Japanese Activity Questionnaire Passes Its Biggest Test Against Wearable Sensors

For decades, the world’s public health machinery has run on a deceptively simple tool: a questionnaire. When the World Health Organization wants to know how active a nation is, when ministries of health design campaigns against sedentary living, and when researchers compare physical activity levels from Tokyo to Toronto, they rely overwhelmingly on people’s own accounts of how much they move. But self-reports are only as good as their validation, and a new study from Japan has now delivered one of the most culturally revealing audits yet of the Global Physical Activity Questionnaire, the WHO’s flagship surveillance instrument. The verdict is nuanced, gendered, and unexpectedly illuminating about how culture shapes not just what we do, but what we remember doing.

The study, published in the Journal of Activity, Sedentary and Sleep Behaviors, was led by Chiaki Tanaka of Tokyo Kasei Gakuin University and the National Institute of Health and Nutrition, together with John J. Reilly of the University of Strathclyde and Shigeho Tanaka of the National Institute of Health and Nutrition and Kagawa Nutrition University. Their goal was straightforward in principle: to determine whether the Japanese version of the Global Physical Activity Questionnaire, known as the GPAQ-J, actually captures what Japanese adults do when researchers are not asking them about it. To find out, the team recruited 266 adults, 181 women and 85 men, all parents of children attending primary schools in Tokyo and Kyoto, and asked them to complete the questionnaire while simultaneously wearing a sophisticated triaxial accelerometer for at least a week.

The device at the heart of the study, the Omron Active style Pro HJA-350IT, is not an ordinary fitness tracker. Unlike most accelerometers used in validation research, it can distinguish between ambulatory activities, such as walking and running, and non-ambulatory activities, such as sweeping floors or lifting objects at work. It does this by calculating the ratio of unfiltered to filtered acceleration signals, a classification approach that has achieved near-perfect accuracy for a battery of laboratory-tested activities. The device also predicts metabolic equivalents, or METs, with sufficient precision that its estimates of total energy expenditure have matched the gold-standard doubly labelled water method, while many competing accelerometers substantially underestimate energy expenditure. This makes it an unusually trustworthy arbiter for judging a questionnaire’s honesty.

The questionnaire itself works by dividing daily life into three domains: work, transportation, and recreation. Respondents estimate how many days per week and how many minutes per day they spend in moderate and vigorous activity within each domain, and they answer a separate question about daily sitting time. The Japanese version was created through careful translation and back-translation, and it has already been deployed in national surveys, including the Sasakawa Sports Foundation’s National Sports-Life Survey. Yet until now, no study had tested whether the GPAQ-J’s estimates of activity duration correspond to what sensors actually record in Japanese adults. The stakes are considerable, because Japan’s national health strategy, Health Japan 21, depends on accurate activity surveillance, and the country’s previous targets, such as 9,000 daily steps for working-age men and 8,500 for women, were met by only about a third of adults.

The headline finding is that the questionnaire works, but unevenly. Among men, total moderate-to-vigorous physical activity reported on the GPAQ-J correlated significantly with accelerometer-measured total MVPA, with a Spearman rank correlation coefficient of 0.375, and even more strongly with sustained bouts of at least ten minutes of MVPA, at 0.419. Among women, the correlations were positive and significant but weaker, at 0.165 for total MVPA and 0.182 for bouts. In the conventional language of validation studies, these values represent weak to moderate validity, which is broadly consistent with what GPAQ studies in other countries have found. The men’s results were comparable to those reported in the United States, China, and the United Kingdom, while the women’s fell somewhat below that range.

The most striking pattern emerged when the researchers broke the results down by activity domain. For men, the transportation section of the questionnaire showed a remarkably strong correlation of 0.615 with accelerometer-measured ambulatory MVPA, the highest association in the entire study. Transportation activity, in other words, is something Japanese men both do and remember doing with reasonable fidelity. For women, by contrast, only the recreation domain showed a significant correlation, and only with ten-minute bouts of MVPA. The work and transportation domains in women’s reports bore little statistical relationship to what the sensors recorded. The researchers suggest a cultural explanation rooted in the division of labor: Japanese women spend far more time than men on household activities, averaging seven hours and thirty-four minutes per day on housework and childcare according to government survey data, compared with one hour and twenty-three minutes for men. Household chores are predominantly non-ambulatory, hovering near the three-MET threshold that separates moderate activity from light activity, and they are notoriously difficult to recall and classify. A woman mopping a floor may not think of it as physical activity at all, even though the compendium of activity energy costs assigns many chores values above three METs.

Sitting time told a parallel story of gender divergence. Among men, the questionnaire’s sitting-time question correlated moderately with accelerometer-measured sedentary time, at 0.332, but among women the correlation was essentially zero, at 0.004. Moreover, both sexes dramatically underestimated how long they sit, with questionnaire-based estimates falling well below sensor measurements. This matters because the WHO issued its first recommendations on sedentary behavior in 2020, urging adults to limit sedentary time for substantial health benefits, and surveillance systems that cannot measure sitting accurately will struggle to track progress against those recommendations.

One of the study’s more forward-looking findings concerns the structure of the questionnaire itself. The GPAQ asks respondents about activity accumulated in bouts of at least ten minutes, a framing that reflected older exercise guidelines. The 2020 WHO guidelines removed the bout requirement, recognizing that activity of any duration counts toward health benefits. Encouragingly, the study found that the questionnaire’s estimates of MVPA fell between the accelerometer’s measurements of total MVPA and its measurements of ten-minute bouts, suggesting that the GPAQ-J captures something meaningful about overall activity volume, not merely its structured, sustained form. This gives the instrument a degree of resilience in an era when the very definition of health-promoting activity has been liberalized.

The authors are candid about the limitations. The sample consisted of parents of primary school children in two large Japanese cities, with a relatively favorable socioeconomic profile and a female employment rate roughly ten points below the national average for that age group, so generalization requires caution. Accelerometers cannot register swimming or cycling, and participants did not wear the devices during every waking moment. Reliability data, which would test whether the questionnaire gives consistent answers on repeated occasions, could not be collected without overburdening participants. Still, the participants’ body sizes closely matched Japanese adult averages, and the study design, pairing a culturally adapted questionnaire with a device capable of separating household from locomotor movement, offers a template for validation work elsewhere.

The broader significance of the study extends well beyond Japan. Physical inactivity is a leading modifiable risk factor for noncommunicable disease worldwide, and international comparisons of activity levels rest on the assumption that the same questionnaire means the same thing in different cultures. The Japanese data show that assumption is only partly safe: where people’s activity is embedded in paid work and commuting, as it tends to be for men in this sample, the questionnaire performs well; where it is embedded in household labor that respondents may not mentally categorize as exercise, as for women, it underperforms. For surveillance systems, the practical lesson is that the GPAQ-J can be used for relative assessments of moderate-to-vigorous activity in Japanese adults, ranking individuals and tracking population trends, but its sitting-time estimates should be interpreted with care, particularly for women. For the global research community, the study is a reminder that the most important sensor in physical activity science may still be human memory, and that memory is shaped by culture as much as by muscle.

Subject of Research: Validation of the Japanese version of the Global Physical Activity Questionnaire against accelerometer-measured physical activity and sedentary time in Japanese adults

Article Title: Correlations between the Japanese version of the Global Physical Activity Questionnaire and accelerometer-measured physical activity and sedentary time in adult women and men

Article References: Tanaka, C., Reilly, J. J., & Tanaka, S. (2024). Correlations between the Japanese version of the Global Physical Activity Questionnaire and accelerometer-measured physical activity and sedentary time in adult women and men. Journal of Activity, Sedentary and Sleep Behaviors, 3(1), Article 29. https://doi.org/10.1186/s44167-024-00069-1

Image Credits: AI Generated

DOI: 10.1186/s44167-024-00069-1

Keywords: GPAQ, physical activity, accelerometer, sedentary behavior, questionnaire validity, Japan, gender differences, MVPA, sitting time, WHO guidelines, public health surveillance, triaxial accelerometer