ABSTRACT
Objectives This study aims to evaluate the gender inequalities in accelerometer-based physical activity (PA) across different age groups using data from five Pelotas (Brazil) cohorts.
Methods The data comes from four birth cohort studies, covering all live births in the urban area of Pelotas for each respective year (1982, 1993, 2004, and 2015), and the ‘Como vai?’ cohort study focusing on 60 years and above. Raw accelerometry data were collected on the non-dominant wrist using GENEActive/Actigraph devices and processed with the GGIR package. Overall PA was calculated at ages 1, 2, 4, 6, 11, 15, 18, 23, 30, and 60+ years, while moderate-to-vigorous PA (MVPA) was calculated from six years onwards. Absolute (difference) and relative (ratio) gender inequalities were calculated and intersectionality between gender and wealth was also evaluated.
Results The sample sizes per cohort ranged from 965 to 3462 participants. The mean absolute gender gap was 19.3 minutes (95%CI: 12.7; 25.9), with the widest gap at 18 years (32.9 minutes; 95%CI: 30.1; 35.7) for MVPA. The highest relative inequality was found in older adults (ratio 2.0; 95%CI 1.92 to 2.08). Our intersectionality results showed that the poorest men being the most active group, accumulating around 60 minutes more MVPA per day compared with the wealthiest women at age 18.
Conclusion Men were more physically active than women in all ages evaluated. PA gender inequalities start at an early age and intensifies in transition periods of life. Relative inequalities were marked among older adults.
What is already known on this topic Gender inequalities in physical activity have been reported globally, but most of the evidence is focused in adolescents and young adults. The literature lacks studies on children and older adults.
What this study adds We present gender inequalities in accelerometer-based physical activity across several age groups, from 1 year olds to older adults.
How this study might affect research, practice or policy Ou study provides a comprehensive description of gender inequalities, identifying key age groups for intervention.
BACKGROUND
Men and women present different patterns of morbidity and mortality[1]. Apart from biological differences, socially constructed gender norms, and historical gender inequalities in access to health resources, have harmful effects on women’s health, including physical activity (PA)[2]. Therefore, gender norms shape behaviors from early childhood, influencing not only on children’s behaviors but the encouragement received for active play or physical activities[3,4], and contributing to the lower levels of PA observed in girls compared to boys[5]. These gender differences in PA increase with age,[6] since girls experience less socio-ecological beneficial factors at the individual, family, school, and environmental levels, exacerbating gender inequalities in adolescence[6,7]. In adulthood, for most women the double burden of having a job and taking on most of the housework, motherhood, and lack of family support also contribute to lower levels of PA[8].
Gender or sex inequalities in PA levels have been reported in different age groups and countries with different economic contexts[9–13]. Globally, a gender difference in the prevalence of meeting PA recommendations of 7.1 percentage points was reported in adolescence[9], and around 8.3 percentage points in adulthood[14]. This inequality ranges from the first year of life[15] to older age[16]. However, despite evidence that sex is a consistent correlate of PA at different ages,[17] most studies investigating gender inequalities in PA levels focused in adolescents[1,11,12] or adults[10,18]. The literature still lacks studies exploring these inequalities in children and older adults.
Moreover, most studies measure PA through self-report[10–13,18], which is susceptible to social-desirability and information bias. This study aims to evaluate the gender inequalities in PA across different age groups using accelerometer data from the Pelotas (Brazil) cohorts, including birth cohorts and the ‘Como vai?’ cohort study addressing older adults.
METHODS
Participants
We are using data from five cohort studies conducted in Pelotas (Brazi), a city a population of 343.826 inhabitants in 2021 and a GDP per capita of R$ 27,586.96 (U$ 5,290.15)[19]. The first birth cohort (BC) was established in 1982 and additional cohorts were set up in 1993, 2004 and 2015, following all live births in the urban area of Pelotas for each respective year. These studies aim to monitor and examine associations between lifestyle, environmental and social exposures, and health outcomes across the life course and to examine differences across generations. Furthermore, in 2014 The ‘Como vai?’ study was initially designed as a cross-sectional survey of older adult population living in the urban area of Pelotas, which afterwards became a cohort study. Detailed information on each cohort can be found elsewhere[20–26].
Our study included only follow-ups with information on accelerometry, which resulted in samples with ages: one, two, four (2015 BC); six, 11, 15 (2004 BC); 18, 22 (1993 BC); 30 (1982 BC); and 60 years or more (‘Como vai?’).
Patient involvement
Not applicable.
Equity, diversity, and inclusion
Given the populational characteristic of the Pelotas cohort studies, representation by gender, race/ethnicity/culture, socioeconomic level, and marginalized groups is guaranteed by design. Also, our research facilities are accessible to participants’ needs, and our research team provides household interviews and measurements when necessary. Furthermore, our study addresses equity by investigating gender and wealth inequalities.
Accelerometry
The protocol for the Pelotas cohorts comprises 24 hours of accelerometer data collection for all ages, including sleep and water activities. For one- and two-year-olds the accelerometer was placed on the left wrist, and data was collected for two days[27]. Four-year-old children wore the device on their left wrist for seven days, while older participants wore it on the non-dominant wrist for seven days. In the six-, 18- and 30-year follow-ups of 2004, 1993, and 1982 birth cohorts, participants wore the accelerometer on the non-dominant wrist from four to seven days[28].
In the earlier accelerometry data collections (six, 18, 30, and 60+ years) GENEActive (Activinsights, Kimbolton, UK) was used, while in more recent follow-ups the device was the Actigraph GT3X/GT3X+ (Actigraph, Pensacola, USA). High comparability between these brands has previously been reported when raw data is analyzed[29,30]. For this reason, we analyzed the raw signal, which are less affected by company-specific filters.
Data were processed with the GGIR package (version 2.2) in R software[31]. This process included verification of sensor calibration error, detection of sustained abnormally high values, and non-wear detection[32,33]. Vector magnitude was calculated using Euclidian Norm Minus One (ENMO) metric to summarize acceleration from axes x, y, and z into a single-dimensional signal (). Data was collected and analyzed in a five-second epoch.
To define the minimum number of measurement days included, we used the Spearman-Brown formula, which is based on the Intraclass Correlation Coefficient (ICC) combined with the Spearman-Brown prophecy. Balancing the reliability values and the number of losses, we considered a minimum of one day for one-year-old infants, two days for four-, six-, 18-, and 30-year-old participants, and three days for other ages. This method was described in detail in previous publications[27,34].
The two main outcome variables used in our study were overall PA and moderate to vigorous PA (MVPA). Overall PA was defined as the average Euclidian Norm Minus One (ENMO) per day, expressed in mg. MVPA was defined as the average of minutes spent above the threshold of 100mg per day[35], using five-minute bouts. We presented both outcome measures to better represent PA behavior in our sample, including the total volume of PA (overall PA) and structured PA (MVPA). The MVPA variable was calculated for individuals aged six years or older.
Statistical analysis
To illustrate the gender gap in PA, we presented the means of overall PA and MVPA in men and women (sex assigned at birth) for each follow-up in an equiplot (equidade.org/equiplot). The means, 95% confidence intervals (CI) and T-test results are shown in the supplementary material.
Absolute inequalities (gender differences) were calculated by subtracting women’s mean PA (overall PA and MVPA) from men’s mean PA, with positive values indicate a higher mean among men. Relative inequalities were calculated as the ratio between women’s and men’s mean PA, where values above 1 indicate higher PA mean among men. For both absolute and relative inequalities, 95%CIs were calculated using the bootstrap method, which provides the standard errors based on a resampling strategy.
We also analyzed our data stratified for socioeconomic status, aiming to illustrate the PA intersectionality between gender and wealth. The concept of intersectionality, first introduced by Collins (1990)[36], refers to the simultaneous overlapping of multiple forms of inequalities[37]. We used a wealth index using items such as house characteristics and household assets, extracting the first component of a principal component analysis[38]. This component was divided into quintiles, with the first (Q1) representing the poorest group and the last (Q5) representing the wealthiest group. We present means of overall PA and MVPA variables in Q1 and Q5 for men and women in an equiplot. The means, 95%CI, and p-values obtained through t-tests for the intersectionality analysis are available in the supplementary material.
We used a meta-analysis approach with random effect to graphically represent gender inequalities in overall PA and MVPA for each age group. Also, a sensitivity analysis was carried out to verify whether the gender gap is influenced by the amount of PA. To answer this research question, we divided the overall PA (mg) and minutes of MVPA in deciles and presented the gender gap in the 10% least active (1st decile) and 10% most active (10th decile). The results of the sensitivity analysis are presented in the supplementary material.
All statistical analyses were performed using Stata 17 (StataCorp. 2021. College Station, TX: StataCorp LLC.)
RESULTS
The present study includes data from five Pelotas cohort studies, comprising, on average, 2590 participants per follow-up with valid accelerometry data (Table 1). Sample size across cohorts ranged from 965 in the ‘Como vai?’ study to 3462 in the 18-year follow-up (1993BC). Table 1 shows the comparison between the original cohorts and the analytical samples, stratified by gender and wealth quintiles (1st and 5th). The 30-year follow-up (1982 BC) had a higher proportion of women (p=0.012) in the analytical sample. None of the remaining comparisons were statistically significant, indicating that the analytical sample is representative of the original cohort when considering gender and wealth distribution.
The gender gap for overall PA and MVPA is illustrated in Figure 1. For overall PA, the gap starts among 1-year-old infants and increasingly widens as PA levels rise, up to 11 years, when the gender gap peaked. There is a drastic decrease in overall PA between 11 and 15 years for both genders, although men still presented higher levels of PA, followed by some stability throughout adulthood. Among older adults, the PA decreases to a lower level than in infants and there is no longer a gender gap (Figure 1 A).
Regarding MVPA (Figure 1 B), from 6 to 15 years there is a declining pattern for PA with stability in the gender gap. At the age of 18, adolescent men showed an increase in MVPA, which is not followed by women, resulting in the widest gender gap. Between the ages of 22 and 30 years, both MVPA and gender gap show little variation, and a steep decrease is observed for older adults, although the gender gap remains, with older men spending more time in MVPA than older women.
The absolute gender inequalities for overall PA and MVPA are described in Figure 2. For overall PA (Figure 1 A), the widest gender gap was found at 11 years (Difference 9.8 mg; 95% CI 8.7 to 11.0), while for MVPA (Figure 1 B) the widest difference was found at 18 years (Difference 32.9 minutes; 95% CI 30.1 to 35.7). For both overall PA and MVPA, men had higher PA than women in virtually all ages, although there was no gap between men and women in overall PA among older adults (Figure 2 A).
Regarding relative gender inequalities (Figure 3 A and B), 18-year-olds presented the highest gender inequalities for overall PA (ratio 1.24; 95%CI 1.22 to 1.26), while older adults presented the lowest inequalities (ratio 1.02; 95%CI 0.98 to 1.07) although not statistically significant. On the other hand, for MVPA older adults presented the highest relative inequalities (ratio 2.0; 95%CI 1.92 to 2.08), while 6-year-olds had the lowest inequalities (ratio 1.47; 95%CI 1.40 to 1.54).
Figure 4 shows the gender gap for PA across the first (Q1, 20% poorest) and last (Q5, 20% richest) wealth quintiles. The double stratification for overall PA showed the higher acceleration among the poorest groups in all ages, although the wealthiest men have similar overall PA levels to the poorest women from the age of 11 years onwards. The highest gaps are shown at ages 11 and 18 years, with the poorest men showing around 20 mg of difference from the wealthiest women (Figure 4 A). For MVPA, during childhood and early adolescence (1 and 11 years) boys showed more MVPA than girls, regardless of wealth. From 15 years to adulthood, the poorest women show similar results to the wealthiest men. The highest gap is at age 18 years when the poorest men show around 60 minutes of MVPA more than the wealthiest women.
Gender inequalities in overall PA and MVPA for the highest and lowest PA deciles are shown in Supplementary Figure 1. The highest decile of overall PA (D10) showed increasing gender inequalities from 1 to 11 years, followed by stability during adulthood and a drastic decline among older adults. The widest gap for overall PA in D10 was observed at age 11, with approximately 20 mg of difference between boys and girls. The MVPA gap in D10 also increased with age until the end of adolescence (18 years), when the highest gap was found (~70 minutes) and narrowed thereafter.
DISCUSSION
This work describes the gender gap for PA in five Pelotas cohort studies, including a broad age range of participants. We attempted to describe the patterns of absolute and relative PA gender inequalities in multiple age groups by exploring overall PA and MVPA, aiming at representing both everyday activities and structured PA, such as sports and exercises. Our findings showed that both overall PA and MVPA were consistently higher among men.
We presented absolute and relative measures of inequalities. While absolute measures have easier interpretation, relative measures can better express proportional differences, allowing comparisons at different scales. They are complementary because their interpretability differ according to the magnitude of outcomes occurrence49. For example, when there is a lower occurrence there could be marked relative inequalities, but modest absolute inequalities. In our case, older adults presented a low level of MVPA so the absolute inequality measure identified a difference of only 7 minutes between men and women. On the other hand, the relative inequality measure identified a mean of MVPA in men 2 times higher than in women for this group - the highest relative inequality among all age groups. This does not mean that one measure is better than the other, but prefereably, both measures should be shown to improve interpretability and better inform policymakers’ decisions[39].
It is important to underline that overall PA measured by accelerometry reflects any activity performed throughout the day, including daily routine activities and intermittent and sporadic movements, and not just structured PA. Also, PA described in mg presents even less interpretability of the data, hampering its translation into daily activities. However, recent evidence has shown that, among inactive adults, the minimum clinically important difference of approximately 1 mg corresponds to about 5 to 6 min of brisk walking per day, and an increase in average daily acceleration of this magnitude is associated with reduced risk of all-cause mortality (Hazard Ratio: 0.95; 95%CI: 0.94; 0.96)[40]. When applying this example to our results, the mean absolute gender gap of 5 mg found in the present study, would correspond to approximately 25 to 30 minutes of daily brisk walking.
The PA gender gap starts early in life, with inequalities already found in one-year-old infants. The intrinsic societal restrictive gender norms are the main drivers of this discrepancy in PA, which indirectly shapes how parents raise their children, including how to dress a young girl to a play-day, which toys are gifted to boys and girls, and above all, the level of incentive to active play they receive[41]. As a result, young girls often receive less favorable influences at the individual, family, school, and environmental levels[42]. When discussing PA promotion at early ages it is imperative to highlight the tracking effect of PA, where behaviours acquired during childhood are more likely to be carried out to adolescence and adulthood[43].
Another relevant finding is that transition phases, from childhood to adolescence (11 years old) and from adolescence to early adulthood (18 years old), showed the widest gender inequalities for overall PA and MVPA, respectively. Some of the barriers surrounding the women’s PA might be more evident in adolescence, especially body dissatisfaction and feeling uncomfortable during the practice, in addition to less social support from peers, family, and teachers than their male counterparts, and less access to safe and welcoming facilities, equipment, transport, and training[44,45]. These factors shape female attitudes towards PA throughout adolescence and could affect this behavior throughout life[44] and this might be intensified by transition life events, such as entering university and/or the job market, and above all motherhood, which often lead women to accumulate even more barriers to PA practice[46,47].
Our results on the PA’s intersectionality between gender and wealth showed higher acceleration among the poorest groups, with the 20% richest women being the least active group, and the 20% poorest men being the most active group. To better interpret this result its imperative to understand that accelerometry-based overall PA includes all PA domains: leisure time, occupational, commuting, and domestic. This is a limitation of accelerometry, since this measure lacks contextual information on the domains of PA, and then we observe controversial wealth patterns for this behavior[48]. In this sense, the literature consistently reports a higher level of domestic, occupational, and commuting PA among socioeconomically vulnerable populations when considering self-reported measures[49]. On the other hand, the richest groups usually have a privileged position in PA practice, being able to pay for private and expensive activities, and taking place during their leisure time[50]. The sensitivity analysis revealed interesting results when stratifying PA levels by gender and the amount of PA. For both outcomes (overall PA and MVPA), we identified a clear pattern of a higher gender gap when looking at the most active group. These results highlight that women, in general, present a much lower level of PA than men, even among the most active group. In other words, the highest decile of PA is a very selected part of the sample (with an average of close to 100 minutes of MVPA). However, we identified a huge difference between most active men and women already at younger ages. Even the most active women are not close to most active men. These findings suggest that the gender gap is also evident in women with greater opportunities, support, and facilitators for PA practice.
Limitations
Because we described data from five different cohort studies, this analysis is not completely longitudinal. PA levels are from different generations, consequently, conclusions should be based on descriptive assumptions of a wide age range instead of a lifecycle analysis. Also, although raw data minimizes the differences between the two device brands used, some residual distinctions could remain. Howerver, since men and women wore the same device, gender inequalities were not affected. Finally, at 15 years, the follow-up was interrupted by the COVID-19 pandemic, which could have impacted PA results.
Policy implications
Our findings demonstrated that PA gender inequalities start at an early age and intensify in transition periods of life, but also, relative gender inequalities are marked among older adults. Actions to promote PA in women must consider a life cycle strategy, starting at an early age, intensifying actions at transition periods, but still being mindful of older adults’ needs.
Competing interests
None declared.
Contributorship
LICR, AW, and DT drafted substantial parts of the manuscript. LICR also performed data analysis and interpretation. HG, FW, BGCS, LT, IS, AB, AM, PCH, MD, UE, RB, IC served as scientific advisors. All authors critically reviewed and approved the final version of this manuscript.
Data Availability
All data relevant to the study are included in the article or uploaded as supplementary information.
Funding
This article is based on data from four “Pelotas Birth Cohort” studies conducted by Postgraduate Program in Epidemiology at Universidade Federal de Pelotas with the collaboration of the Brazilian Public Health Association (ABRASCO). The cohorts were funded by Wellcome Trust, The International Development Research Center (IDRC), World Health Organization, Overseas Development Administration, European Union, National Support Program for Centers of Excellence (PRONEX), Children’s Pastorate, Fundação de Amparo a Pesquisa do Estado do Rio Grande do Sul (FAPERGS), the Brazilian National Research Council (CNPq), and the Brazilian Ministry of Health.
Ethical approval
The 1982, 1993, and 2004 Pelotas birth cohorts study protocols were approved by the Medical School Ethics Committee of the Federal University of Pelotas (respective registration numbers: 16/12; 1.250.366; 40601116) and the 2015 birth cohort, by the School of Physical Education Ethics Committee of the Federal University of Pelotas (registration 26746414.5.0000.5313). The ‘Como vai?’ study protocol was approved by the Research Ethics Committee of the Faculty of Sciences of the Federal University of Pelotas (registration 24538513.1.0000.5317). The participation of the individuals in each study was voluntary, and informed consent was obtained from all participants.
Data sharing
All data relevant to the study are included in the article or uploaded as supplementary information.
Acknowledgements
The authors would like to thank all members from the Grupo de Estudos e Pesquisa em Acelerometria (GEPEA) for the valuable scientific discussions during the conception of this manuscript.