physical Activity Questionnaire was compiled [
16].
The Global Physical Activity Questionnaire (GPAQ) was developed by the World Health Organization (WHO) in 2002 and was endorsed as STEPwise Approach to the Chronic Disease Risk Factor Surveillance (STEPS). The questionnaire was constructed with special attention to the physical activity habits of the population of developing countries [
17].
The first version of the GPAQ was validated in 9 countries, mostly in Asia, Africa, and South America. Based on the experience of the GPAQ v1, the GPAQ v2 was developed after minor revisions in 2005 with 16 items reflecting work, transportation, leisure time activities, and assessment of daily sitting time. GPAQ v2 was initially validated in Europe in Portugal and in Great Britain [
18].
To ensure cultural adaptation of the tool, Mathews et al. developed a modified version of GPAQ according to local cultural tradition for adult women in India [
13,
19]. The comparative validation study revealed significant but weak to moderate correlation between GPAQ and accelerometer data. The European validation studies showed weak to moderate correlation for moderate to vigorous PA (MVPA) [
12,
20].
Furthermore, based on the study of Riviere et al. the IPAQ long version questionnaire proved to be an adaptive instrument to validate the GPAQ. These two quantitative techniques are similar as they contain the same domains (except the household activities which is not part of the GPAQ) and for this reason it is a relevant measurement tool to examine the concurrent validity [
21].
The aim of the present study was to adapt and validate the self-administered GPAQ - Hungarian version (GPAQ-H) against accelerometer data and IPAQ-Hungarian long version (IPAQ-HL) in Hungarian healthy young adults.
Discussion
This study showed the validity and reliability of the GPAQ-H measurement tool in comparison with accelerometer and IPAQ-HL data. Our results demonstrated fair to moderate validity of the Hungarian GPAQ compared to the accelerometer data and moderate and good correlation with IPAQ-HL questionnaire. We examined the correlation between accelerometer and questionnaires according to moderate, vigorous, MVPA activities, and sitting time values. Our results are consistent with other studies according to the intensity of the correlation coefficients.
The GPAQ-H vigorous data were showed significant moderate correlation with accelerometer-moderate and accelerometer-MVPA results, but there were no significant results with accelerometer-vigorous data. The GPAQ-H moderate values did not correlate with MVPA, only with accelerometer-moderate results. The GPAQ-H MVPA showed significant correlation with moderate and MVPA accelerometer values. The GPAQ-H sitting time did not correlate with the examined accelerometer parameters. In case of the subgroup analysis our results were similar according to genders. We noticed significant difference only by vigorous activities irrespective of the measurement method (GT3X p = 0.048, GPAQ-H p = 0.046, IPAQ-HL p = 0.017), and by objectively measured sitting time (p = 0.018). Otherwise, in case of the total sample, sitting time did not show a significant correlation between questionnaire and accelerometer data, but there was a significant negative correlation between accelerometer sitting time value, the GPAQ-H MVPA (R = -0.296, p < 0.001), and vigorous values (R = -0.325, p < 0.001). The GPAQ-H and IPAQ-HL questionnaires showed moderate and good correlation and similar mean values, but the overestimation of the MVPA, moderate and vigorous activities was higher in IPAQ-HL.
In the French validation study of GPAQ, Riviere et al. applied similar study design as our research group: they measured PA patterns of staff members and students (
N = 92, age 30.1 ± 10.7, 76.9% BMI 18.5–24.9) of the University of Lorraine, using IPAQ-LF for concurrent and ActiGraphs for criterion measures. Multiple overestimation of PA – in particular for vigorous intensity (more than tenfold) – was characteristic in case of self-reports. Regarding intensity, Riviere et al. found correlation only between vigorous activities (
R = 0.38) and not any significant relationship between moderate activities (
R = 0.10). Comparing total activities across all domains of GPAQ with accelerometer-moderate activity (
R = 0.40) and with accelerometer vigorous values (
R = 0.24), modest significance was found. They observed poor significant relationship when examining the correlation between self-reported sitting time, accelerometer-sitting time (
R = 0.42), and accelerometer-moderate activities (
R = -0.22). By retest, the research found poor values by moderate leisure and total PA (ICC = 0.37 and 0.58 respectively) but good or almost perfect values by total sedentary and vigorous PA at work (ICC = 0.80 and PABAK = 0.91). Comparing GPAQ and IPAQ-LF, important discrepancies were found, and the classification with level of PA was only poorly to moderately correlated by the concurrent validity (Phi coefficient 0,22–057) [
21].
Mumu et al. found fair to moderate correlation between objective and subjective monitoring, still claimed GPAQ as an acceptable measure, particularly among women with higher level of education despite the under-estimation of sedentary behaviour (
R = 0,23,
p < 0.001) [
31]. The authors explain divergence of the results by genders with PA habits, contrary of other studies [
12] in favour of females. In Bangladesh – a least developed country according to United Nations classification – walking is more specific by work activities for females which is more reliably monitored with accelerometers than upper-body motions of males during intensive farming or carriage of heavy loads e.g. swimming or cycling. 60% of the sample in the study by Mumu et al. belongs to the rural population – in our study, 96% of the sample belongs to urban population. This difference may be behind more equal PA habits between genders.
Meeting PA guidelines but being highly sedentary for the rest of the days is also an emphasized risk factor [
32]. Chu et al. negotiated sedentary behaviour measures, using a domain-specific Adult Sedentary Behaviour Questionnaire (ASBQ) and the Global Physical Activity Questionnaire’s (GPAQ) single-item sitting question against triaxial ActiGraph wGT3X-BT accelerometers. They found significant correlation between accelerometer and GPAQ in sedentary time, while the GPAQ under-estimated the time spent sedentary. However, moderate to good test-retest reliability (
R = 0.74, ICC 0.62–0.82) was presented [
22].
Measurement of inactivity proved to be doubtful in other studies as well. Cleland et al. found poor correlation with daily sitting time in minutes (
R = 0.187), and reported that those people who are more sedentary were less likely to under-report their level of SB. The authors stated that this questionnaire could not be considered as a valid tool to measure sitting time [
12]. They postulated both long (
R = 0.33) and short (
R = 0.34) form of IPAQ with higher sitting items more appropriate referring to a previous research [
33]. We observed the same tendency in our study and we hypothesise that this may be due to the context of health literacy and health behaviour of participants. Describing perceived levels of activities is difficult, not only in case of SB but also regarding MVPA. Cleland et al. found only moderate correlation between the objective and subjective measurement (
R = 0.484) of MVPA (10).
However, GPAQ was originally designed to be interviewer-administered by the WHO. Yet, similarly to many authors, we also decided to record the questionnaires in self-administrated form. The way of query did not justify bias or discrepancy in prior examinations. In the study of Chu et al. data with self-administration were not weaker than with interviewer-administration, yet they found only fair-to-moderate correlations for moderate-to-vigorous physical activity (R = 0.30, R = 0.46 respectively). Strongest correlations were observed for vigorous-intensity activity (self-report R = 0.38, with interviewer R = 0.52). Bias were illustrated with Bland-Altman plots toward overestimation of higher levels of vigorous- and moderate-intensity activities, and underestimation for lower levels PA, parallel to similar studies in general. Reliability for MVPA revealed moderate correlations (self-report R = 0.61, with interviewer R = 0.63). To reduce bias in the GPAQ measurements they advised to incorporate accelerometers, particularly by the measurement of different intensity PA (A. H. Chu, Ng, Koh, & Muller-Riemenschneider, 2015).
Wanner et al. measured the validity of GPAQ in European context. They found significant results as other Western countries, like fair-to-moderate validity of the GPAQ questionnaire. The range of the overestimation of the GPAQ was between 2.8–4.2 times, which mean that GPAQ results are notably higher than the accelerometer data. Total activities showed fair correlation between GPAQ and the accelerometer (
R = 0.22), but the MVPA showed weak correlation (
R = 0.11), while vigorous activities were moderately correlated with accelerometer (
R = 0.46) like sitting time (
R = 0.47). The results of the Wanner et al. study showed significant difference between gender, where male participants were more likely to overestimate their vigorous activities [
20]..
The reason for overestimation of time spent with MVPA may also be in relation with the lack of appropriate knowledge on adequate value of health enhancing physical activity and the perception of importance of physical activity [
34]. Besides, a better understanding of the questionnaires could help to receive more accurate results. Cleland et al. also found higher validity in higher-income countries due to higher education levels [
12].
Contrary to the above results, Laeremans et al. compared GPAQ results with another wearable sensor (SenseWear) in a multi-centre (Antwerp-Barcelona-London) study and demonstrated significantly lower (
p < 0.05) time and energy expenditure (MET) in GPAQ MVPA than with SensWear. Nevertheless, the study found significant correlation (0.45–0.64) between these variables. They reported also unusual findings in relation to SB, which did not differ by various instruments, yet it was poorly correlated (R < 0.25). However, vigorous PA values showed high similarity (R > 0.59) [
17].
To improve the validity of GPAQ data, Metcalf et al. highlighted the utility of a Mean Squared Prediction Errors model for calibration. In this study data were collected in Ottumwa (IA, USA) and accelerometer data were predicted with a multiple regression model regarding gender, age, GPAQ PA domains by intensity and SB as covariates. The authors found weak correlation between self-reported and objectively measured data in Outcome Matching Model (R
2 = 0.025–0.177), but using Break Factor Cut-offs the Final Calibration Model showed considerable improvement (R
2 = 0.097–0.364). In both models the proportion of variance explained of vigorous PA was the highest and SB the lowest. Mean Squared Prediction Errors reduced from 66.4–98.3% to 61.3–98.6% [
19]. Majority of these studies show that, compared to other PA questionnaires, the GPAQ is more appropriate for monitoring physically active people and activities with higher energy expenditure.
While this current study focused on young adults, aging people belonging to a high risk population, should be negotiated with particular attention. Results from the GPAQ study of Hamrik et al. (carried out in the Czech Republic, a region which is socio-economically similar to Hungary) highlighted that more than 60% of the studied population across all ages could be described as sedentary, but the levels of PA decrease more with age (OR/95% CI1.011/1.005–1.017; F
age = 8.002,
p < 0.001). They reported the highest level of sedentary behaviour over 65 years [
35]. These facts indicate the need for repeated monitoring of PA through the lifespan. While GPAQ is not suitable for reporting changes in individual PA habits, it appears to be a valid tool for monitoring national strategies for PA promotion, especially for MVPA [
12].
Limitations
However, the Hungarian results confirmed that GPAQ is a valid and reliable tool to examine the Hungarian population’s physical activity level, it should be borne in mind that self-administration of data can be a challenge [
20], GPAQ as other subjective measurements based on self-reported data, can over or underestimates values of the physical activity level [
36]. On the other hand, opposite to the self-report measures accelerometer do not register the cycling, contact sport and swimming time and it was not wearing all day.
GPAQ is a widely used tool to measure the effect of interventions at population- or community level, but it is not an efficient tool to measure changes in an individual’s physical activity [
12].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.