Introduction

Despite the well-documented physical and psychological health benefits of regular physical activity [1, 2], most adults are inactive in their daily lives [3]. According to the multilevel ecological framework, interactions between individual, social, physical, and environmental factors in settings in which people live, work, and play are important for predicting physical activity behavior [4]. However, most studies examine the availability of or access to environmental features within one’s neighborhood in relation to total physical activity levels [5–7]. A growing concern about research in this area is that there is spatial or temporal uncertainty about the actual settings that exert contextual influence on the health behaviors under investigation [8]. Known as the uncertain geographic context problem (UGCoP) [9], this methodological issue is characterized lack of clarity about (1) the specific context or setting that has a direct causal influence on health-related behaviors and (2) the timing and duration of individuals’ actual exposures to these contextual influences. It is thought that the UGCoP may account for many of the inconsistencies observed in research on the effects of neighborhood built environmental features on health behaviors [10–12]. This problem may be addressed by measuring where physical activity actually takes place [13].

Understanding more about the contexts of physical activity may help to address disparities in physical activity levels among different populations, such as the significant gender gaps [14, 15]. Men spend more overall time in physical activity and are more likely to engage in vigorous intensity activities than women [16–18]. Furthermore, men are more likely to participate in group-based activities (e.g., basketball), whereas women are more likely to practice more individualized activities (e.g., swimming, planned walking) [19]. Despite this initial evidence, few studies have systematically examined gender differences in social and physical contexts of physical activity. Dunton and colleagues [20] found that men were more likely to exercise outdoors and at work, whereas women were more likely to exercise at home. However, this study used a 24-h recall methodology, which may be prone to errors and biases.

The methodological and substantive limitations of prior work in this area may be partially overcome through ecological momentary assessment (EMA). EMA utilizes electronic devices such as mobile phones to prompt real-time self-report surveys at random times throughout the day [21–24]. Thus, immediate exposures to physical and social environments can be assessed in free-living, natural settings. EMA has been used to measure the types, locations, and contexts of children’s physical and sedentary activity [25–28]. However, no known studies have employed EMA methodology to describe the contexts of adults’ physical activity and sedentary behavior. The current study used EMA with mobile phones to (1) describe where and with whom adults’ self-reported physical and sedentary activities occur (i.e., social and physical contexts); (2) determine whether adults’ objectively measured levels of physical and sedentary activities differ across social and physical contexts; (3) determine whether the perceived outdoor environmental features (i.e., safety, traffic, litter, and greenness/vegetation) are associated with objectively measured levels of physical activity in those locations; and (4) determine whether the patterns and relationships described above differ for men and women.

Methods

Participants and Procedures

This study used the baseline data (collected in 2011) from participants enrolled in Project MOBILE (Measuring Our Behaviors in Living Environments). All participants were healthy, low-active (i.e., engaged in <150 min/week physical activity) adults (ages 27–73 years) living in/around Chino, California who were able to answer electronic EMA surveys while at work. Individuals were excluded if they (a) did not speak and read fluently in English; (b) had annual household income greater than $210,000; and (c) had physical disabilities limiting physical activity. Participants were scheduled for a data collection appointment at a local community site or their home, where body measurements were taken and monitoring equipment was distributed with verbal and written instructions. This research was reviewed and approved by the Institutional Review Board at the University of Southern California.

Measures

EMA

Electronic EMA surveys were delivered through an HTC Shadow mobile phone (T-Mobile USA, Inc.) with a custom software program (MyExperience) installed (http://myexperience.sourceforge.net). All other functions of the mobile phone were disabled. The monitoring period was from 6:30 am on Saturday to 10:00 pm on Tuesday. Eight EMA surveys were prompted each day at a random time during pre-programmed intervals to ensure appropriate sampling spacing across the day. Participants were asked to complete a short question sequence on the display screen upon hearing the signal from the mobile phone. If a signal occurred during an incompatible activity (e.g., showering, driving), participants were instructed to ignore it. If a survey prompt was not answered, the mobile phone emitted up to three reminder signals at 5-min intervals, after which the survey became inaccessible until the next prompt. Survey responses were stored on the phone in an electronic file until downloaded by researchers. Project staff checked the survey response rate when participants returned the mobile phone. Participants with a low response rate (e.g., less than 70 %) would have an option to re-wear the mobile phone for another 4 days. Participants were compensated $50 for participating in the study.

The EMA questions assessed self-reported information on current activity, physical location, type of social company, and perception of outdoor environments. The response choices were designed to reflect the situations that are most likely to occur throughout adults’ daily lives as determined through pilot testing. During each EMA question sequence, participants were asked for their current activity type (see Fig. 1, screens 1–3). These items have been found to be valid measures [29]. Next, participants were asked for their current location. More specific location of Home (Indoors), Home (Outdoors), and Outdoors (not at home) was also assessed (see Fig. 1, screens 4–9). Further, participants were asked to assess their outdoor environments (see Fig. 1, screens 10–12). Participants were also asked to answer either “yes” or “no” to indicate whether they were alone. If not alone, they received a series of follow-up questions requiring “yes” or “no” responses to indicate whether they were with their spouse, child (ren), other family members, friend (s), coworkers, other types of acquaintances, or people they did not know. To reduce participant burden, each EMA item appeared 60 % of the time in a randomly programmed sequence with the exception of current activity type, which occurred during every EMA prompt.

Fig. 1
figure 1

Screen shots for EMA questions assessing current activity, physical context, and outdoor environmental features

Full size image

For the purpose of multilevel data analysis, responses indicating “Physical Activity/Exercise” and “Jogging/Running” were coded as physical activity, and those indicating “Reading/Computer,” “Watching TV/Movies,” and “Sitting” were coded as sedentary activity. The physical contexts for physical activity were recoded as either (1) Home, indoors, (2) Home, outdoors, (3) Outdoor, park, (4) Outdoor, other, or (5) Other. The physical contexts for sedentary activity were recoded as either (1) Home, bedroom, (2) Home, living room, (3) Home, others, (4) Work (indoors), or (5) Other. The social contexts were recoded as either (1) Alone, (2) With spouse only, (3) With child (ren) only, (4) With friends and/or co-workers and/or other acquaintances, or (5) Others (multiple types of social accompany). Each EMA survey was also coded for the time of day that it occurred as morning (6:30 am to 11:59 am), afternoon (12:00 pm to 5:59 pm), or evening (6:00 pm to 10:00 pm).

Objectively Measured Physical and Sedentary Activity

To assess physical and sedentary activity objectively, all participants also wore an accelerometer device (Actigraph, GT2M model) throughout the monitoring period. The device was attached to an adjustable belt and placed on the right hip, along with the mobile phone. The moderate-to-vigorous physical activity (MVPA) threshold was 2,020 activity counts per minute (equivalent to three METs) [30]. Sedentary activity was defined as less than 100 activity counts per minute [31]. All accelerometer data points were time-stamped in order to be linked with the EMA data. A 30-min window was created around each answered EMA survey, which covered the 15-min before and the 15-min after the EMA prompt. A total of zero activity counts in the 30-min window was considered as accelerometer non-wear and was therefore excluded from analyses.

Height and Weight

Project staff measured height and weight in duplicate using the electronically calibrated digital scale (Tanita WB-110A) and professional stadiometer (PE-AIM-101) to the nearest 0.1 kg and 0.1 cm, respectively. Body mass index (BMI) was calculated (kg/m2). Weight category was classified as follows: under and normal weight (BMI < 25), overweight (BMI greater than or equal to 25 and less than 30), and obese (BMI greater than or equal to 30).

Demographic Variables

Participants reported their age, gender, ethnicity, and annual household income through a paper-and-pencil survey. Annual household income was coded into quartiles (less than $40,001, $40,001-$70,000, $70,001-$100,000, and above $100,000).

Data Analyses

Multilevel analyses were conducted using SUDAAN 11.0.0. To describe the physical and social contexts of self-reported physical activity (Objective #1), only EMA entries indicating physical activity as the current activity were included in the analyses. Since a relatively small number of EMA entries reported physical activity at home (indoors) by men (n < 5), the physical context variable for this particular analysis was further collapsed into (1) Home (indoors & outdoors), (2) Outdoor, park, (3) Outdoor, others, and (4) Other. To describe the physical and social contexts of self-reported sedentary activity (Objective #1), analyses only included EMA entries indicating sedentary activity as the current activity. Separate multinomial logistic regression models were then fitted using physical and social context as the dependent variables and gender as the predictor variable (Objective #4). To test differences in objectively measured physical and sedentary activity levels across physical/social contexts (Objective #2) and by perceived environmental features (i.e., greenness/vegetation, traffic, litter, safety) (Objective #3), multilevel linear regression models were fit using MVPA and sedentary activity minutes in the 30-min window around each EMA survey as the dependent variables, and physical/social contexts and perceived outdoor environmental features as the predictor variables, separately. In a separate step, interactions between the predictor variables and gender were tested (Objective #4). All models additionally controlled for age, ethnicity (Hispanic vs. non-Hispanic), annual household income, weight category, day of the week (weekdays vs. weekend days), and time of day (morning, afternoon, vs. evening).

Results

Descriptive Statistics

Participants (N = 114) answered 82 % (range 25–100 %) of EMA prompts (n = 2,834). Of these, 110 participants had valid accelerometer data and were matched with data from the EMA prompts (n = 2,278). The likelihood of not answering an EMA prompt was not related to day of the week, time of day, sex, age, race/ethnicity, annual household income, or weight category. No device was lost or damaged. Greater details about the EMA and accelerometer data availability and sources of missing data are reported elsewhere [29]. Demographic characteristics for the sample are shown in Table 1.

Table 1 Demographic characteristics (N = 114)
Full size table

Overall, physical activity, and sedentary activity was reported as the main activity in 7.8 and 44.4 % of EMA responses, respectively (see Fig. 2). Of the EMA-reported physical activity, 50.7 % were walking, 6.3 % were running/jogging, 5.0 % were using cardiovascular equipment, 4.5 % were weightlifting/strength training, 3.2 % were bicycling, and 30.3 % were other types of physical activity. Of the EMA-reported sedentary activity, 35.9 % were reading/using computer, 33.0 % were watching TV/movies, and 31.1 % were sitting. The likelihood of EMA-reported sedentary activity differed by time of day (p < .01). Sedentary activities were more likely to occur in the morning as compared with the afternoon and evening. Within the 30-min window surrounding each answered EMA prompt, participants on average engaged in 0.8 min (SD = 2.4) of MVPA and 20.0 min (SD = 6.4) of sedentary activity, as measured by accelerometers.

Fig. 2
figure 2

Overall EMA reported main activities

Full size image

Physical and Social Contexts for Physical and Sedentary Activity

Figure 3 shows the proportions of physical and sedentary activity reported in each physical context. Half of the EMA-reported physical activity took place either indoors (24 %) or outdoors (29 %) at home. When reported in indoor home locations, the majority of the physical activity took place in the bedroom (32.3 %) or living/family room (29 %). When reported in outdoor home locations, most of the physical activity took place on a driveway (33.3 %) or in a yard (30.6 %). Thirty-seven percent of the EMA-reported physical activity occurred at an outdoor (not at home) location. Thirty-eight percent of these outdoor physical activity occurred in an area with a lot of trees and plants, and 16 % was in an area with no trees or plants. Forty-two percent of the outdoor activity was reported in a place with no traffic, and only 11 % was reported in a place with a lot of traffic. The majority of the outdoor activity (76 %) occurred in a place where adults felt very safe, only 4 % outdoor activates occurred in a place where adults felt unsafe or somewhat unsafe. Sixty percent of the outdoor activity was reported in a place with no litter, and only 2 % outdoor activity was reported in a place with a lot of litter. Sixty-nine percent of the outdoor activity took place more than 1 mile away from home, and 18 % took place less than half mile away from home. Most (76 %) of these outdoor activity locations were travelled to by car, 20 % were by walking, and only 4 % were by biking.

Fig. 3
figure 3

EMA reported physical contexts for physical and sedentary activities

Full size image

The physical contexts for sedentary activity differed between weekdays and weekend days (adj. Wald F = 9.73, p < .01). Specifically, adults were more likely to engage in sedentary activity in bedroom and living room during weekend days (pred. prob. = .41, SE = .032; pred. prob. = .42, SE = .038; respectively) than in weekdays (pred. prob. = .28, SE = .033; pred. prob. = .29, SE = .036; respectively), and at work during weekdays (pred. prob. = .29, SE = .040) than in weekend days (pred. prob. = .04, SE = .016). There was no difference between weekdays and weekend days for the physical contexts of physical activity.

Figure 4 shows the proportions of physical and sedentary activity reported in each social context. When participants reported engaging in physical activity while alone, 41.9 % were walking, 7.0 % were running/jogging, 7.0 % were weightlifting/strength training, 7.0 % were bicycling, 4.7 % were using cardiovascular equipment, and 32.6 % were some other types of physical activity. When participants reported being physically active with family members and/or friends (the multiple categories), the majority of that physical activity was either walking (49.1 %) or weightlifting/strength training (14.7 %). When participants reported being sedentary alone, 41.9 % were reading/using computer, 30.0 % were watching TV/movies, and 28.2 % were sitting. When participants reported being sedentary with family members and/or friends (the multiple categories), half (53.6 %) of the sedentary activities were watching TV/movies, 25.1 % were sitting, and 21.2 % were reading/using computer.

Fig. 4
figure 4

EMA reported social contexts for physical and sedentary activities

Full size image

The social contexts for sedentary activity also differed between weekdays and weekend days (adj. Wald F = 6.39, p < .01). Adults were more likely to engage in sedentary activity with spouse and children during weekend days (pred. prob. = .17, SE = .028; pred. prob. = .15, SE = .029; respectively) than on weekdays (pred. prob. = .09, SE = .021; pred. prob. = .11, SE = .020, respectively), and alone or with co-workers/friends during weekdays (pred. prob. = .46, SE = .039; pred. prob. = .19, SE = .032, respectively) than on weekend days (pred. prob. = .35, SE = .040; pred. prob. = .09, SE = .018, respectively). There was no difference between weekdays and weekend days for social contexts of physical activity.

Gender Differences in Physical and Social Contexts for Physical and Sedentary Activity

As shown in Fig. 5, there was a significant gender difference in the physical context of physical activity (adj. Wald F = 4.24, p < .01). Women were more likely to report engaging in physical activity when they were at home (pred. prob. = .61, SE = .061) than men (pred. prob. = .24, SE = .077), and men were more likely to report engaging in physical activity when they were outdoors at a park (pred. prob. = .32, SE = .085) than women (pred. prob. = .10, SE = .034). No gender differences were found in the social contexts of physical activity.

Fig. 5
figure 5

Predicted probability of physical context of physical activity by gender. Adjusted for age, ethnicity (Hispanic vs. non-Hispanic), weight category, annual household income, day of week (weekday/weekend day), and time of day (morning/afternoon/evening)

Full size image

The social contexts of sedentary activity differed by gender (adj. Wald F = 2.46, p < .05; see Fig. 6). When with friends/co-workers/acquaintances, women were more likely to report engaging in sedentary activity (pred. prob. = 17, SE = .027) than men (pred. prob. = .07, SE = .022). There were no gender differences in the physical contexts of sedentary activity.

Fig 6
figure 6

Predicted probability of social context of sedentary activity by gender. Adjusted for age, ethnicity (Hispanic vs. non-Hispanic), weight category, annual household income, day of week (weekday/weekend day), and time of day (morning/afternoon/evening)

Full size image

Differences in Objectively Measured Physical and Sedentary Activity Levels by Physical and Social Context

Physical activity levels differed across physical contexts (p < .01; see Table 2). Adults engaged in fewer MVPA minutes when they were at home indoors compared with all the other physical contexts. The interaction between physical context and gender was also significant (p < .05). Women engaged in more MVPA minutes when outdoors at home than men whereas men engaged in more MVPA minutes when outdoors at a park than women. MVPA minutes did not differ across social contexts for men or women.

Table 2 Differences in physical activity minutes across contextsa
Full size table

As shown in Table 3, sedentary activity levels differed across physical contexts (p < .01). Adults engaged in more sedentary minutes when they reported being at home in the bedroom than any other locations, except when at work. The physical context by gender interaction was significant (p < .05). When at home (other than in the bedroom or living room), men engaged in more sedentary minutes than women whereas when at work, women engaged in more sedentary minutes than men. Sedentary minutes did not differ across social contexts between genders.

Table 3 Differences in sedentary activity minutes across contextsa
Full size table

Relation Between Perceived Outdoor Environmental Features and Objectively Measured Physical Activity Level

Perceptions of traffic and litter within settings were unrelated to the number of MVPA minutes simultaneously recorded in those settings. However, there was a significant interaction between perceived greenness/vegetation and gender (adj. Wald F = 3.95, p < .05, see Fig. 7). MVPA was higher in outdoor environments with higher levels of perceived greenness/vegetation among men but not women.

Fig. 7
figure 7

Predicted mean MVPA (moderate to vigorous physical activity) minutes by gender and self-reported outdoor vegetation. Adjusted for age, ethnicity (Hispanic vs. non-Hispanic), weight category, annual household income, day of week (weekday/weekend day), and time of day (morning/afternoon/evening). MVPA minutes were measured using accelerometer in ±15 min window around each EMA survey prompt

Full size image

Discussion

The current study used a novel methodology, EMA, to examine where and with whom adults’ physical and sedentary activity occurs during their daily lives. Home was the most commonly reported physical context for both physical and sedentary activity. Most of the reported physical and sedentary activity occurred when participants were alone. When alone, the most commonly reported physical activity was walking and the most commonly reported sedentary behavior was reading/using computer.

The combine use of electronic EMA surveys and accelerometers allows researchers to more clearly understand the different types of activities people perform during their everyday lives, and where and with whom these activities take place. This information could be useful for designing targeted intervention. For example, similar to previous research, we found that men perform less physical activity at home as compared with women [20, 32]. With the advantage of EMA data, we were able to explore this finding further. A post hoc analysis of our current data showed that when women reported engaging in physical activity in outdoor areas of their homes, 36 % occurred in the driveway, 25 % occurred in the yard, 14 % occurred on a deck or patio, 4 % occurred in the pool, and 21 % occurred in other outdoor areas of the home. More than half of these self-reported activities involved walking and more than half occurred alone. Given the large proportion of these activities were reported in the driveway or yard, it is possible that women answered the randomly prompted EMA surveys at the very beginning or tail end of going for a walk in the neighborhood (before leaving or immediately after returning to their property). If so, these results would be in line with previous research suggesting that women are more likely to perform solo walking activities than men [19].

Further post hoc analyses found that when men reported being physically active outdoors at a park, they reported being with another person on 83 % of the EMA surveys and with a dog on 50 % of the surveys. Furthermore, males engaged in higher levels of MVPA in settings perceived to have higher levels of greenness/vegetation. These results may begin to reveal a pattern that men prefer socially oriented physical activities and dog-walking in outdoor green spaces such as parks. It is also possible that when at the park, men are more likely than women to engage in moderate-to-vigorous intensity physical activity, as suggested by previous evidence [33].

The perceived amount of traffic and litter was unrelated to the outdoor physical activity levels. It may partly due to the lack of variation for these two environmental factors in our sample (64 % of the EMA-reported outdoor physical activity occurred in a place with no or a little traffic, and 84 % of these outdoor activities took place in a place with no or a little litter). This result might also imply that adults tend to choose a place that is free from traffic and litter to engage in physical activity.

Despite the advantages of using EMA methodology to assess contexts and accelerometer to objectively measure activity level, this study had a few limitations. First, our sample consisted of adults from low-to-middle income families who did not engage in much physical activity. Although this may not be a representative sample of the general population, this is a high-risk group for obesity and other obesity-related health problems [34]. The gender distribution in the sample was not even. Due to the relatively smaller proportion of male participants, we may not have enough power to detect small gender differences. Further, since our participants were limited to a particular geographic area (Southern California), the results describing gender differences may not be applicable to other populations in other areas of the USA or the globe. Instead, our results demonstrate the novel opportunities offer by EMA method to assess contextual information as a way to explore disparities in physical activity levels among different populations. Similarly, we only assessed the perceived environmental features when participants reported they were outdoors, not at home (n = 157). Thus, this limited number of EMA episodes may not be sufficient to test the relationship between perceived environmental factors and physical activity level. Another limitation is that we combined reading with using computer as one choice when reporting current activity. While both activities are considered as sedentary, they might have different patterns and health implications. Furthermore, the data were collected only for 4 days. This short monitoring period might not be fully representative of adults’ daily lives.

Overall, this study demonstrates the potential for using real-time EMA to capture context-specific physical and sedentary activity among adults. Future research should investigate the conditions under which gender differences in physical and sedentary activity contexts are based upon preferences versus opportunities. Interventions could then develop strategies to encourage activity in locations where individuals are naturally inclined to be more active, such as capitalizing on men’s predisposition for park-based activity or women’s tendency to engage in less sedentary behavior at home.