Weather conditions, including high temperatures and precipitation levels, have been identified as barriers to participation in physical activity [1
]. While physical activity generally peaks in summer [1
], a lower level of physical activity in summer was noted in Texas residents when the average temperature of the study month was 29 °C (84 °F) [4
]. Another study showed a moist tropical climate may be one of the strongest deterrents against physical activity in the U.S., reducing the percent of adults meeting physical activity recommendations by ~20% [5
]. Adults in southeastern U.S. had the highest prevalence of physical inactivity (28.0%) based on Behavioral Risk Factor Surveillance System (BRFSS) data [6
]. According to the World Health Organization, the prevalence of insufficient physical activity among adults is as high as 40–50% in countries in the subtropical or tropical zones such as Saudi Arabia, India, Brazil, etc. [7
], where high temperatures may be a barrier to physical activity.
Different types of interventions (e.g., one-to-one counselling, self-directed physical activity, supervised physical activity, etc.) have been conducted to promote physical activity among children, adolescents, and older adults [8
]. However, a review by Foster et al. (2014) suggested that the effect of interventions on self-reported physical activity was mixed with significant heterogeneity in reported effects [12
]. New procedures are needed to meet perceived convenience, accessibility, safety, and aesthetic requirements in a given climatic condition, especially in humid tropical conditions [5
As some studies have suggested, initiating and maintaining strenuous exercise programs is difficult [13
]. Zimmerman et al. (2009) suggested the use of anchors such as social norms, habits and a cultural frame to influence people’s preferences for action to promote physical activity [15
]. Nudging, which alters people’s behavior in a predictable way without forbidding any option, has been identified as an effective approach to promote physical activity [16
]. For example, Bellettiere et al. (2017) found that stair use increased when placing signs at the bottom of stairs to encourage people to go up [18
Evidence suggests that time spent outdoors is positively related to reduced sedentary time and moderate and strenuous exercise in adults [19
]. Harada et al. showed that time spent outdoors was significantly and positively associated with physical activity measured as daily steps among 192 older adults, and suggested the health benefits of time spent outdoors were primarily mediated by physical activity [24
]. Higher frequency of going outdoors was associated with less likely decline in the activity of daily living score among older adults [25
]. Beyer et al. suggests the association between time spent outdoors and increased physical activity could be an opportunity to promote physical activity among youth [19
]. However, the nudge approach related to time spent outdoors combined with technology support providing visible feedback (e.g., pedometer) for increased physical activity has largely been unexplored [17
]. Encouraging even a small amount of additional time spent outdoors, which is positively associated with increased physical activity and reduced sedentary time from previous studies, could increase physical activity.
We hypothesized an intervention of spending an additional 30 min outdoors daily beyond normal activity would provide physical activity benefits with minimal risk to increased heat exposure, particularly when people are free to choose the time of day. Summers in Alabama (AL), U.S., are characterized by subtropical temperature and humidity, where the average high temperature is ~33 °C (91 °F), average low temperature is ~22 °C (71 °F), humidity is ~75% and there are ~12 days with precipitation per month [26
]. Participants could freely choose time of the day and activity to spend an additional 30 min outdoors, and participants were instructed on methods to avoid heat stress and safely carry out the intervention. Effectiveness may be different across urban and rural settings; therefore, feasibility and compliance of this “nudge” intervention were estimated in both an urban and a rural setting in AL.
Participants’ characteristics are presented in Table 1
. We excluded one participant due to non-compliance with protocol (Consolidated Standards of Reporting Trials (CONSORT) flowchart in Supplemental File 1
). All participants were women and 173 out of 177 (98%) participants self-identified as Black or African American. Thirty-two participants from Birmingham were outdoor workers (i.e., Urban OutWor). Urban OutWor participants were significantly younger (p
-value 0.03), had a lower measured body fat (%) (p
-value 0.04) and a higher measured body water (%) (p
-value 0.02) compared to urban non-outdoor worker participants (i.e., Urban residents). Prevalence of diabetes was higher among Rural compared to Urban residents (35 out of 88 (40%) vs. 7 out of 57 (12%)). Rural participants on average had a higher measured body fat (%) (p
-value 0.04) and a lower measured body water (%) (p
-value 0.02) compared to Urban residents. We observed no significant differences in education, annual household income levels, BMI and obesity prevalence when comparing Rural vs. Urban residents, or Urban OutWor vs. Urban residents. When compared to the U.S. census data in these two locations, a higher percent of the participants self-identified as African American (95% vs. 71% in Birmingham, 100% vs. 71% in Wilcox County), had high school and above education (91% vs. 86% in Birmingham, 88% vs. 77% in Wilcox County), and had lower median annual household income (<USD 20,000 vs. USD 35,346 in Birmingham, <USD 20,000 vs. USD 27,237 in Wilcox County) [43
]. A total of 166 out of 177 (94%) participants had a valid PANES score. Participants in the urban location had a significantly higher PANES score compared to participants in the rural location (3.4 out of 7 (95%CI (3.0, 3.7)) among participants in the urban location vs. 1.6 out of 7 (95%CI (1.3, 1.9)) among participants in the rural location).
Participants spent an additional 30 min outdoors on 736 (83%) intervention person-days. A total of 104 (59%) participants spent an additional 30 min outdoors on every intervention day while only four (2%) participants never carried out the intervention (Figure 1
). There was a statistically significant difference in the compliance days between Rural residents and Urban residents (Chi-Square = 7.99, Degrees of Freedom = 3, p
-value = 0.046), but no significant difference between Urban residents and Urban OutWor (Chi-Square = 3.29, Degrees of Freedom = 3, p
-value = 0.35).
The frequency of self-reported difficulty in intervention compliance is shown in Figure 2
. Participants reported difficulty in intervention compliance on 316 (36%) person-days, and Urban residents reported more person-days with difficulty in intervention compliance compared to Rural residents (126 out of 285 (44%) person-days vs. 128 out of 440 (29%) person-days). We observed similar frequencies of reported difficulty between Urban residents and Urban OutWor (126 out of 285 (44%) person-days vs. 62 out of 160 (39%) person-days). The self-reported reasons for difficulty in intervention compliance are shown in Figure 3
. Rain, heat, and time conflicts were the leading reasons for intervention compliance difficulties (Figure 3
). We presented the factors associated with the probability of intervention compliance in Supplemental File 2
. The effect sizes of most fixed effects are small; participants who were physically inactive had a 15.70% (95%CI (8.94%, 22.46%)) reduced probability of intervention compliance.
The population average of individual mean steps on baseline and intervention days is shown in Figure 4
, where Rural residents and Urban residents walked more steps during intervention although the difference was statistically insignificant (Figure 4
). In the mixed models, participants on average walked 637 (95%CI (83, 1192)) more steps on intervention days (Table 2
). We did not find a significant interaction effect between intervention and groups (Supplemental File 3
). In separate groups, Rural residents had a significant increase of 1063 (95%CI (273, 1851)) mean daily steps during intervention days, after accounting for ambient conditions and other individual-level factors (Table 2
). Participants in urban locations had a smaller increase in steps on intervention days compared to Rural residents (Table 2
). Participants walked more steps on intervention weekends than intervention weekdays (Supplemental File 4
). Intent-to-Treat results and Per-Protocol results are similar; we found slightly lower estimated intervention effect in Per-Protocol, with an average 579 (95%CI (5, 1154)) additional steps on intervention days (Supplemental File 5
). Participants had fewer steps on intervention days in the minimal processed dataset compared to primary dataset, with the β estimate of intervention −271 (95%CI (−960, 418)) in minimal processed dataset (Supplemental File 6
Rural and Urban participants had similar average daily mean or max individually experienced HI on intervention days, but Urban OutWor had significantly lower daily mean or max individually experienced HI during the intervention, after accounting for WS HI (Figure 5
). When we included ambient conditions and individual-level factors in models, we found overall participants had a 0.59 °C (95%CI (0.30, 0.88)) lower daily mean and a 1.40 °C (95%CI (0.53, 2.27)) lower daily max individually experienced HI on intervention days (Table 3
and Table 4
). An interaction term between intervention and group was significant (Supplemental File 7
). In separate groups, Rural residents and Urban OutWor participants on average experienced a 0.49 °C (95%CI (0.09, 0.89)) and a 1.74 °C (95%CI (1.09, 2.38)) lower daily mean HI[individual] during intervention days, respectively (Table 3
). Urban OutWor experienced a 6.60 °C (95%CI (4.11, 9.09)) lower daily max HI[individual] during the intervention (Table 4
Overall, participants had lower daily mean or max HI[individual] on intervention days during weekends compared to intervention days on weekdays (Supplemental File 8
). ITT results and PP results were similar, with slightly smaller estimated effect sizes in PP (β estimate of intervention −0.59 (95%CI (−0.88, −0.30)) in ITT vs. −0.49 (95%CI (−0.79, −0.20)) in PP on daily mean of individually experienced HI, and −1.40 (95%CI (−2.27, −0.53)) in ITT vs. −0.99 (95%CI (−1.90, −0.08)) in PP on daily max of individually experienced HI) (Supplemental File 8
). Outlier removal minimally affected the intervention effect on daily mean HI difference (β estimate of intervention −0.59 (95%CI (−0.88, −0.30)) in ITT vs. −0.51 (95%CI (−0.83, −0.19)) in ITT with no outlier removal). Outlier removal affected the intervention effect on daily max HI difference more (β estimate of intervention −1.40 (95%CI (−2.27, −0.53)) in ITT vs. −0.58 (95%CI (−1.93, 0.76)) in ITT with no outlier removal) (Supplemental File 9
Body measurement change ratios (%) are shown in Table 5
. Overall, participants had a small decrease in weight, body fat and muscle mass and a small increase in body water. These change ratios were only statistically significant in participants who were obese. A −0.29% (95%CI (−0.45, −0.13)) weight change ratio is equivalent to a 0.52 lb. (95%CI (0.23, 0.81)) weight loss for a participant weighing 180 lbs. at baseline. There was no significant difference in body measurement change ratios in sensitivity analysis including participants with extreme measurement change ratios (Supplemental File 10
This study investigates whether spending an additional 30 min outdoors daily in summer is feasible in an urban versus rural environment, and if it changes daily steps and individually experienced HI of participants. Rain, heat, and time conflicts were leading factors hindering participants from spending an additional 30 min outdoors in both environments. This result is consistent with findings in previous studies [1
], suggesting heat is a barrier for physical activity in summer. Since it is hot and humid with frequent storms in the summer in the southeastern states of the U.S., heat and rain may be barriers to outdoor time and associated physical activity benefits. We found participants who self-identified as physically inactive had a 15.70% (95%CI (8.94%, 22.46%)) lower probability of intervention compliance. These results indicate participants starting with less physical exercise might perceive higher barriers to spending time outdoors, suggesting efforts to improve time spent outdoors among participants with less physical exercise may require initially reducing the amount of time (e.g., start with 15 min) or other methods of encouraging behavior change.
Participants increased daily steps by 637 (95%CI (83, 1192)) on intervention days. This relation was driven by increased daily steps in Rural residents, who walked a mean of 1063 (95%CI (273, 1851)) more steps (baseline daily 4346 steps, 24% increase) on intervention days. In contrast, Urban OutWor participants, with much higher baseline steps, only had a small increase in daily steps on intervention days. The results suggested that the benefits of the increased time spent outdoors may be more significant in physically less-active participants. The built environment (e.g., sidewalks, trails, recreational facilities) impacts physical activity [45
]. Birmingham is the second most walkable city in AL while Wilcox County is considered a car-dependent, less walkable location, based on the walk score metric [49
]. Among participants, Birmingham was rated as a more activity-friendly, walkable location with more access to recreational facilities compared to Wilcox County in the PANES results, although some neighborhood environment variables in the PANES may not be relevant for rural neighborhoods [30
]. These differences in the built environment could at least partially explain differences in neighborhood-level microclimates and might impact the intervention effects on promoting physical activity among participants. The generalizability of the results presented to other populations with similar or different demographics should be evaluated in future studies. Spending an additional 30 min outdoors daily is minimally limited by socioeconomic status (SES), although we acknowledged that conflicts of time/limited free time associated with lower SES from participants were reported (Figure 2
). We believe our results may be useful to provide an additional intervention method to promote physical activity among populations with similar SES in both urban and rural settings, especially in subtropical/tropical states in the U.S. Small but significant changes in body measurement change ratios were detected among participants who were obese after participation, suggesting the intervention benefits may be more significant among people with higher BMI.
While previous studies use weather station data to estimate the effect of ambient conditions on physical activity [3
], in the current study we have additionally measured microclimates experienced by participants within urban and rural neighborhoods and individual HI experienced by participants as they move through outdoor and indoor environments. This is important as previous studies have shown a wide variation in temperature and humidity experienced within cities, suburban, and rural environments [54
]. Overall, participants experienced lower daily mean or max HI on intervention days after accounting for ambient conditions, suggesting the additional 30 min outdoors did not result in increased heat exposure. Urban and Rural participants experienced a similar small change in daily mean or max HIs on intervention and baseline days, while outdoor workers had significantly reduced HI exposure during intervention days. Outdoor workers may have carried out the intervention in the cool hours of the day, thereby reducing their overall daily heat index exposure. Since participants were free to choose the time of day to spend the additional 30 min outdoors, we think most of the participants carried out the intervention either on early mornings or after sunset to avoid the hottest hours. Additionally, because the estimated prevalence of home central air-conditioning was not high for participants, outdoor environment may be cooler than homes when participants carried out the intervention, leading to reduced individually experienced HI. However, there was high missingness for the central air-conditioning response, so it is difficult to draw conclusions.
Two baseline days were weekdays while two out of five intervention days were weekends. To remove the weekend effect, we compared the daily mean or max individually experienced HI on baseline days vs. intervention on weekdays. However, this step considerably reduced the observation sample size. We observed that weekends augmented the negative association between the intervention and daily mean or maximum individually experienced HI in participants. Our results show that non-outdoor worker participants increased daily steps during the weekend but did not increase individually experienced heat indexes.
To address the concern that the thermometer on the shoe might pick up high temperatures due to artifacts (e.g., close to warm surfaces) when the actual environment was not hot, we removed upper outliers. The removal had minimal impact on the intervention effect on daily mean HI[individual]. Pedometer data imputation changed the intervention effect substantially.
In future studies, researchers may use pedometers with built-in daily reading features, or accelerometers to monitor physical activity more accurately. The benefits of additional time spent outdoors would likely include increased physical activity and may be more pronounced after longer term compliance, although this requires further study. Using advanced wearable technologies (e.g., FitBit, Apple Watch), albeit more expensive, to incorporate heart rate, time spent in different intensity activity, energy expenditure and total distance to measure physical activity more accurately would be an important next step to quantify the physical activity benefits. Participants could be further encouraged to engage in physical activity from these additional real-time feedback measures. Benefits beyond improved physical activity, such as improved mental health, an improved sense of well-being and blood pressure etc. suggested by previous studies could also be included [60