Interrelationships of physical activity in different domains: Evidence from the Geographic Research on Wellbeing (GROW) study

Abstract

Research has rarely distinguished between non-work (NW) and work (W) active transport (AT) or investigated relationships to other domains of physical activity ([PA], like leisure time [LTPA] or work [WPA]). We investigated correlates of AT by employment status, accounting for LTPA and WPA, in a population-based sample of California mothers (N=2906) in the Geographic Research on Wellbeing (GROW) study (2012–2013). AT was measured by the National Household Travel Survey. LTPA was measured using the Stanford Leisure-Time Activity Categorical Item. WPA was measured with the Stanford Brief Activity Survey. Most employed mothers (53%) worked in sedentary jobs, and few (<10%) used NWAT or WAT. Over 20% of unemployed mothers used NWAT, although LTPA levels were similar to employed mothers. Multiple regression models found employed and unemployed with low education and income, and unemployed African American or Latina immigrant mothers had higher odds of using NWAT. Younger employed and unemployed mothers, and unemployed who had ≥4 children or had “light” LTPA had lower odds of using NWAT. Multiple regression models demonstrated that low education or income employed mothers, African American mothers, those who worked part time, and those with relatively low LTPA had higher odds of using WAT, while younger women had lower odds of using WAT, compared with reference groups (ps<0.05). WPA was associated with WAT in unadjusted models, but not in adjusted models. Different AT patterns were seen for employed vs unemployed women, but women who used AT did so for most trips. LTPA was associated with NWAT among unemployed mothers and with WAT among employed mothers. Most women were underactive across all domains, suggesting no compensatory effect of PA done in one domain reducing PA done in another domain, with few meeting minimal guidelines. Policy and practice strategies should support infrastructure to encourage a variety of domains of PA.

1. Introduction

Physical activity is causally related to multiple physical and psychological health benefits (Centers for Disease Control and Prevention (CDC), 2015). Despite these benefits, fewer than one in four adults meet even minimal, recommended guidelines, with only about one in five women meeting guidelines (Blackwell et al., 2014; CDC, 2014; CDC, 2015). The situation is even worse among women of color: compared to about 20% of white women meeting guidelines, about half as many African American (11%) and Hispanic or Latina (12%) women meet guidelines (Blackwell et al., 2014). Despite long time recognition of these disparities, there remain deficits in whether these statistics solely represent women who have sufficient leisure time and training to participate in leisure time physical activity, potentially missing women who may achieve sufficient physical activity via transportation or occupational demands.

As measurement of physical activity has improved, recent studies have begun to investigate the interrelationships among types of physical activity, with particular interest on how domains of physical activity vary by socioeconomic position (Beenackers et al., 2012). Traditional measures of physical activity have focused on physical activities done during leisure time (Amireault and Godin, 2015; Craig et al., 2003; Godin and Shephard, 1985), which may only capture a fraction of the physical activity that women do regularly (Baecke et al., 1982). Measurement of physical activity done for transportation (e.g., walking, cycling) or done during the work day (e.g., lifting, carrying, walking, cleaning) may capture a broader range of physical activities that women may do during a typical day. Although the proportion of people reporting active transportation (AT) is relatively low (3.4%) in the population at large (CDC, 2014), it is reasonable to expect that prevalence might be higher among people with more modest resources who do not have an automobile at their exclusive disposal. For example, among the approximately 3.5 million participants who completed the nationwide annual American Community Survey administered by the Census Bureau between 2008–2012, those who travel most by bicycling or walking were in the lowest household income brackets with participation declining as income increased, with the exception of those at the highest end of the income bracket (≥$200,000), who had higher participation; similar patterns were observed by education (McKenzie, 2014). In addition to higher use of AT, people with lower SES tend to hold positions with higher levels of work-related physical activity (Beenackers et al., 2012; Hadgraft et al., 2015; Vandelanotte et al., 2013; Vandelanotte et al., 2015).

Other investigations have suggested that people from lower SES who have high levels of physical activity at work might compensate with less physical activity participation in leisure time (Chau et al., 2012; Seiluri et al., 2011), but still achieve overall high levels of physical activity that would not be captured by measures of leisure time physical activity (LTPA) alone (Beenackers et al., 2012; Kirk and Rhodes, 2011; Vandelanotte et al., 2015). For example, non-Hispanic white women with higher SES tend to report more LTPA than other women with lower SES, but they also tend to be employed in occupations that require less physical activity (Ekenga et al., 2015). However, other studies have reported that this was not the case; rather, women with low levels of work physical activity also had lower levels of LTPA (Ekenga et al., 2015; JaKa et al., 2015), or that no relationship was found (Marquez et al., 2010).

Relationships among participation in different domains of physical activity are not clearly understood, and it is unknown whether people who do more physical activity at work or during leisure time are more or less likely to use AT. Further complicating understanding is the reality that women and ethnic minorities are overrepresented in lower SES strata compared with men and non-Hispanic/Latino whites (Lee and Cubbin, 2009; U.S. Census Bureau, 2014). These data underscore the need for investigations that rigorously control for SES when determining physical activity patterns among women and ethnic minorities. People from most ethnic minority groups have higher rates of AT than whites (McKenzie, 2014). Some research has also reported that Mexican-American women and women who were recent immigrants to the US were more likely to meet guidelines via transportation and work-related physical activity (Gay and Buchner, 2014; Martinez et al., 2011).

The gap in understanding the relationships among different types of physical activity may hamper efforts to help women meet physical activity guidelines. There are few studies that have systematically investigated these relationships in samples large and diverse enough to account for a range of ethnicities and socioeconomic position, including employment status. The purpose of this manuscript is to investigate physical activity done during leisure-time, work and transport in a representative sample of California mothers, and determine correlates of non-work and work AT. Findings from this study may help guide future intervention and policy strategies to increase physical activity across multiple domains.

2. Methods

2.1. Data source

The Geographic Research on Wellbeing (GROW) study was a follow-up survey of participants in the Maternal and Infant Health Assessment (MIHA) survey, a collaborative project of the California Department of Public Health’s Maternal, Child, and Adolescent Health Branch and researchers at the University of California, San Francisco. MIHA is an annual, statewide-representative mail/phone survey of mothers delivering live infants in California during February through May, and administered in English and Spanish (California Department of Public Health, 2016; Cubbin et al., 2002; Galbraith et al., 2003; Heck et al., 2003). The sampling frame is birth certificates, MIHA response rates typically exceed 70%, and the data are weighted to be representative of all eligible births statewide.

To be eligible for GROW, MIHA respondents from 2003–2007 had to live in one of six largely urbanized counties at the time of the MIHA survey, which represented over half of all MIHA respondents. They also had to have given permission to be re-contacted for a potential study in the future; which was nearly universal. During 2012–2013, GROW respondents answered about 80 questions relating to their sociodemographic, health, and neighborhood characteristics and received a $20 gift card as an incentive (with the option to also participate in a raffle). Nearly 75% of women who were located responded to the GROW survey (N=3016), and over 90% of them still lived in one of the six counties. Over half completed the survey by phone and over one quarter completed it in Spanish. Weights were created to account for nonresponse and produce data that were representative of the six GROW counties. Further details about the GROW study have been published (Cubbin, 2015). The analytic dataset excluded women whose race/ethnicity was reported as American Indian/Alaska Native, missing, or “other” because of small sample size, and those who were missing data on employment status (sample size for analysis: N=2906). The GROW protocol was approved by the Institutional Review Boards at the University of Texas at Austin, the University of California, San Francisco, and the California Department of Public Health; all participants gave informed consent.

2.2. Measures

Physical activity was self-reported for transportation, leisure time, and work domains. Active transport (AT) was measured by items from the National Household Travel Survey (U.S. Department of Transportation, 2009). Non-work AT was measured by how the respondent got to most places she went during the past seven days, not counting to/from work (drove or got rides vs. walked, walked and took public transit, or rode a bike). Work AT (WAT) was categorized similarly, except that it referred to the past two weeks among employed women at a paid job. Leisure time physical activity (LTPA) was assessed using the Stanford Leisure-Time Activity Categorical Item which has demonstrated reliability, validity and sensitivity to change over time (Kiernan et al., 2013). Participants selected one response which best described the type, intensity, frequency and duration of physical activity usually done during non-work time in the past 30 days to produce six classifications of LTPA ranging from inactive to vigorously active five or more times per week. Classifications and responses are presented in Fig. 1. Work physical activity (WPA) was measured with modified items from the Stanford Brief Activity Survey that asked respondents to select one option that best described the kinds of physical activity that she usually did at her job. Options included sitting or standing (such as at a desk job), walking or using her hands and arms (such as a seamstress, shop assistant, hairdresser, childcare helper), or definite physical effort including lifting or handling heavy objects, or doing hard physical labor (such as a cleaner, hospital nurse, gardener, postal delivery worker). Spearman correlations between all domains of physical activity were low (i.e., less than 0.15) except for the correlation between non-work AT and work AT among employed women (0.62) suggesting that employed women who use active transportation do so for most trips.

Fig. 1.

Summary of instructions and responses from the Stanford Leisure-Time Activity Categorical Item measure used to describe physical activity not done during work in the Geographic Research on Wellbeing (GROW) study (Kiernan et al., 2013).

Sociodemographic variables included age, race/ethnicity, marital status, educational attainment, income (annual family income, in increments of the federal poverty level: 0–100%, 101–200%, 201–400%, 401+%, missing) and, for employed women, whether they worked part time or full time. Missing income was imputed via hot deck methods using the following variables: Age, race/ethnicity, education, marital status, neighborhood poverty, and employment status.

2.3. Analyses

We first examined the means and distributions of all variables by employment status. Next, we examined odds ratios of non-work AT, with a focus on whether LTPA was associated with non-work AT, in models stratified by employment status. Last, we examined odds ratios of WAT among employed women only, with a focus on whether LTPA and/or WPA were associated with WAT. (Non-work AT was initially considered as an independent variable, but we chose to keep it out of the model based on the high Spearman correlation between non-work AT and WAT (0.62) and a very high unadjusted OR in a model regressing non-work AT on WAT (OR 24.27), indicating that working women practicing non-work AT were also at very high odds of practicing WAT, which might be expected.) Crude and adjusted models were constructed as indicated in Tables 3 and 4, and all analyses were conducted using SAS version 9.4 (Cary, NC), incorporating weights and the complex sample design.

Table 3.

Odds ratios of non-Work active transport, Geographic Research on Wellbeing (GROW) study (n = 2,906).

	Employed (n = 1,628)		Unemployed (n = 1,278)
	Unadjusted	Adjusteda	Unadjusted	Adjusteda
Age (years)
20–29	0.99 (0.54–1.82)	0.48 (0.24–0.98)	1.03 (0.67–1.59)	0.40 (0.24–0.68)
30–39	1.22 (0.77–1.92)	0.83 (0.49–1.38)	1.12 (0.79–1.58)	0.69 (0.46–1.05)
40 +	1.00	1.00	1.00	1.00
Race/ethnicity
African American	2.12 (0.98–4.55)	1.19 (0.53–2.65)	9.16 (4.67–17.97)	5.23 (2.30–11.87)
Asian or Pacific Islander	0.80 (0.23–2.71)	0.42 (0.12–1.46)	2.16 (0.81–5.78)	2.21 (0.76–6.39)
Latina, US born	1.29 (0.59–2.81)	0.72 (0.33–1.56)	2.40 (1.10–5.24)	1.83 (0.76–4.40)
Latina, immigrant	638 (3.76–10.82)	1.42 (0.72–2.81)	12.18 (6.83–21.72)	5.00 (2.50–10.01)
White, non-Hispanic	1.00	1.00	1.00	1.00
Marital status
Previously or never married	2.80 (1.77–4.42)	1.31 (0.78–2.19)	1.31 (0.88–1.95)	0.99 (0.62–1.59)
Married or living together	1.00	1.00	1.00	1.00
Number of children in household
0–1	1.00	1.00	1.00	1.00
2–3	0.73 (0.37–1.42)	0.70 (0.36–1.35)	1.07 (0.61–1.88)	0.60 (0.29–1.24)
4+	1.27 (0.60–2.70)	0.68 (0.31–1.46)	1.17 (0.64–2.12)	0.40 (0.18–0.86)
Education
Less than high school	12.76 (7.11–22.90)	2.57 (1.18–5.58)	9.65 (5.30–17.58)	2.86 (1.06–7.72)
High school graduate/GED	4.67 (2.50–8.71)	1.30 (0.61–2.77)	5.81 (3.14–10.72)	2.09 (0.78–5.64)
Some college	1.57 (0.78–3.18)	0.80 (0.36–1.78)	2.21 (1.13–4.32)	1.49 (0.64–3.48)
College graduate	1.00	1.00	1.00	1.00
Income
< = 100% federal poverty level	12.35 (6.88–22.16)	5.28 (2.18–12.78)	13.65 (6.71–27.76)	4.92 (1.53–15.87)
101–200% federal poverty level	4.41 (2.30–8.44)	2.53 (1.01–6.32)	5.53 (2.56–11.91)	2.21 (0.71–6.90)
201–400% federal poverty level	1.02 (0.43–2.41)	0.94 (0.35–2.49)	1.20 (0.45–3.18)	0.86 (0.26–2.83)
> 400% federal poverty level	1.00	1.00	1.00	1.00
LTPA
Inactive	2.43 (0.74–8.01)	1.33 (0.36–4.93)	1.00 (0.34–2.94)	0.26 (0.08–0.88)
Light activity, 1–2× per week	2.83 (1.14–7.06)	1.42 (0.56–3.58)	1.52 (0.69–3.34)	0.65 (0.26–1.64)
Moderate activity, 3× per week	2.80 (1.11–7.03)	1.57 (0.63–3.94)	1.54 (0.69–3.44)	0.70 (0.27–1.84)
Moderate activity, 5+ times per week	0.83 (0.22–3.08)	0.61 (0.15–2.44)	1.92 (0.81–4.55)	1.71 (0.63–4.66)
Vigorous activity, 3× per week	0.85 (0.27–2.72)	0.92 (0.28–3.06)	0.57 (0.18–1.84)	0.56 (0.14–2.26)
Vigorous activity, 5+ times per week	1.00	1.00	1.00	1.00

Note. Bold face type indicates a significant relationship between variables.

^aAdjusted for all variables

Table 4.

Odds ratios of work-related active transport, Geographic Research on Wellbeing (GROW) study (n = 1,628).

	Employed
	Unadjusted	LTPAa	Work physical activityb	Fullc
Age (years)
20–29	0.84 (0.46–1.51)	0.39 (0.20–0.79)	0.42 (0.21–0.84)	0.39 (0.19–0.79)
30–39	0.87 (0.56–1.37)	0.58 (0.35–0.95)	0.61 (0.37–1.00)	0.57 (0.35–0.94)
40+	1.00	1.00	1.00	1.00
Race/ethnicity
African American	2.78 (1.37–5.63)	2.39 (1.11–5.15)	2.50 (1.16–5.39)	2.49 (1.15–5.38)
Asian or Pacific Islander	1.50 (0.63–3.59)	1.39 (0.59–3.31)	1.23 (0.49–3.08)	1.41 (0.59–3.36)
Latina, US born	1.17 (0.53–2.56)	0.92 (0.43–2.00)	0.95 (0.44–2.04)	0.91 (0.42–1.97)
Latina, immigrant	5.66 (3.38–9.49)	1.66 (0.85–3.26)	1.76 (0.90–3.45)	1.68 (0.85–3.30)
White, non-Hispanic	1.00	1.00	1.00	1.00
Marital status
Previously or never married	2.20 (1.37–3.52)	1.05 (0.63–1.75)	1.06 (0.62–1.80)	1.07 (0.63–1.81)
Married or living together	1.00	1.00	1.00	1.00
# of children in household
0–1	1.00	1.00	1.00	1.00
2–3	0.83 (0.43–1.62)	0.78 (0.42–1.44)	0.86 (0.46–1.63)	0.77 (0.42–1.41)
4+	1.01 (0.47–2.20)	0.60 (0.28–1.28)	0.68 (0.31–1.48)	0.59 (0.27–1.28)
Education
Less than high school	8.20 (4.78–14.08)	2.52 (1.25–5.08)	2.40 (1.16–4.96)	2.56 (1.24–5.30)
High school graduate/GED	2.93 (1.64–5.25)	1.31 (0.66–2.61)	1.26 (0.61–2.59)	1.36 (0.66–2.80)
Some college	0.98 (0.49–1.94)	0.62 (0.30–1.26)	0.64 (0.32–1.30)	0.64 (0.31–1.30)
College graduate	1.00	1.00	1.00	1.00
Income, % of federal poverty level
< = 100%	7.55 (4.38–13.02)	4.00 (1.70–9.39)	4.22 (1.85–9.60)	4.17 (1.79–9.70)
101–200%	2.64 (1.41–4.95)	1.96 (0.83–4.65)	2.09 (0.91–4.81)	1.98 (0.85–4.61)
201–400%	0.80 (0.38–1.67)	0.88 (0.37–2.07)	0.85 (0.35–2.03)	0.87 (0.37–2.06)
> 400%	1.00	1.00	1.00	1.00
Part-time employment
< 40 hours/week	2.44 (1.61–3.70)	1.70 (1.08–2.67)	1.54 (0.95–2.47)	1.69 (1.04–2.74)
+40 hours/week	1.00	1.00	1.00	1.00
LTPA
Inactive	2.83 (0.75–10.69)	1.90 (0.49–7.35)		1.89 (0.49–7.32)
Light activity, 1–2× per week	4.00 (1.35–11.85)	2.52 (0.92–6.92)		2.50 (0.90–6.96)
Moderate activity, 3× per week	3.65 (1.22–10.93)	2.61 (0.95–7.20)	–	2.61 (0.93–7.27)
Moderate activity, 5+× per week	3.03 (0.89–10.35)	2.44 (0.71–8.47)		2.65 (0.76–9.26)
Vigorous activity, 3× per week	1.10 (0.29–4.28)	1.18 (0.32–4.40)		1.19 (0.31–4.50)
Vigorous activity, 5+× per week	1.00	1.00		1.00
Work physical activity
Sit or stand most of day	1.00		1.00	1.00
Physical labor	2.14 (1.20–3.83)	–	0.73 (0.37–1.44)	0.69 (0.34–1.39)
Walked or used arms	2.39 (1.51–3.77)		1.10 (0.63–1.92)	1.08 (0.62–1.89)

Note. Bold face type indicates a significant relationship between variables.

^aAdjusted for age, race/ethnicity, marital status, # of children in household, education, income, part-time employment, and LTPA.

^bAdjusted for age, race/ethnicity, marital status, # of children in household, education, income, part-time employment, and work physical activity.

^cAdjusted for all variables.

3. Results

Table 1 presents characteristics of the sample stratified by employment status. The average age of the mothers was about 36 years, and did not vary by employment status. By race/ethnicity, higher proportions of employed women were Asian/Pacific Islander and White, non-Hispanic compared with unemployed women. In contrast, higher proportions of unemployed women were immigrant Latina compared with employed women. The large majority of women were married or living with a partner, and proportions of married women were higher for unemployed compared with employed women. On average, women had about three children living with them. Among unemployed women, a higher proportion were of low SES, based on education and income, compared with employed women. For example, 45% of unemployed women had incomes below the federal poverty level compared with 20% of employed women. Over half of employed women worked in sedentary jobs, and their levels of LTPA were similar to non-employed women. Few employed women engaged in non-work and work AT (9%), compared with over 20% of non-work AT among unemployed women.

Table 1.

Relationships between sociodemographic characteristics, physical activity and employment status, Geographic Research on Wellbeing (GROW) study (n = 2,906).

	Employed (n = 1,628)	Unemployed (n = 1,278)
Age (mean)	36.5 (36.2–36.9)	35.6 (35.2–36.0)
Race/ethnicity (%)
African American	6.9 (6.2–7.7)	6.3 (5.3–7.2)
Asian or Pacific Islander	18.9 (16.5–21.4)	10.8 (8.6–13.0)
Latina, US born	17.8 (15.8–19.8)	14.1 (12.0–16.1)
Latina, immigrant	28.0 (25.6–30.5)	46.4 (43.7–49.2)
White, non-Hispanic	28.4 (26.3–30.5)	22.5 (20.4–24.6)
Marital status (%)
Previously or never married	19.0 (16.9–21.2)	14.2 (12.1–16.3)
Married or living together	81.0 (78.8–83.2)	85.8 (83.7–87.9)
# of children in household (mean)	2.6 (2.5 – 2.7)	3.1 (3.0–3.2)
Education (%)
Less than high school	13.9 (12.0–15.8)	27.7 (25.1–30.2)
High school graduate/GED	18.6 (16.4–20.8)	27.4 (24.5–30.3)
Some college	25.2 (22.9–27.6)	20.3 (18.0–22.6)
College graduate	42.3 (39.8–44.8)	24.7 (22.4–27.0)
Income (% of the federal poverty level)
< = 100%	20.3 (18.1–22.5)	44.6 (41.7–47.5)
101–200%	20.0 (17.8–22.2)	21.4 (18.7–24.0)
201–400%	22.7 (20.4–25.0)	14.6 (12.5–16.6)
>400%	37.0 (34.5–39.4)	19.5 (17.4–21.6)
Leisure time physical activity
Inactive	6.1 (4.6–7.6)	4.9 (3.6–6.2)
Light activity, 1–2× per week	37.9 (35.3–40.6)	42.7 (39.7–45.8)
Moderate activity, 3× per week	30.5 (28.0–33.0)	30.2 (27.4–33.1)
Moderate activity, 5+ times per week	7.6 (6.3–9.0)	10.2 (8.5–12.0)
Vigorous activity, 3× per week	10.5 (8.9–12.1)	5.8 (4.5–7.1)
Vigorous activity, 5+ times per week	7.3 (6.0–8.7)	6.1 (4.7–7.5)
Work physical activity (%)
Sit or stand most of day	52.6 (49.9–55.3)
Walked or used arms	33.4 (30.8–36.1)	–
Physical labor	14.0 (12.1–15.9)
Non-work AT (%)
Drove, got a ride	91.4 (89.8–93.1)	79.2 (76.7–81.6)
Walk, bike, public transportation	8.6 (6.7–10.2)	20.8 (19.4–23.3)
Work AT (%)
Drove, got a ride	90.8 (89.1–92.5)	–
Walked, biked, took public transportation	9.2 (7.5–10.9)

Table 2 presents prevalences of non-work and work AT. Patterns of observed AT were similar for non-work and work AT. Immigrant Latinas used AT over four times as frequently as US born Latinas. African American mothers were the second highest users of AT. Mothers who were not married and those with lower SES used AT more than their counterparts. Those mothers who did light to moderate LTPA used more AT than those who were inactive or vigorously active, and those with more WPA used more AT than those with more sedentary work.

Table 2.

Prevalences of active transportation, Geographic Research on Wellbeing (GROW) study (n = 2,906).

	% using Active Transportation for non-work	% using Active Transportation for work*
Age (%)
20–29 years	14.2	8.5
30–39 years	15.1	8.9
40+ years	12.8	10.0
Race/ethnicity (%)
African American	16.6	10.2
Asian or Pacific Islander	4.8	5.8
Latina, US born	6.6	4.6
Latina, immigrant	28.4	18.8
White, non-Hispanic	3.9	3.9
Marital status (%)
Previously or never married	19.8	15.4
Married or living together	13.1	7.6
# of children in household
0–1	13.0	10.0
2–3	12.8	8.5
4+	17.7	10.1
Education (%)
Less than high school	33.4	27.7
High school graduate/GED	19.9	12.0
Some college	7.5	4.4
College graduate	3.9	4.5
Income (% of the federal poverty level)
< = 100%	31.4	23.5
101–200%	13.9	9.7
201–400%	3.3	3.1
> 400%	2.9	3.9
Leisure time physical activity
Inactive	11.7	8.3
Light activity, 1–2× per week	16.0	11.3
Moderate activity, 3× per week	15.7	10.4
Moderate activity, 5+ times per week	15.4	8.8
Vigorous activity, 3× per week	5.3	3.4
Vigorous activity, 5+ times per week	8.7	3.1
Work physical activity (%)
Sit or stand most of day	4.1*	6.0
Walked or used arms	13.1*	13.1
Physical labor	15.0*	11.9

^*among employed women, n = 1,628

Table 3 presents odds ratios for non-work AT in unadjusted and adjusted models, stratified by employment status. In unadjusted models for employed women, immigrant Latinas, unmarried women, women with low education or income, and women who are sedentary or who report moderate LTPA 3x/week had higher odds of non-work AT compared with their reference groups. However, in the adjusted model the only significant associations that remained were for low SES women. In addition, younger women had marginally lower odds of non-work AT compared with women aged 40 and over. Similar to employed women, low SES unemployed women also had higher odds of non-work AT in the adjusted model, and younger women had lower odds compared with their reference groups. African American and immigrant Latinas had about 3.5 times higher odds of non-work AT compared with non-Hispanic White women, and women with 4 or more children as well as those reporting light LTPA had lower odds of non-work AT compared with women with 0–1 children or those reporting the most LTPA, respectively.

Table 4 presents similar models for work AT among employed women. LTPA and work physical activity were both associated with work AT in unadjusted models. For example, inactive women during leisure time had nearly 4 times higher odds of work AT compared with the most active women. While LTPA remained significant after adjusting for sociodemographic characteristics (“LTPA” model), the effect of work physical activity was no longer significant after adjusting for sociodemographic characteristics (“Work physical activity” model). In the “Full” model adjusting for both LTPA and work physical activity, we found that African American, women with less than a high school degree or with incomes below the federal poverty level, those working part-time, and both inactive and those reporting moderate activity 3 times/week during their leisure time had higher odds of work AT compared with their reference groups. In contrast, younger compared with older women had lower odds of work AT.

4. Discussion

The purpose of this manuscript was to investigate systematically the relationships among participation among different domains of physical activity accounting for a range of sociodemographic characteristics, including employment status, in a representative sample of California mothers. Findings suggested that participation in physical activity among different domains were related, possibly pointing to a more “active type” of woman rather than a compensatory strategy: LTPA was associated with non-work AT among unemployed mothers and with work AT among employed mothers after accounting for sociodemographic factors. Patterns of LTPA were similar among employed and non-employed mothers, in contrast to what others have found, possibly related to measurement or sample characteristics (i.e., motherhood) (Kwak et al., 2016; Van Domelen et al., 2011). WPA was not related to work-related AT among working mothers. These relationships differed somewhat by employment status, but work and non-work AT were strongly related, suggesting women who use AT do so for most trips.

This study demonstrates that among California mothers in urbanized areas, lower SES is associated with higher participation in AT, which is somewhat inconsistent with the findings from the literature review of European studies (Beenackers et al., 2012) but similar to other studies in the US (Berrigan et al., 2006; Tudor-Locke and Ham, 2008). Further, employed and unemployed mothers with low education and income and unemployed African American or Latina immigrant mothers had higher odds of using non-work AT compared with their reference groups. It is likely that lower SES contributed more to these relationships given the confounding between SES and race or ethnicity. For work AT, mothers who were younger and unemployed women who had four or more children or had “light” LTPA had lower odds of using non-work AT compared with their reference groups. These findings suggest that AT may occur primarily among those with lower socioeconomic position, and it might reflect fewer alternative transportation options (i.e., personal car ownership). As well, transporting children via AT can be cumbersome and stressful, possibly making AT an unattractive option unless absolutely necessary. Very few mothers in the sample responded that they rode a bicycle to get from place to place (n=8, data not shown), possibly suggesting low bicycle ownership or lack of cycling infrastructure. Increasing bicycle access, promoting the desirable benefits of AT (e.g., burns calories, improves air quality, reduces stress), and improving the quality of the pedestrian and cycling infrastructure and public transportation systems could help to increase AT. Of those who did use AT to get from place to place, public transportation was an important part of their travel: 215 used public transportation compared to 142 who walked (data not shown).

Different AT patterns were seen for employed versus unemployed women. Over half of employed women worked in sedentary jobs, and employed women had relatively low levels of both non-work and work AT (9%), compared with unemployed women who reported over 20% of non-work AT. Although WPA was associated with work AT in unadjusted models, it was not associated with work AT in the adjusted models, suggesting that physical activity done at work and during AT were not themselves directly related. However, both WPA and AT appeared to be largely related to socioeconomic status. Working mothers who were African American, and those with less than a high school education or had lower incomes were much more likely to be in labor intensive jobs and use AT to get to work, which is consistent with other reports of people from lower socioeconomic position (Beenackers et al., 2012; Hadgraft et al., 2015; Vandelanotte et al., 2013; Vandelanotte et al., 2015). Future research might investigate whether this additional physical activity done in lower SES populations might have a buffering effect against developing health compromising conditions associated with inactivity; however, these groups may also experience higher stress from less leisure time and lower access to resources which could counteract a buffering effect. Regardless, the fact that most women were generally low across all domains of physical activity suggests that there was not an appreciable compensatory effect of physical activity done in one domain suppressing physical activity not done in another. Thus, most mothers of young children at all socioeconomic strata are in need of innovative intervention strategies to meet physical activity guidelines.

4.1. Strengths and limitations

Strengths of the study include a sizable, diverse and representative sample of women with young children in urbanized areas of California (Cubbin, 2015) and carefully crafted, translated, pilot tested and administered surveys. Although the measures were carefully selected to be the best available to balance reliable and valid assessments without compromising participant burden, only self-reported information was available. Self-reported information may produce measurement bias. For example, although the question measuring leisure time physical activity specifically instructed respondents not to include physical activity done at work, there is no way to know whether respondents understood or read these instructions. Method of administration (phone vs mail) may also have affected responses to LTPA; however, sensitivity analyses did not reflect important bias (results available upon request). This study solely included measures of individual SES, and inclusion of area SES might help to explain additional variability in outcomes. Minor differences were seen in the direction of the odds for a few variables by method of administration compared with the combined sample; however, our conclusions regarding significant effects are unchanged. Differences likely reflect increased variability based on smaller sample size rather than a true difference by administration.

4.2. Conclusion

Measures of physical activity may be underestimating physical activity among low SES women and some racial/ethnic groups if investigators only measure leisure-time or work-related physical activity. Policy and environment strategies should support a variety of messages to encourage these unique and seemingly unrelated domains of physical activity. Public health messages should be framed to normalize and enhance perceptions of AT infrastructure and behavior to encourage broader adoption across all sectors of the population. This study found no support for the notion that people who are more active in one domain may compensate with less activity in another domain. Rather, use of AT in this diverse sample of employed and unemployed California mothers appears to be largely related to SES differences such that those who may not have other options are more likely to use AT.