Individual, social environmental, and physical environmental ...

Download PDF
2 downloads 0 Views 185KB Size Report
Comment
Individual, Social Environmental, and Physical Environmental Influences on Physical Activity. Among Black and White Adults: A Structural Equation Analysis.
Individual, Social Environmental, and Physical Environmental Influences on Physical Activity Among Black and White Adults: A Structural Equation Analysis Lorna Haughton McNeill, Ph.D., M.P.H. School of Public Health Harvard University

Kathleen W. Wyrwich, Ph.D. Department of Research Methodology Saint Louis University

Ross C. Brownson, Ph.D. Department of Community Health Saint Louis University School of Public Health

Eddie M. Clark, Ph.D. Department of Psychology Saint Louis University

Matthew W. Kreuter, Ph.D., M.P.H. Department of Community Health and Health Communication Research Laboratory Saint Louis University School of Public Health

forms of activity, self-efficacy was the strongest direct correlate of physical activity, and evidence of a positive dose–response relation emerged between self-efficacy and intensity of physical activity. Conclusions: Findings from this research highlight the interactive role of individual and environmental influences on physical activity.

ABSTRACT Background: Social ecological models suggest that conditions in the social and physical environment, in addition to individual factors, play important roles in health behavior change. Using structural equation modeling, this study tested a theoretically and empirically based explanatory model of physical activity to examine theorized direct and indirect effects of individual (e.g., motivation and self-efficacy), social environmental (e.g., social support), and physical environmental factors (e.g., neighborhood quality and availability of facilities). Method: A community-based sample of adults (N = 910) was recruited from 2 public health centers (67% female, 43% African American, 43% < $20,000/year, M age = 33 years) and completed a self-administered survey assessing their current physical activity level, intrinsic and extrinsic motivation for physical activity, perceived social support, self-efficacy, and perceptions of the physical environment. Results: Results indicated that (a) perceptions of the physical environment had direct effects on physical activity, (b) both the social and physical environments had indirect effects on physical activity through motivation and self-efficacy, and (c) social support influenced physical activity indirectly through intrinsic and extrinsic motivation. For all

(Ann Behav Med

2006, 31(1):36–44)

INTRODUCTION Physical inactivity is a significant public health problem, contributing to over 280,000 deaths per year in the United States (1). Recent federal initiatives and programs, in addition to national health objectives in Healthy People 2010, have been developed to track and promote physical activity in the United States. Yet, rates of leisure time inactivity have remained relatively stable from 1989 to 1996 and are now slowly declining at a rate of 1% per year (2). Among minorities, women, and individuals of low socioeconomic position, the prevalence of physical inactivity remains large, reflecting a continuing pattern of social and racial/ethnic disparities in health behaviors (3). Also in recent years, the prevalence of overweight and obesity among adults has been increasing (4), prompting epidemiologists and social scientists to seek a better understanding of the factors that influence adult physical activity so that health promotion programs and policies can more effectively target specific contributing factors (5,6). Although it is widely recognized that there are both environmental and individual determinants of physical activity, explanatory models have only seldom explicitly included both types of variables (7–9). This is a growing area of research with few studies assessing how these types of factors are interrelated and interact to explain activity. Social Cognitive Theory and ecological models describe behavior as a dynamic process that

This research was supported by grant R06/CCR71721602 from the Centers for Disease Control and Prevention, and the CDC/ASPH Minority Fellowship Program (U48/CCU710806, Prevention Research Centers Program). We also thank Drs. Nancy Krieger and Karen Emmons for their review of early drafts of this article. Reprint Address: L. Haughton McNeill, Ph.D., M.P.H., Society, Human Development, and Health, School of Public Health, Harvard University, 677 Huntington Avenue, 7th Floor, Boston, MA 02115. E-mail: [email protected] © 2006 by The Society of Behavioral Medicine.

36

Volume 31, Number 1, 2006 is simultaneously influenced by aspects of the social and physical environments as well as personal attributes of individuals (10–13). To date, however, explanatory models of adult physical activity have focused primarily on personal attributes (e.g., self-efficacy) and less so on aspects of the social environment, such as social support (14). Combined with sociodemographic variables, these factors explain only a small portion of the variance in physical activity behavior, indicating the potential for other important contributing factors (15). The inclusion of environmental factors in physical activity models is theoretically and empirically justified and an important next step in physical activity research (16). Although Social Cognitive Theory and ecological models specify a direct relationship between one’s environment and behavior, it is also important to understand the role of mediators in the relationship between environmental factors and physical activity. For example, though many empirical studies have shown a direct relationship between the physical environment and physical activity, the strength of this association has been shown to be attenuated in the presence of social cognitive factors, such as self-efficacy (17). Therefore, to increase our understanding of physical activity correlates, this study tested an explanatory model of physical activity hypothesizing direct and indirect effects of individual, social environmental, and physical environmental influences on physical activity.

METHOD Study Design, Population, and Data Collection Procedures Structural equation modeling was used to test a series of mediating relationships, elucidating the potential pathways by which social cognitive and ecological factors influence physical activity. This study used baseline data from a 3-year randomized trial funded by the Centers for Disease Control and Prevention, completed in 2003. The trial, “Optimal Segmentation Strategies in Health Communication,” evaluated the effectiveness of three different targeted magazines on physical activity (18). Participants were 1,090 African American and White lower and middle-income adults, recruited from the waiting rooms of two public health centers in the St. Louis, Missouri, area during a 3-month period in spring 2002. Criteria for participation included the participant being between 18 and 65 years of age, being either African American or White, having daily access to a working telephone (for follow-up telephone interviews conducted in the intervention trial), and being able to read material written at the fifth-grade level. At both health centers, individuals were approached by trained graduate assistants, given a description of the project, and offered the opportunity to participate. Potential participants were basically “healthy” adults who either had a routine appointment at the health center (for themselves or their children) or were there with someone for such an appointment (i.e., family members or friends). If interested, potential participants were screened for eligibility, and they gave written informed consent to participate. The baseline survey was self-administered.

Influences on Physical Activity

37

Description of the Proposed Conceptual Model Separate models were developed for three physical activity outcomes: walking, moderate-intensity activity, and vigorous-intensity activity. These models (see Figures 1–3) were developed to test the hypothesized relationships between individual (e.g., self-efficacy and intrinsic and extrinsic motivation), social environmental (e.g., social support), and physical environmental correlates (e.g., neighborhood quality and availability of facilities) of physical activity. The correlates used in developing the models are based on models and theories of health behavior (10,11) in addition to empirical evidence that demonstrates a relationship between these factors and physical activity (14,19–21). On the basis of this evidence, we proposed three major hypotheses. First, we hypothesized that the relationship between social support and physical environmental factors, and physical activity would be mediated by individual-level factors (19,20). Second, we hypothesized that social support would positively influence self-efficacy directly, which in turn would influence physical activity. Past and more recent studies have demonstrated that social support may influence physical activity indirectly through self-efficacy (21,22). Third, we hypothesized that physical environmental factors would positively influence physical activity directly (6,23). These hypotheses test whether the model would hold for the total population of adults, regardless of age, gender, race/ethnicity, education, and income. Future analyses will examine effect modification by sociodemographic factors. Measures Individual-Level Variables: Self-Efficacy and Motivation Self-efficacy. This construct was measured using several items developed by Marcus, Selby, Niaura, and Rossi (24) and supplemented by additional items developed for the study, resulting in a seven-item scale. Participants were asked to rate their level of confidence using a 5-point Likert scale from 1 (strongly disagree) to 5 (strongly agree) that they can be physically active under the following conditions: feeling too tired to be physically active, poor weather conditions, lack of time, being injured, recovering from an injury, feeling sick, and having sore muscles. The seven items extracted a single factor in exploratory factor analysis, and internal consistency of the scale was acceptable (α = .87). Motivation. Intrinsic and extrinsic motivation for physical activity was measured using items from the Motivation for Physical Activities Measure (25) and supplemented by additional items developed for the study, resulting in an 18-item scale. Intrinsic motivation is behavior engaged in for pleasure and enjoyment, whereas extrinsic motivation is behavior engaged in for reasons outside the activity itself or as a means to an end, such as engaging in physical activity to produce weight loss (26). Participants were asked what they thought about physical activity and whether they agreed or disagreed (on a 5-point Likert scale from 1 [strongly disagree] to 5 [strongly agree]) with several statements. Examples include “I think physical ac-

38

McNeill et al.

tivity is fun,” “I want to be physically active to control my weight,” and “I want to be physically active so that others will find me more attractive.” Exploratory factor analysis identified three factors: intrinsic motivation (11 items, α = .92), extrinsic motivation for social pressure (3 items, α = .52), and extrinsic motivation for peer acceptance (3 items, α = .58). Social Environmental Variables: Social Support Social support. Social support was measured using items from a scale originally developed by Sallis, Grossman, Pinksi, Patterson, and Nader (27) and modified by Eyler et al. (28) for use with racially/ethnically diverse populations (e.g., African American, Hispanic, American Indian/Alaskan Native, and Whites). The items in this modified version assessed both emotional (three items, α = .77) and informational (three items, α = .71) social support using a 5-point Likert scale that measured participants’ agreement or disagreement from 1 (strongly disagree) to 5 (strongly agree) with the following statements: “There are people in my life willing to do physical activity with me,” “I’d be more physically active if I had help with other responsibilities in my life,” “My family members are supportive of me being physically active,” “I’d be physically active if I had more information about it,” “My friends are supportive of me being physically active,” and “I’d be active if my family or friends were more supportive.” Physical Environmental Variables: Neighborhood Quality and Access to Facilities Physical environment. Participants’ perceptions of their environment were measured using a seven-item measure developed by Brownson et al. (29). This scale measured availability of physical activity facilities (four items, α = .68) and neighborhood quality (three items, α = .79). Participants were asked to rate the quality of their neighborhood in reference to criminal activity, traffic, and pleasantness for engaging in physical activity using a 4-point Likert scale from 1 (very unsafe/unpleasant) to 4 (very safe/pleasant). To assess availability of facilities, participants were asked if their neighborhood had walking/biking trails, parks, and outdoor/indoor places to exercise. Participants responded by answering yes or no. Dependent Variable: Physical Activity Physical activity. Three dimensions of physical activity were measured in this study. Walking was defined as time spent walking for work, recreation, exercise, to get to places, or for any other reason; moderate-intensity activity included such tasks as brisk walking, bicycling, vacuuming, gardening, or any other activity that causes small increases in breathing and heart rate; and vigorous-intensity activity included running, aerobics, heavy yard work, or anything else that causes large increases in breathing or heart rate. Walking, moderate-intensity physical activity, and vigorous-intensity physical activity were assessed using lifestyle activity items from the Behavioral Risk Factor Surveillance System Survey (30) that captures elements of

Annals of Behavioral Medicine household- and transported-related physical activity in addition to leisure time activity (2). These items have been used in numerous national studies and have been previously validated (31). Participants were asked to respond to two questions: How many days per week do you walk for at least 10 minutes at a time, and on days when you walk for at least 10 minutes at a time, how much total time do you spend walking? Response options for the first question ranged from 0 to 7 days per week, and response options for the second question ranged from 10 min to 60 or more min (in 10-min increments). The midpoint of the response range was then used as the time spent walking. For example, the midpoint between 10 min to 20 min is 15 min. Responses to Questions 1 and 2 were then multiplied to form a continuous measure of total minutes engaged in walking per week (min per day × days per week), which represents a single-item indicator serving as the dependent variable. The same two-question format was used to assess weekly minutes of moderate-intensity and vigorous-intensity physical activity. Total possible minutes ranged from 0 to 455 min per week for each form of activity (moderate intensity, vigorous intensity, and walking). Sociodemographic variables. Age, educational attainment (both measured in years), race/ethnicity (White or African American/Black), gender, and pretax household income during the preceding year (< $5,000 to > $60,000 in eight categories) were measured. Statistical Analysis Plan and Treatment of Missing Data Structural equation modeling (SEM) using LISREL for Windows, version 8.52 (32) was used to test the fit of the hypothesized physical activity models (Figure 1) to the data collected from participants. SEM is an extension of the general linear model, in which variables are assumed to have an additive linear relationship (33), and it enables researchers to test a series of regression equations simultaneously (34). It is designed to test the theoretical relationships between latent constructs and allows the testing of direct and indirect effects, or mediating relationships among variables. Unlike regression analysis, where the independent variables are equally assumed to be measured perfectly, SEM assumes that all variables measured have some measurement error and are accounted for in the explanatory model (34). A strength of SEM is that the estimated relations among latent variables are not biased because of unreliability in the construct indicators (34). However, it is noted that SEM has the same limitations of any other type of analyses applied to cross-sectional data (i.e., inability to assess causal relationships). The chi-square statistic, based on the maximum likelihood method of estimation, is commonly used to determine if the model is a good representation of the process that generated the data in the population (35). A small chi-square with a nonsignificant p value relative to degrees of freedom indicates a good-fitting model. In this case, a nonsignificant result is desir-

Volume 31, Number 1, 2006 able because it indicates that the hypothesized model and the underlying covariance structure matrix are the same. We used the Sattora–Bentler scaled chi-square statistic, which adjusts the analyses for any non-normality present by weighting the original data using the asymptotic correlation matrix, thereby making the chi-square statistic and standard errors less biased (37). Chi-square is also sensitive to sample size. Therefore, several fit indexes have been proposed to compensate for large sample sizes that increase power to reject models. Four fit indexes will be used. The Bentler–Bonnett non-normed fit index (NNFI) and the comparative fit index (CFI) both compare the fit of the estimated model to that of a baseline or null model. Using these indices, the null value (0) specifies that all measured variables are uncorrelated, there are no latent constructs, and a value of 1.0 represents perfect fit. In general, a value of .90 or greater indicates a good-fitting model (35). The root mean square error of approximation (RMSEA) is a measure of discrepancy between the true population model and the hypothesized model per degree of freedom, thus favoring a more parsimonious model. A value of less than .08 indicates an acceptable fit, and a value of less than .05 indicates a good fit (35). Last, the standardized root mean residual (SRMR) captures the average difference between the observed correlation and the model-implied correlation. In general, a value of less than .08 indicates an acceptable fit and a value of less than .05 indicate a good fit (35). SEM requires complete data with no missing values on any variable (e.g., sociodemographic or social cognitive construct) to conduct model testing. Of the 1,090 participants who completed baseline questionnaires, 66% (n = 727) provided complete data. The average number of missing values per variable was 19, with “income” having the most missing values (n = 62)

Influences on Physical Activity

and “current level of physical activity” having the fewest (n = 4). To minimize the exclusion of participants from the sample, imputation of missing ordinal and continuous data values was performed using the appropriate methods for each type of data (32). LISREL has a program component named PRELIS that is used to screen and summarize data before specifying the actual model in LISREL (37). Using PRELIS, missing values are imputed by matching on other variables used in the model. This resulted in an effective sample size of 910. Sensitivity analyses comparing initial and final samples found that excluded participants (n = 180) were more likely to be African American (65% vs. 35%, p =.00), be employed (59% vs. 41%, p =.03), report greater neighborhood quality (Ms = 10.0 vs. 9.5, p =.00) and report less informational support (Ms = 9.2 vs. 9.8, p =.00). RESULTS Participants Descriptive characteristics of the study population and predictor variables are shown in Table 1. The study sample was predominantly female (n = 727, 67%) and lower income (43% < $20,000/year), with a mean age of 33 years (SD = 13.1, Mdn = 28) and 13 years of education (SD = 2.1). African Americans made up 43.2% of the sample. Mean total minutes of moderate-intensity activity was 122.0 per week, with almost 30% meeting the Healthy People 2010 goal for moderate-intensity activity (minimum of 150 min of activity per week), and mean total minutes of vigorous-intensity activity per week was 61.0, with about 34% meeting the Healthy People 2010 recommendations (minimum of 60 min per week). Overall, 46.5% met the recommendations for either moderate-intensity or vigor-

TABLE 1 Characteristics of Study Participants and Health Parameter Measures, 2002 Characteristics Gender (% women/men) Age group (% 18–29/30–65) Race/ethnicity (% White/African American) Educational attainment (% < high school/> high school) Annual household income (% < $20,000/> $20,000) Social support (M ± SD) Information support (range = 3–15) Emotional support (range = 3–15) Motivation (M ± SD) Intrinsic motivation (range = 12–60) Extrinsic motivation for social pressure (range = 3–15) Extrinsic motivation for peer acceptance (range = 3–15) Self-efficacy (range = 7–35) (M ± SD) Physical environment (M ± SD) Availability of facilities (range = 0–4) Neighborhood quality (range = 3–12) Physical activity (M ± SD) Walking (range = 0–455) Moderate-intensity activity (range = 0–455) Vigorous-intensity activity (range = 0–455) Note. N = 910.

39

67.7/32.3 48.0/52.0 56.8/43.2 55.6/44.4 41.4/58.6 9.3 ± 2.4 11.0 ± 2.3 44.3 ± 8.5 11.9 ± 2.1 9.9 ± 2.4 21.9 ± 5.8 2.6 ± 1.3 9.9 ± 1.9 154.1 ± 141.8 122.0 ± 117.9 61.0 ± 95.0

40

McNeill et al.

ous-intensity activity; this is similar to the 2001 national average of 45.1% (2). Measurement and Structural Equations Prior to modeling any relationships between constructs, measurement models were evaluated using confirmatory factor analysis to confirm the factor structure of the all model constructs. The overall fit of the measurement models were adequate based on standard fit index criteria (NNFI > .90, CFI > .90, RMSEA < .08, SRMR < .08). The confirmatory factor analysis factor loadings were significant at the .05 level and ranged between .74 and .79, .46 and .84, .49 and .85, and .67 and .89, for self-efficacy, intrinsic and extrinsic motivation, emotional and informational social support, and availability and neighborhood quality of the physical environment, respectively. The latent construct physical activity was represented with a single-item indicator. As a result, the error variance was set to variance x (1 – α), where α represents the test–retest reliability estimate of physical activity and variance is the observed variance of the item (36). The resulting error variances were .58, .48, and .66, for walking, moderate-intensity activity, and vigorous-intensity activity, respectively. The overall fit of the structural models predicting walking, moderate-intensity activity, and vigorous-intensity activity were also adequate based on the same standard fit criteria. Walking As proposed in the theoretical model, both social environmental factors (e.g., emotional and informational social support) influenced all dimensions of motivation for physical activity (see Figure 1). All paths between emotional support and social pressure (β = 0.382, t = 6.52), peer acceptance (β = 0.356, t = 6.32), and intrinsic motivation (β = 0.492, t = 10.71) were positive and statistically significant. The association between informational support and intrinsic motivation was not significant, indicating that provision of advice and suggestions was not directly associated with intrinsic reasons for physical activity (e.g., enjoyment and goal attainment); however, informational support was directly associated with both dimensions of extrinsic motivation (social pressure, β = 0.294, t = 4.70; peer acceptance, β = 0.679, t = 10.69) for the model of walking behavior. However, contrary to our hypothesis, the association between social support and self-efficacy was not statistically significant, providing evidence that self-efficacy did not directly mediate the relationship between social support and walking but instead indirectly mediated this relationship through intrinsic motivation (β = 0.289 indirect vs. β = 0.174 direct). Also, as proposed in the model, both physical environmental factors (e.g., neighborhood quality and availability of facilities) were associated with motivation for physical activity. However, this direct association was found for intrinsic motivation only—one negatively and the other positively, respectively. The inverse association between neighborhood quality and intrinsic motivation suggests that greater quality of the neighborhood

Annals of Behavioral Medicine χ

FIGURE 1 Individual, social, and physical environmental influences on walking. Only statistically significant paths are included in this figure. All bold paths are statistically significant at p < .05. RMSEA = root mean square error of approximation; CFI = comparative fit index; NNFI = non-normed fit index; SRMR = standardized root mean residual.

(e.g., low traffic and crime) was associated with less intrinsic motives for physical activity. Yet availability of facilities was associated with more intrinsic motives. When assessing the direct relationship between the physical environment and walking, availability of physical activity facilities was associated with more walking (β = 0.269, t = 6.74), but neighborhood quality was not associated with walking. Self-efficacy had the greatest total effect on physical activity. Results indicate a positive moderate direct effect on walking (β = 0.281, t = 5.12). There was also a strong association between intrinsic motivation and self-efficacy. Results indicate that greater intrinsic motivation, or engaging in activity for enjoyment and goal attainment, was positively associated with self-efficacy. Extrinsic motivation due to social pressure and self-efficacy were inversely associated.

Moderate-Intensity Activity The moderate-intensity activity model yielded results similar to the walking model (see Figure 2). Both dimensions of social support were positively associated with motivation. Similar to walking, intrinsic motivation was a significant mediator between physical environmental correlates and physical activity. However, unlike walking, both neighborhood quality and availability were directly associated with moderate-intensity physical activity (neighborhood quality, β = 0.135, t = 2.57; availability, β = 0.137, t =3.42), though this effect is marginal. In this model, self-efficacy was more strongly associated with moderate-intensity activity (β = 0.353) compared to walking (β = 0.281).

Volume 31, Number 1, 2006

Influences on Physical Activity χ

41

χ

FIGURE 2 Individual, social, and physical environmental influences on moderate-intensity physical activity. Only statistically significant paths are included in this figure. All bold paths are statistically significant at p < .05. RMSEA = root mean square error of approximation; CFI = comparative fit index; NNFI = non-normed fit index; SRMR = standardized root mean residual.

FIGURE 3 Individual, social, and physical environmental influences on vigorous-intensity physical activity. Only statistically significant paths are included in this figure. All bold paths are statistically significant at p < .05. RMSEA = root mean square error of approximation; CFI = comparative fit index; NNFI = non-normed fit index; SRMR = standardized root mean residual.

Vigorous-Intensity Activity Similar to the findings for both walking and moderate-intensity activity, social support was directly associated with motivation (see Figure 3). Physical environmental correlates were also associated with vigorous-intensity activity indirectly through intrinsic motivation. However, unlike walking and moderate-intensity activity, neighborhood quality was the only physical environmental correlate associated with vigorous-intensity activity (β = 0.104, t = 2.52). Self-efficacy exhibited a strong direct effect on vigorous-intensity activity (β = 0.443), demonstrating a slightly greater influence on vigorous-intensity activity than either walking or moderate-intensity activity. The structural equations illustrated that for all physical activity outcomes, social and physical environment dimensions accounted for 27% of the variance in intrinsic motivation, 26% of the variance in extrinsic motivation for social pressure, and 63% of the variance in extrinsic motivation for peer acceptance. Social environmental factors and motivational dimensions accounted for 42% of the variance in self-efficacy, and self-efficacy and physical environment dimensions accounted for 15% of the variance in walking, 18% of the variance in moderate-intensity activity, and 21% of the variance in vigorous-intensity activity.

through individual-level factors, which in turn influenced physical activity; likewise, physical environmental factors were found to have a direct effect on physical activity. Although many studies have incorporated both individual and social environmental measures, this latter finding suggests that including environmental factors in social ecological models of physical activity may provide evidence of its relative influence on activity behavior, specifically when in the presence of other wellestablished correlates (38). We proposed that self-efficacy would mediate the relationship between social support and activity, as has been found in other studies. Prior studies conducted among older adults found an association between social support and physical activity through self-efficacy (22,39). However, our study population was adults ages 18 through 65, which suggests that the relationship between social support and self-efficacy may be different in different populations. Other studies among sedentary individuals have also shown that self-efficacy mediates the relationship between social support and activity (21,40). Our diverse sample was moderately active, and our findings support the hypothesis that social support may influence physical activity indirectly through motivational reasons for engaging in activity. Social support was also associated with motivational reasons of social pressure and peer acceptance for engaging in physical activity. Engaging in activity to obtain peer acceptance or engaging in activity due to social pressure may lead to feelings of tension and stress, because one is engaging in an activity

DISCUSSION Results from this study indicated that both social and physical environmental factors influenced physical activity indirectly

42

McNeill et al.

to gain approval from others and not for pleasure derived from the activity itself (26). Health care professionals and public health educators commonly recommend engaging in activity for external reasons, such as health improvement and weight loss. These have been identified as salient motives for the adoption of physical activity (41). Studies show that although extrinsic motives are reported as reasons for initiating an exercise program, intrinsic motivation is significantly associated with adherence over time (41). Our findings suggest that such extrinsic motives may undermine self-efficacy, which is a consistent predictor of physical activity participation and as a result should not be promoted as the sole justification for engaging in physical activity. Our findings also provide support for intrinsic motivation as an important correlate of self-efficacy. Possessing intrinsic reasons for engaging in physical activity, such as enjoyment or goal attainment, may increase one’s confidence in the ability to be physically active. Self-efficacy was the strongest direct correlate of activity in this study. It was most strongly associated with vigorous-intensity activity, providing evidence of a positive dose–response gradient based on intensity of activity. This finding was somewhat divergent from other studies, which have shown a greater or equal association with moderate-intensity activity (42,43). Few studies have assessed the strength of the association between self-efficacy and all three forms of physical activity. Our findings show that in fact self-efficacy appears to be most important under challenging circumstances, which in this case is engaging in vigorous-intensity activities. For all forms of activity, availability of facilities was also positively associated with intrinsic reasons for engaging in activity. In contrast, we found that neighborhood quality was inversely associated with intrinsic motivation. Few studies have assessed the influence of neighborhood quality on individual factors. It has been suggested that further exploration of the relationship between physical environmental and individual factors will help to better understand the role of ecological models in explaining physical activity behavior (44). Though these findings support our hypothesis of an indirect association, the direction of the association is puzzling. Measurement issues may in part explain some of this negative relationship. In this study, “neighborhood quality” includes measures of both safety (e.g., traffic and crime) as well as pleasantness of one’s neighborhood environment. Many researchers have measured these two concepts separately, finding associations in the expected directions (15). Conversely, others have found opposite directions than expected between certain physical environmental factors (e.g., poor lighting, safety) and engaging in physical activity (43,44). Physical environmental correlates are often measured with single items and not as latent constructs or scales (45,46). In our study, exploratory and confirmatory factor analysis results indicated that this was a well-measured construct, suggesting the need for further research exploring the relationship between these factors. We found that physical environmental factors were associated with walking, moderate-intensity activity, and vigorous-intensity activity; however, the relationships varied by form

Annals of Behavioral Medicine of activity. Availability of facilities was positively associated with walking. Walking is generally an outdoor activity for which physical environmental influences would be major factors (6). This is consistent with results from Brownson et al. (47), who found that availability of a walking trail increased activity among users of the trail. Neighborhood quality, however, was unrelated to walking but important for both moderate-intensity and vigorous-intensity activity—differing from other studies that found associations between aspects of neighborhood quality (e.g., aesthetics) and walking specifically (48). It is possible that individuals who report walking are doing much of this activity in other places (e.g., work environment), whereas those engaged in moderate- and vigorous-intensity activity do so outdoors. This differential outcome by form of activity is key, supporting researchers who have suggested the need to determine whether environmental influences differ by type of activity (49).

Study Limitations and Strengths Several study limitations warrant attention. First, although the larger study was a longitudinal randomized controlled trial, the study presented here used baseline data only, thereby serving as a cross-sectional study unable to ascertain causal relationships. Second, this proposed model is but one of many potential plausible explanatory models explaining physical activity behavior. Despite the fact that our conceptual model was based on substantive theory and prior research, other plausible models could be tested using these same variables. Third, the physical activity measure did not differentiate between walking intensity, and therefore it is likely that some self-reported walking overlaps with moderate-intensity activity, reflecting walking at higher intensities. Last, this study assessed only some of the important variables of interest in physical activity research. Model misspecification, due to omitted variables (e.g., social norms for physical activity), is possible. Such correlates could have been added to the model to better understand its relationship to activity in the presence of other factors. This study also has several major strengths. The use of social ecological models in understanding physical activity behavior is of importance to social and behavioral science research (12,14). This study is among the first to use SEM to test a conceptual model of physical activity among adults using individual, social, and physical environmental correlates as important explanatory variables. Although this model cannot make causal assertions, it does provide a first step in elucidating the relationship between these kinds of physical activity correlates. In addition, most physical activity models are tested in predominantly female and White populations or among college students; minority groups tend to be underrepresented in population-based studies of physical activity (50). This has important consequences because of low levels of activity among minority groups. This study used a large, multiracial sample, which strengthens support for the study findings across similar diverse populations.

Volume 31, Number 1, 2006 Future studies should continue to assess the relative importance of specific individual, social, and environmental factors as important contributors to physical activity behavior and to better understand the pathways through which they influence physical activity. REFERENCES (1) Allison DB, Fontaine KR, Mason JE: Annual deaths attributable to obesity in the United States. Journal of the American Medical Association. 1999, 270:2207–2212. (2) Macera CA, Jones DA, Yore MM, et al.: Prevalence of physical activity, including lifestyle activities among adults—United States, 2000–2001. Morbidity and Mortality Weekly Report. 2003, 52(32):764–769. (3) Crespo CJ, Smit E, Andersen RE, Carter-Pokras O, Ainsworth BE: Race/ethnicity, social class and their relation to physical inactivity during leisure time: results from the Third National Health and Nutrition Examination Survey, 1988–1994. American Journal of Preventive Medicine. 2000, 18:46–53. (4) Mokdad AH, Ford ES, Bowman BA, et al.: Prevalence of obesity, diabetes, and obesity-related health risk factors, 2001. Journal of the American Medical Association. 2003, 289:76–79. (5) Dishman RK, Sallis JF: Determinants and interventions for physical activity and exercise. In Bouchard C, Shephard RJ, Stephens T (eds), Physical Activity, Fitness, and Health: International Proceedings and Consensus Statement. Champaign, IL: Human Kinetics, 1994, 214–238. (6) Saelens BE, Sallis JF, Frank LD: Environmental correlates of walking and cycling: Findings from the transportation, urban design, and planning literatures. Annals of Behavioral Medicine. 2003, 25:80–91. (7) Eyler AA, Matson-Koffman D, Vest JR, et al.: Environmental, policy, and cultural factors related to physical activity in a diverse sample of women: The Women’s Cardiovascular Health Network Project—Introduction and methodology. Women & Health. 2002, 36(2):1–15. (8) Giles-Corti B, Donovan RJ: The relative influence of individual, social and physical environment determinants of physical activity. Social Science & Medicine. 2002, 54:1793–1812. (9) Giles-Corti B, Donovan RJ: Relative influences of individual, social environmental, and physical environmental correlates of walking. American Journal of Public Health. 2003, 93:1583–1589. (10) Bandura A: Social Foundations of Thought Action. Englewood Cliffs, NJ: Prentice Hall, 1986. (11) McLeroy KR, Bibeau D, Steckler A, Glanz K: An ecological perspective on health promotion programs. Health Education Quarterly. 1988, 15:351–377. (12) Sallis JF, Owen N: Ecological models. In Glanz K, Lewis FM, Rimer BK (eds), Health Behavior and Health Education. San Francisco: Jossey-Bass, 1997, 403–424. (13) Stokols D: Establishing and maintaining healthy environments: Toward a social ecology of health promotion. American Psychologist. 1992, 47:6–22. (14) Sallis JF, Owen N: Physical Activity and Behavioral Medicine. Thousand Oaks, CA: Sage, 1999. (15) De Bourdeaudhuij I, Sallis JF, Saelens BE: Environmental correlates of physical activity in a sample of Belgian adults. American Journal of Health Promotion. 2003, 18:83–92.

Influences on Physical Activity

43

(16) Humpel N, Owen N, Leslie E: Environmental factors associated with adults’ participation in physical activity: A review [see comment]. American Journal of Preventive Medicine. 2002, 22:188–199. (17) Troped PJ, Saunders RP, Pate RR, Reininger B, Addy CL: Correlates of recreational and transportation physical activity among adults in a New England community. Preventive Medicine. 2003, 37:304–310. (18) Kreuter MW, Sugg-Skinner CS, Holt CL, et al.: Cultural tailoring for mammography and fruit and vegetable intake among low-income African-American women in urban public health centers. Preventive Medicine. 2005, 41:53–62. (19) Dzewaltowski DA: A social cognitive theory of older adults exercise motivation. In Ostrow AC (ed), Aging and Motor Behaviour. Indianapolis, IN: Benchmark, 1989, 257–281. (20) Duda JL, Tappe MK: Personal investment in exercise among adults: The examination of age and gender-related differences in motivational orientation. In Ostrow AC (ed), Aging and Motor Behaviour. Indianapolis, IN: Benchmark, 1989, 239–256. (21) Duncan TE, Stoolmiller M: Modeling social and psychological determinants of exercise behaviors via structural equation systems. Research Quarterly for Exercise & Sport. 1993, 64:1–16. (22) McAuley E, Jerome GJ, Elavsky S, Marquez DX, Ramsey SN: Predicting long-term maintenance of physical activity in older adults. Preventive Medicine. 2003, 37:110–118. (23) Sallis J, Owen N: Ecological models. In Glanz K, Lewis FM, Rimer BK (eds), Health Behavior and Health Education. San Francisco: Jossey-Bass, 1997, 403–424. (24) Marcus BH, Selby VC, Niaura RS, Rossi JS: Self-efficacy and the stages of exercise behavior change. Research Quarterly for Exercise & Sport. 1992, 63:60–66. (25) Frederick CA, Ryan RM: Differences in motivation for sport and exercise and their relationships with participation and mental health. Journal of Sport Behavior. 1993, 16:125–145. (26) Vallerand RJ: A hierarchical model of intrinsic and extrinsic motivation in sport and exercise. In Roberts GC (ed), Advances in Motivation in Sports and Exercise. Champaign, IL: Human Kinetics, 2001, 263–320. (27) Sallis JF, Grossman RM, Pinski RB, Patterson TL, Nader PR: The development of scales to measure social support for diet and exercise behaviors. Preventive Medicine. 1987, 16:825–836. (28) Eyler AA, Brownson RC, Donatelle RJ, et al.: Physical activity social support and middle- and older-aged minority women: Results from a US survey. Social Science & Medicine. 1999, 49:781–789. (29) Brownson RC, Eyler AA, King AC, et al.: Reliability of information on physical activity and other chronic disease risk factors among US women aged 40 years or older. American Journal of Epidemiology. 1999, 149:379–391. (30) Centers for Disease Control and Prevention: Behavioral Risk Factor Surveillance System User’s Guide (U.S. Department of Health and Human Services, ed). Atlanta, GA: Centers for Disease Control and Prevention, 1998. (31) Ainsworth BE, Bassett DR Jr, Strath SJ, et al.: Comparison of three methods for measuring the time spent in physical activity. Medicine & Science in Sports & Exercise. 2000, 32(9 Suppl.):S457–S464. (32) Jöreskog KG, Sörbom D: LISREL 8: Structural Equation Modeling With SIMPLIS Command Language. Hillsdale, NJ: Scientific Software International, 1993.

44

McNeill et al.

(33) Tabachnick BG, Fidell LS: Using Multivariate Statistics (2nd Ed.) New York: Harper & Row, 1989. (34) Hoyle RH: Structural Equation Modeling: Concepts, Issues, and Applications. Thousand Oaks, CA: Sage, 1995. (35) Hu L, Bentler PM: Evaluating model fit. In Hoyle RH (ed), Structural Equation Modeling. Thousand Oaks, CA: Sage, 1995, 76–99. (36) Schumacker RE, Lomax RG: A Beginner’s Guide to Structural Equation Modeling. Mahwah, NJ: Lawrence Erlbaum Associates, Inc., 1996. (37) du Toit M, du Toit S: Interactive LISREL: User’s Guide. Lincolnwood, IL: Scientific Software International, 2001. (38) Owen N, Humpel N, Leslie E, Bauman A, Sallis JF: Understanding environmental influences on walking: Review and research agenda. American Journal of Preventive Medicine. 2004, 27:67–76. (39) Rovniak LS, Anderson ES, Winett RA, Stephens RS: Social cognitive determinants of physical activity in young adults: A prospective structural equation analysis. Annals of Behavioral Medicine. 2002, 24:149–156. (40) Duncan TE, McAuley E: Social support and efficacy cognitions in exercise adherence: A latent growth curve analysis. Journal of Behavioral Medicine. 1993, 16:199–218. (41) Ryan RM, Frederick CA, Lepes D, Rubio N, Sheldon K: Intrinsic motivation and exercise adherence. International Journal of Sport Psychology. 1997, 28:335–354. (42) Motl RW, Dishman RK, Ward DS, et al.: Examining social–cognitive determinants of intention and physical activity among Black and White adolescent girls using structural equation modeling. Health Psychology. 2002, 21:459–467.

Annals of Behavioral Medicine (43) Courneya KS, McAuley E: Are there different determinants of the frequency, intensity, and duration of physical activity? Behavioral Medicine. 1994, 20:84–90. (44) Spence J, Lee R: Toward a comprehensive model of physical activity. Psychology of Sport and Exercise. 2003, 4:7–24. (45) Booth ML, Owen N, Bauman A, Clavisi O, Leslie E: Social–cognitive and perceived environment influences associated with physical activity in older Australians. Preventive Medicine. 2000, 31:15–22. (46) Humpel N, Owen N, Leslie E, et al.: Associations of location and perceived environmental attributes with walking in neighborhoods. American Journal of Health Promotion. 2004, 18:239–242. (47) Brownson RC, Housemann RA, Brown DR, et al.: Promoting physical activity in rural communities: Walking trail access, use, and effects. American Journal of Preventive Medicine. 2000, 18:235–241. (48) Humpel N, Owen N, Iverson D, Leslie E, Bauman A: Perceived environment attributes, residential location, and walking for particular purposes. American Journal of Preventive Medicine. 2004, 26:119–125. (49) Owen N, Leslie E, Salmon J, Fotheringham MJ: Environmental determinants of physical activity and sedentary behavior. Exercise & Sport Sciences Reviews. 2000, 28:153–158. (50) King AC, Stokols D, Talen E, Brassington GS, Killingsworth R: Theoretical approaches to the promotion of physical activity: Forging a transdisciplinary paradigm. American Journal of Preventive Medicine. 2002, 23(Suppl. 2):15–25.