Activity Levels over Four Years in a Cohort of Urban-Dwelling Adolescent Females

Abstract

Purpose

Evidence suggests that female adolescents and those living in urban environments may have lower physical activity (PA) levels compared to their peers. Yet, few studies report PA for urban adolescent females and there is no consensus regarding potential causes for low PA in this subgroup. We examined PA levels, in a large, diverse cohort of 14–17 year old urban-dwelling females and assessed the impact of socio-economic, personal, and neighborhood/environmental factors on PA.

Methods

One week of time-stamped step count data was collected on 926 girls from the Pittsburgh Girls Study (PGS) at four annual visits. Valid recordings (worn at least 10 hours on 3+ days) were examined and compared to normalized step count values from a U.S. population-representative sample. Relationships between important covariates and average steps/day were examined with regression models.

Results

Adjusted mean (stdev) step counts/day at baseline were 5,614 (2,434) after controlling for important covariates with less than 6% of girls achieving at least 10,000 steps/day. PGS girls accrued ~45% of their steps during school hours. Age specific, median step counts/day for study participants were similar to the 25^th percentile of U.S. population normalized values and did not significantly change over follow-up. Non-Hispanic African American race/ethnicity was associated with higher average step counts/day; obesity and a recent childbirth were associated with lower average step counts/day.

Conclusions

Step counts in this cohort of urban adolescent girls were considerably lower than expected for U.S. adolescent females. Targeted efforts to improve PA levels in urban youth should consider the importance of school-based activity while increasing PA opportunities outside of school.

Introduction

Low levels of physical activity (PA) have been associated with poor quality of life, the development of chronic diseases, and other negative health outcomes in youth.(23, 24) Currently, the U.S. Centers for Disease Control (CDC) and the World Health Organization (WHO) recommend 60 minutes/day of PA for youth.(37, 41) Compared to the general youth population, lower percentages of disadvantaged and urban youth appear to be meeting the recommendations. (8, 15, 16, 19) The known health benefits of physical activity support prioritizing the identification and characterization of these and other subgroups at high risk for inactivity. In addition, the recent Surgeon General’s Call to Action, titled, “Step it up!”(38), highlighted the need to identify factors related to PA levels in these subgroups that may better inform targeted PA intervention efforts for those most in need of improvement.

Over the last two decades a consistent picture has emerged in which activity levels among youth decrease with increasing age and females are less active than males.(18, 22, 36) For example, adolescent females aged 12–19 from the National Health and Nutrition Examination Survey (NHANES), recorded less steps/day on average than similar aged boys (9,000 versus 11,000), and less when compared to both younger girls and boys (12,000 and 13,000 step/day, respectively; aged 6–11 years).(36) Aspects of the built environment that vary across metropolitan settings have also emerged as factors that may lead to differences in PA levels across subgroups.(8, 16). In relation to this, one study reported that urban youth had significantly lower average (stdev) step counts/day compared to suburban and rural youth (10,856 ± 3,706 versus 12,297 ± 3,616 and 11,934 ± 3,374, P<0.05); with the lowest mean step counts/day in urban females.(16) Furthermore, studies of environmental factors suggest that differences in opportunities to engage in PA in/outside of school and neighborhood safety (perceived and objectively recorded) may contribute to activity level variations across metropolitan settings.(10, 12, 33) Other factors highly correlated with metropolitan setting that may be related to activity levels include race/ethnicity and poverty.(8)

Few studies report PA data specifically for adolescent females living in an urban environment (8, 12, 16, 40) and even less do so using objective measures of PA (12, 16, 40) which may be more accurate and precise compared to subjective measures (e.g., questionnaires). The few published studies using objective measures have been cross-sectional studies of younger adolescent females (< 13 years of age) and none of these studies have examined urban cohorts with sufficient size or ethnic and socio-economic diversity to fully explore the roles of race/ethnicity, environmental factors, poverty factors, and personal/lifestyle factors on PA levels within this population.

This paper aims to describe PA levels, in a large, population-based, racially and socio-economically diverse cohort of 14–17 year old urban dwelling females followed over a 4 year period using time-stamped step counts. The activity level of this cohort will be compared to known step goals and to published normalized population values for average step counts/day derived from female adolescents participating in NHANES. Finally, the impact of socio-economic, personal/lifestyle, and neighborhood/environmental factors on step counts in this cohort will be examined.

Methods

Participants were recruited from the Pittsburgh Girls Study (PGS; begun in 1999). The PGS study design, methods, and primary results have been published.(20) Briefly, 2,451 girls comprising four age cohorts (5 to 8 years in wave 1) were recruited from the 89 zip code defined residential neighborhoods of Pittsburgh, Pennsylvania using random household sampling techniques. In the 30 lowest income neighborhoods, 100% of households were enumerated. Approximately 50% of households were randomly enumerated in the remaining neighborhoods. Overall, 83.7% of girls listed by the U.S. Census were identified. Of the 2992 eligible families identified, 2451 (82%) agreed to participate. Sampling weights were determined, based on the oversampling pattern, to maintain population representativeness. Prior to age 18 written informed consent was obtained from the caregiver and participants gave verbal assent. From age 18 onward the young women provided written informed consents. The study was approved by the University of Pittsburgh institutional review board.

The PGS pedometer ancillary study began in assessment wave 10 (2010) at which time the girls were 14–17 years old and continued to wave 13 (2013). A total of 1045 girls were recruited sequentially from PGS interview respondents in 2010 to determine objectively assessed PA levels, reported by pedometers, in this cohort. The mean retention rate of the original PGS sample was over 90% in wave 10.

Data Collection

Trained interviewers conducted separate, annual in-home interviews for girls and caregivers at which socio-economic and neighborhood perception variables were collected from caregivers. Ethnicity/race was defined as non-Hispanic (NH) African American, NH Caucasian, and All Other due to the small numbers of caregivers reporting other ethnic/racial categories. Family poverty was defined as: receiving public assistance [y/n]. In addition, single parent household [y/n], living in a neighborhood with primarily single family households [y/n], and parent’s education less than high school [y/n] were also examined. Socio-economic status (SES), as a composite variable, was calculated from: parental education (less than high school), public assistance, and single parent household.

Neighborhood perception variables included measures of neighborhood safety, problems, convenience, appreciation, and neighbor helpfulness. Each measure was based on responses to established questions.(26) Additionally, neighborhood was characterized according to trained interviewers’ perception about whether the participant’s block was predominantly single-family housing. U.S. Census statistics (2010) for households below poverty (%) and households with less than a high school education (%) were based on 2010 postal zip code defined neighborhoods.

Girls provided data on personal/lifestyle factors including current smoking [y/n], recent childbirth status (gave birth in the past year), and body mass index (BMI). BMI was calculated from interviewer measured height and weight. Weight categories were defined using accepted age/gender specific BMI cutpoints: normal and underweight (BMI <85^th %tile), overweight (BMI ≥85^th%tile, but < 95^th %tile), mild/moderate obesity (BMI ≥95^th %tile but less than 1.2×95th %tile), and severe obesity (BMI ≥1.2×95th %tile).(9) Families were compensated for their participation.

Physical activity measurement

PA data was collected using a reliable and valid pedometer, the Omron HJ-720ITC (Kyoto, Japan). Previously validated algorithms used in the HJ-720ITC convert recorded displacement along the two measurement axes into step counts.(30) Data are output as time-stamped step counts in one hour increments.

Monitors were distributed at the participants’ interviews during four consecutive annual visits, (waves 10–13). Participants wore the monitors on their waist for the 7 days following each interview visit. Identification of each recorded hour as worn/not worn was provided with the standard monitor output utilizing proprietary algorithms. Data were screened for mailback time and monitor wear time. Girls were asked to rewear the monitor if they had <3 days of identifiable monitor wear. Valid days of monitoring were defined using the criteria of ≥10 hours of wear time and a lower threshold of at least 150 steps.(4) Girls with <3 valid days of recording were not included in the final analyses. (1)

Statistical Analyses

Population weights were used in all analyses. Chi-squared and Kruskal-Wallace tests were used to determine differences between PGS girls with/without valid pedometer data. The PGS population weighted values for average total steps/day by age at different percentiles of the distribution were compared to published normalized values to provide a context for step counts recorded in PGS. The normalized reference values were generated using step count data from a U.S. population representative sample of females aged 14–20 (NHANES 2005–2006).(3)

Regression models were used to determine differences in step counts both between and within groups at baseline and longitudinally. Several a priori factors; monitor wear time, weekend day included (Saturday or Sunday) [y/n], seasonal variation indicated as recorded in winter (December through March) [y/n], and age cohort were included in all models. In the case of missing values for neighborhood perception variables, and socio-economic variables, the previous years’ values were carried over. A sensitivity analysis in which socio-economic and neighborhood variables were not carried over did not change the findings.

Adjusted generalized linear regression models were used at baseline (PGS wave 10) and mixed models were used for longitudinal data analyses (PGS waves 10–13) to calculate values for least-square (adjusted) means and standard errors (stderr) for average total steps/day. Significant differences across and between subgroups were established using analysis of covariance and Tukey’s test, respectively. Linear mixed models were used to determine important predictors of step counts over the entire follow-up. Univariate models, including the a priori covariates, were run to determine the relationship between each demographic, lifestyle, neighborhood, and socio-economic variable and average steps/day. Additional multivariable analyses included all potential predictors. A step down approach was used to determine the most appropriate final model. Statistical analyses were conducted in SAS v 9.3 (SAS Institute, Cary, NC).

Results

Baseline (PGS study wave 10) step counts

A subsample of 1,045 of the original 2,451 PGS girls were invited to participate in the pedometer study. There were no significant differences in key demographic and lifestyle factors between girls recruited/not recruited into the ancillary study (based on p<0.05; data not shown). At baseline of the ancillary study (PGS wave 10) 926 (88.6%) of 1045 enrolled girls had complete pedometer data (≥3 days with ≥10 hours of recording time and ≥150 steps [Table 1]). Compared to girls without complete data, ancillary study participants with complete data were more likely to live in neighborhoods with more household incomes below poverty level (15.6 versus 20.7%, p=0.02).

Table 1.

Demographic characteristic of Pittsburgh Girls Study pedometer study participants at baseline (wave 10 assessment; n=1045)

Variable	Without complete PED data N=119	With complete PED data N=926	P-value
Study cohort (%)			0.0041
5 (~ age 14)	14.4	23.8
6 (~ age 15)	24.0	26.2
7 (~ age 16)	26.0	28.4
8 (~ age 17)	35.6	21.6
Race (%)			0.77
NH African American	43.9	46.2
NH Caucasian	48.6	47.2
Mixed/Other	7.5	6.6
BMI percentile (median,IQR)	81.0 (54.9, 94.4)	78.6 (51.9,93.6)	0.91
Previous birth (%)	1.7	2.3	0.23
Smoking, current (%)	16.5	14.6	0.60
Single parent house (%)	50.7	44.1	0.18
Parent less than high school education (%)	35.7	44.6	0.07
Public assistance, current (%)	34.4	34.5	0.98
SES, composite variable (%)			0.13
Low	48.8	41.0
Middle	16.9	24.5
High	34.4	34.5
Neighborhood (%)
Below Poverty (median,IQR)2	15.6 (11.5, 25.9)	20.7 (11.4, 30.9)	0.02
Less than high school education (median,IQR)2	13.4 (4.0, 18.8)	14.3 (13.0, 11.9)	0.99

NOTE: PED= Pedometer; BMI= Body mass index; IQR= Inter-quartile range; SES= Socio-economic status; NH= Non-Hispanic

¹Based on chi-squared p-value for differences, p-trend was not significant (p=0.49);

²Neighborhood statistics based on 2010 census

Of the 926 girls with complete pedometer data, 38.4% were NH Caucasian and 55.6% were NH African American. They were comparable to nationally representative samples of similarly aged girls for reporting a pregnancy (2.3% versus 2.4%)(13), and reporting currently smoking (14.6% versus 15.7%).(17, 19) The percentage from families receiving public assistance was similar to 2011 statistics for adolescents (11–15 years) in the U.S. population (34.5% versus 32%),(6) while the percentage defined as overweight (body mass index ≥85%tile) was somewhat higher compared to NHANES 2009–2010 data for 12–19 year olds (43.9% versus 32.6%).(27)

Average [standard deviation; stdev] monitor wear time was 15.8 [1.9] hours/day. Baseline (age 14–17; PGS study year 10) step counts were relatively low (Table 2). Average unadjusted mean [stdev] steps/day were 5,614 [2,434]. Average median [inter-quartile range; IQR] steps/day were 5,368 [3,866, 7,069]. Only 5.3% of girls had at least 10,000 steps/day which is an estimate of steps needed to meet the 60 minute CDC activity goal for adolescent females.(34) Based on the commonly used 5000 steps/day cut-point, 43.7% of the girls were considered inactive or sedentary at baseline.(35)

Table 2.

Mean (stdev) daily activity at baseline (wave 10) for PGS girls with valid pedometer data (n=926)

	Step counts	Maximum hourly step counts1	Percentage of wear hours with < 100 steps	Percentage of inactive girls2
All (n=926)	5,614 (2,434)	1,429 (639)	41.9 (15.9)	43.7
Age in years
14 (n= 217)	5,520 (2,285)	1,372 (583)	40.0 (15.2)	44.7
15 (n= 240)	5,763 (2,474)	1,505 (697)	41.6 (16.1)	40.5
16 (n= 262)	5,843 (2,489)	1,497 (649)	42.0 (15.4)	40.1
17 (n= 207)	5,252 (2,434)	1,311 (594)	44.3 (16.9)	51.4
Race/ethnicity
NH African American (n= 515)	6,087 (2,373)	1,522 (563)	43.0 (15.1)	37.0
NH Caucasian (n= 356)	5,244 (2,426)	1,357 (721)	40.3 (16.8)	48.6
All Other (n= 55)	5,024 (2,423)	1,296 (659)	45.0 (16.3)	56.4

Note: Stdev= Standard deviation; NH= Non-Hispanic;

¹Daily average for highest step counts recorded during an hour of the day;

²Population level value based on average step counts/day; inactive is defined as <5000 steps/day

NH African American girls recorded slightly more mean [stdev] steps/day when compared to NH Caucasian girls (6,087 [2,373] versus 5,244 [2,426]; Table 2). Based on 10,000 steps/day, 6.1% of NH African American girls would have been meeting the activity goal compared to 2.8% of NH Caucasian girls and 3.2% of girls reporting other racial/ethnic groups. The results of a comparison of adjusted least-squares mean steps/day (models included cohort, monitor wear time, season of wear, and if a weekend day was recorded) between NH African American and NH Caucasian girls across SES groups suggested that step counts were higher for NH African American girls in all three SES groups reaching significance (Tukey’s test p< 0.05) in the lower and moderate SES groups (see Figure, Supplemental Digital Content 1, Adjusted means and standard errors for total steps per day at wave 10 visit for NH African Americans and NH Caucasians across Socio-economic status groups).

Step counts/day were significantly lower (p<0.0001) for girls who wore the monitors in winter (n=474; median [IQR]:5,585 [3,817, 7,404]), December-March) compared to girls who wore the monitors during other months (n=452; median [IQR]:6,341 [4,774, 8,247]). Although activity levels were significantly higher on week days compared to weekend days (p<0.05; data not shown), girls with a recorded weekend day did not have significantly more or less steps than girls without a recorded weekend day (n=728; median [IQR]:5,348 [3,866, 6,976]) versus (n=198; median [IQR]:5,566 [3,865, 7,506]; p=0.32).

Diurnal activity patterns suggested that average step counts were significantly different across hours of the day for week days and weekend days (both p<0.0001; see Figure Supplemental Digital Content 2, Median and inter-quartile range (IQR) for average pedometer steps by time of the day during the school year). On week days, median steps/hour quickly peaked near the start of the school day (8–9 AM; median [IQR]: 279 [100, 530] steps) and again in the late afternoon (4–5 PM median [IQR]: 463 [230, 754] steps). However, more steps overall were accrued from 8 am to 3 pm (corresponding to school hours; median [IQR]: 2,635 [1,724, 3,589]) when compared to other periods of the day (12AM–6AM, 6AM–8AM, 3–7PM, and 7PM–12AM).

PGS step count values by age: Comparison to NHANES

The number of records for the PGS sample (recorded for all waves; 10–13) was highest at age 17 (n=754); as all four birth cohorts wore the pedometers at this age. Median [IQR] values for steps/day among PGS girls were quite stable over time (Figure 1). At 14, 17, and 20 years of age, median (IQR) step counts/day were 5,177 [2,209, 10,243], 5,116 [1,776, 10,458], and 4,599[1,700, 10,584], respectively.

Figure 1.

PGS cohort steps per day percentiles by age (collected during waves 10–13) and normalized steps per day percentiles from NHANES 2005–2006 data

Note: NHANES data for ages 14–19 adapted from Barriera et al. 2015(2); NHANES 2005–2006 data for age 20 provided via personal correspondence with Dr. John M. Schuna.

Overall, total steps/day in PGS girls were lower than normalized step values from NHANES for similar aged girls (figure 1). PGS 14 year old girls at the 5^th and 95^th percentiles recorded an average of 444 and 1,098 steps/day less than the normalized values for 14 year old girls at the 5^th and 95^th percentiles, respectively. For each age group the results suggest that the 50^th percentile step count value in PGS girls were similar to the 25^th percentile for normalized population step count values from NHANES. However, PGS step count values at the 5^th and 95^th percentiles tracked relatively closer to the normalized values at the U.S. population’s 5^th and 95^th percentile.

Longitudinal PGS step count values: Associations with personal and environmental factors

Longitudinal data from PGS was further examined with mixed models to account for repeated measures. Only girls with valid baseline data (n=926) were included in the longitudinal analyses. A large proportion of these girls also provided data with ≥3 valid days, in subsequent waves: 832 (wave 11), 790 (wave 12) and 758 (wave 13) individuals, respectively. All models were adjusted for the a priori variables of monitor wear time, inclusion of a weekend day [y/n], season of pedometer wear (winter/non-winter), and age cohort. The adjusted mean [standard error; stderr] estimates for average steps/day were 5,623.2 [90.7], 5,222.4 [102.7], 5,585.7 [105.0], and 5,315.7[108.8], at PGS study wave assessments 10–13, respectively.

In the individual univariate models, which also contained the four a priori variables, significant predictors of average steps/day (p<0.05) were race/ethnicity, weight category, recent birth, and all four socio-economic variables (single family home, neighborhoods with predominately single family homes, parental education, and receiving public assistance). Six of seven neighborhood perception variables were significantly related to step counts (p<0.05, data not shown); only neighborhood appreciation was not. Indicators of higher poverty and poorer neighborhood perception were generally associated with higher, not lower, step counts/day (data not shown). The relationship between weight categories and average steps/day was not monotonic; girls in the severely obese category recorded the lowest step count/day followed by girls with normal/underweight BMI (data not shown).

Significant predictors of average steps/day in the final multivariable model (Table 3), were season of monitor use (winter versus non-winter), average monitor wear time/day, birth cohort, race/ethnicity, recent child birth, body weight category, and whether the girl’s home neighborhood was predominantly single-family housing [y/n]. Although step counts/day did not differ over time in this model (p=0.62) there was a difference in step counts across birth cohorts, with steps decreasing with age. The strongest predictor of steps was “having given birth in the past year” which was associated with an average [stderr] estimated reduction of 1,476.3 [294.0] step counts/day (p<0.0001). Girls identified as being severely obese had a significant average [stderr] estimated reduction of 707.3 [190.0] step counts/day when compared to overweight girls (reference category; p-dif <0.001). NH African American ethnicity/race and living in a neighborhood with less single-family housing were associated with higher estimated step counts.

Table 3.

Reduced multivariable model predicting average steps/day in PGS girls from pedometer measurement baseline (wave 10) to wave 13.

Variable	Estimate (Stderr)1	P-value
Intercept	−631.0 (739.7)	0.39
Worn in winter (y)	−641.1 (102.9)	<0.0001
Average Wear time (hours/day)	303.08 (25.0)	<0.0001
Recording included weekend day (y)	134.2 (125.9)	0.29
Time (~1 year increment)	−22.5 (45.5)	0.62
Birth cohort (age at wave 10)		0.01
5 (~ age 14)	Reference	----------
6 (~ age 15)	37.3 (137.7)	0.79
7 (~ age 16)	−87.7 (136.0)	0.52
8 (~ age 17)	−431.4 (153.8)	<0.01
Race		<0.0001
NH Caucasian	Reference	----------
NH African American	662.0 (111.4)	<0.0001
All Other	343.4 (210.0)	0.10
BMI weight category		0.001
Low to normal (<85th %tile)	−396.5	<0.01
Overweight (85th – 94th %tile)	Reference	----------
Mild/moderate obesity (95th %tile to < 1.2×95th %tile)	−171.1 (195.5)	0.38
Severe obesity (≥1.2×95th %tile)	−707.3 (190.0)	<0.001
Recent childbirth; past year (y)	−1476.3 (294.0)	<0.0001
Multi-family house (y)	337.7 (120.0)	<0.01

Note:

¹For categorical variables estimate represents step count difference compared to reference group. For wear time the estimate is the difference for each additional hour of the day and time is for each ~1 year period between recordings.

Discussion

This study is the first we are aware of to report objective PA data for urban adolescent females over a four year period. The results of this study are in-line with other studies of younger adolescents suggesting that urban females may be less active than other youth their age.(8, 16) For each year of pedometer recording, only about 5% of PGS girls recorded at least 10,000 steps/day; the suggested minimum number of steps needed to meet the 60 minute/day activity goal.(34) Overall, daily step counts for PGS girls were lower than expected values, based on a nationally representative sample of similar aged girls. Additionally, changes in step counts over the follow-up period were small and not significant in the PGS cohort.

These findings highlight the importance of school-based activity for these urban youths. Based on time-stamped pedometer data school-based activity was important in this cohort, accounting for nearly 45% of recorded steps/day at baseline. The number of steps recorded during school hours was similar to other cohorts. (32, 39) However, school-based activity was a larger contributor to overall activity levels in this cohort, as the PGS girls reported lower step counts after school and on weekends compared to these other cohorts. The dominance of school-based activity in this sample may also suggest a lack of alternative physical activity options. Young urban females may benefit from the development of new opportunities for activity during the after-school period and on weekends.

PGS girls identifying as NH African American had slightly higher average step counts/day than girls identifying with other racial/ethnic groups. Previous studies have reported inconsistent findings for the differences in PA levels across racial/ethnic groups (7, 15, 21), with several recent studies supporting this study’s findings of higher PA levels in African American youth compared to similar aged Caucasian youth.(7, 11) It should be noted, that although significant, the differences in step counts across racial/ethnic groups in PGS appear smaller than differences between the PGS cohort, as a whole, and the normalized values for the U.S. population representative sample suggesting the relative importance of the urban environment on overall activity levels, regardless of race/ethnicity.

Reasons cited for why youth living in urban settings may have lower PA levels when compared to youth living in other metropolitan settings include less safe neighborhood environments and increased poverty.(12, 31) However, some urban youth studies have reported increased PA levels with increased poverty.(7, 29) One possible reason for this may be increases in transportation activity observed among those living below the poverty line.(15, 29) The association between higher poverty and poorer parental perception of neighborhood with higher PA levels in this cohort (although attenuated in the full model) may support the notion that disadvantaged youth may, of necessity, walk more for transportation.

Longitudinal changes in step counts (over the 4 years) were not significant in the PGS sample. While the cross-sectional NHANES data shows little difference in activity for adolescent girls across a similar age range, other longitudinal studies using objective PA measures have shown significant declines in activity levels during adolescence.(5, 14, 28) Most of these longitudinal studies report MVPA which cannot be separated out from total pedometer steps. For this reason, it is not possible to know whether MVPA levels could have similarly declined in the PGS girls across adolescence, despite the small overall changes in step counts.

A strength of this study compared to many previous studies was the use of an objective measure of PA. Objective measures, including pedometers can provide more accurate assessments of PA than questionnaires. However, interpretations of step count comparisons between monitors, such as Omron monitors (PGS) and ActiGraph monitors (NHANES), should consider the possibility that differences in reported steps could be due to differences in monitor design and processing. Although some bias is likely, both the Omron and ActiGraph monitors have validated well against manual counting, recording 97.5% and 98.5% of steps, respectively, and have shown good agreement with each other.(25) This would suggest a high degree of confidence in reported values from both monitors as well as cross-monitor comparisons.

The size and diversity of this cohort was advantageous and made it possible to examine the role of personal and environmental factors that may affect activity levels within the urban environment. Our study was also not confounded by the relationships between metropolitan setting and socio-economic status and metropolitan setting and race/ethnicity. However, without suburban and rural youth it was not possible to gain a complete understanding of factors that may be contributing to low PA levels among youth living in urban settings, but not other metropolitan settings. Therefore, future efforts should seek to examine a large cohort, similar to the PGS, that also includes youth living in other metropolitan settings.

Conclusions

The overall low levels of PA in the PGS cohort suggest that adolescent girls living in urban environments should be considered a priority group for targeted PA improvement. The daily patterns of activity in the PGS girls suggests that future intervention efforts in similar urban adolescents should consider the potential importance of school-based activity. Increasing opportunities for activity outside of school, particularly on weekends should also be considered a priority.

Supplementary Material

Supplemental Data File _.doc_ .tif_ pdf_ etc.__1

Supplemental Digital Content 1.pdf—Adjusted means and standard errors for total steps per day at wave 10 visit for NH African Americans and NH Caucasians across Socio-economic status groups

Supplemental Data File _.doc_ .tif_ pdf_ etc.__2

Supplemental Digital Content 2.pdf—Median and inter-quartile range (IQR) for average pedometer steps by time of the day during the school year