Abstract
The purpose of this systematic review is to analyze current lifestyle intervention literature conducted in U.S. rural areas to identify the most effective and impactful interventions on physical activity outcomes. Quality of studies was assessed using the Cochrane Collaboration’s risk of bias tool. Exploratory calculations of effect size and 95% confidence intervals were performed to demonstrate trends in clinical importance. Eight trials which included 1,399 adult participants met the inclusion criteria for review. Two trials reported a significant difference in the increase of physical activity between groups with medium to large effect sizes. Interventions which are very personalized or tailored and/or include many intervention contacts appear to be most effective. However, the small number of studies, mixed findings, and the risk of bias limit our ability to draw conclusion.
Rural areas are home to approximately 60 million people, or 20% of the U.S. population (U.S. Census Bureau, 2015). Compared with their urban counterparts, rural residents experience higher rates of chronic diseases and a higher prevalence of all-cause mortality (Meit et al., 2014). Differences in obesity rates are likely a major contributor to these geographic health disparities. Based on the 2011 National Health Interview Survey, self-reported obesity varied significantly by urbanization level with rates of obesity increasing as levels of rurality increased (Meit et al., 2014). Furthermore, when examining objectively measured body mass index, the obesity prevalence was 40% among rural adults and 33% among urban adults, and the disparity in these obesity rates remained significant after adjusting for demographics, diet, and physical activity (Befort, Nazir, & Perri, 2012).
Insufficient physical activity is one of the major factors contributing to obesity and other chronic diseases (Physical Activity Guidelines Advisory Committee [PAGAC], 2008). Few Americans meet national physical activity recommendations, with rural adults at particular risk (Fan, Wen, & Kowaleski-Jones, 2015). Rural residents accumulate more light intensity activity but less moderate-to-vigorous physical activity when compared with urban residents (Fan, Wen, & Kowaleski-Jones, 2014).
Previous reviews of lifestyle interventions have found mixed levels of effectiveness. One review was unable to determine specific lifestyle intervention elements that contributed to intervention effectiveness, although the authors did report that effectiveness was not associated with the number of intervention components, intensity, or duration of an intervention (Blue & Black, 2005). A second review concluded that well-controlled interventions targeting multiple health behaviors are effective in the short term; however, maintenance of behavior change post-intervention has remained elusive (Marcus et al., 2006). Another systematic review of lifestyle interventions among general adult populations concluded that group-based interventions are effective and reduce cost of intervention delivery (National Institute for Health and Clinical Excellence, 2012). However, recommendations such as this may not be suitable for rural populations.
Reduced levels of physical activity among rural Americans may be due to unique barriers. For example, urban communities typically have greater access to exercise classes and physical fitness facilities when compared with rural communities (Cleland, Ball, King, & Crawford, 2012). In addition, rural areas tend to have fewer health care resources than urban areas which makes it even more critical that health behavior interventions are both efficient and effective (Jones, Parker, & Ahearn, 2009).
Furthermore, rural Americans are much less likely to report access to safe streets or neighborhood parks to be active in (Parks, Housemann, & Brownson, 2003). Also, rural women describe important personal and socio-economic barriers to physical activity, including child care, family needs, and work demands (Olsen, 2013). This review of determinants highlight important intervention points specific to rural Americans. However, to our knowledge, no systematic review has explored this issue. The purpose of this systematic review is to analyze current lifestyle intervention literature conducted in U.S. rural settings to identify the most effective and impactful interventions on physical activity outcomes. This review includes both an assessment of the quality of the studies and effect size (ES) analysis.
Method
Search Strategy
We conducted a review of studies published from January 1990 to June 2015. The following databases were searched: PubMed, CINAHL, PsycINFO, SportDiscus, and Cochrane Library. Keywords used in the search included physical activity OR exercise OR walking AND rural. The search term intervention was not included in the search strategy to avoid missing relevant articles. The search was limited to English in all databases and adult populations in PubMed, CINAHL, and PsycINFO. In addition, ancestry search was conducted by reviewing relevant sources cited in identified publications. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines provided the structure for this systematic review (Moher, Liberati, Tetzlaff, & Altman, 2009).
Trials were selected for review if they (a) were randomized or quasi-randomized controlled trials, (b) recruited adults (≥18 years of age), (c) conducted interventions in rural areas in the United States, (d) measured physical activity or exercise as an outcome (either self-report or objective measures), (e) reported adequate data for ES calculation (i.e., mean, measure of variability, and sample size), and (f) had a minimum of five participants per group. For the purpose of this review, the term rural was not strictly defined, rather studies were included if they referred to the setting or sample as rural. Physical activity or exercise in this review was defined according to the 2008 U.S. Physical Activity Guidelines which describes physical activity as bodily movement that enhances health (PAGAC, 2008). The control group in randomized controlled trials (RCT) is defined as the group that does not receive experimental treatment or receives an alternative intervention, and whose performance provides a baseline against which the effects of the treatment can be measured.
The initial computerized search yielded 652 citations (see Figure 1). Both authors independently reviewed the studies to assess fit with the inclusion criteria. After removing duplicates and screening titles and abstracts, 252 potentially relevant studies were identified for evaluation. Following full review of these articles, 244 were excluded. The reasons for exclusion were primarily because studies (a) were not intervention studies, (b) focused on youth, (c) were not conducted in the United States, (d) were focused on urban populations, (e) did not measure physical activity as an outcome, or (f) lacked data needed for ES calculation. One additional study was identified through ancestry searches on all potential primary studies. A total of eight studies were included in this review (see Figure 1).
Flowchart of selection of studies.
Appraisal Strategy
Each study was assessed independently by two reviews using the Cochrane Collaboration Risk of Bias tool for potential bias in domains of sequence generation (i.e., randomization and recruitment), allocation concealment (i.e., attempts to conceal allocation of intervention assignment and methods for generation of the sequence of allocations), blinding of participants, personnel and outcome assessors, incomplete data addressed, selective outcome reporting, and other bias. Risk of bias in each domain was assessed as high, low or unclear (see Figure 2; Higgins & Green, 2011). High risk of bias are interpreted as plausible bias that seriously weakens confidence in the results; unclear risk of bias are plausible bias that raises some about the results; and low risk of bias are plausible bias unlikely to seriously alter the results. While the critical appraisal of included studies is commonly used in the context of systematic review methods, assessing risk of bias squarely targets the extent to which results of included studies could be flawed.
Risk of bias summary.
Analysis Strategy
Identified studies were examined by selected variables (if available), using the following data collection categories: (a) design and sample: publication year, U.S. region, study design; (b) intervention characteristics: theoretical framework, setting, content/delivery mode, duration; (c) sample characteristics; (d) physical activity measures and time points; and (e) key physical activity findings (see Table 1). Descriptive analyses identified major themes. ES with 95% confidence intervals was calculated to demonstrate trends in clinical importance by using an online calculator (Coe, 2000). For reporting consistency, we calculate the value of Cohen’s d in each study for each PA outcome, using the means (
Study Characteristics of Included Studies.
Note. Unless otherwise stated, all reported findings are significant at p < .05 level. PA = physical activity; RCT = randomized control trial; SCT = social cognitive theory; SOC = stage of change transtheoretical framework; HBM = Health Belief Model; SSM = social support models; TX = intervention group; LHA = lay health advisors; N = number who completed the study; PAR = physical activity recall; MET = metabolic equivalent; Wk = week; FU = follow-up; TTM = transtheoritical model; TRA = the Theory of Reasoned Action; CCM = chronic care model; IPAQ = International Physical Activity Questionnaire; MPA = moderate physical activity; VPA = vigorous physical activity; MVPA = moderate or vigorous physical activity; EE = energy expenditure; AHA = American Heart Association; GLTEQ = Godin Leisure Time Exercise Questionnaire; BMI = body mass index; HPM = health promotion model.
Cohen (1988) defined ESs as “small: d = 0.2,” “medium: d = 0.5,” and “large: d = 0.8.”
Results
Populations Studied
Five of the eight trials were conducted in the Midwest, one trial was conducted in the West, and the remaining two in the South. The eight trials represented a total 1,399 adult participants with mean ages from 35 to 58 years. Sample sizes ranged from 32 to 587. Four of the trials targeted rural women; on average, only 11% of participants across all studies were male. A majority of participants were Caucasian. Three trials included participants who had underlying health conditions: overweight (Carr et al., 2008; Zoellner et al., 2013) or pre-hypertension (Hageman, Pullen, Hertzog, & Boeckner, 2014).
Study Design and Intervention Attributes
Six of the eight trials were guided by social and behavioral theories, including the Social Cognitive Theory (SCT; Campbell et al., 2004; Carr et al., 2008; Ely et al., 2008; Folta et al., 2009), Transtheoretical Model/Stage of Change (TTM/SOC; Campbell et al., 2004; Carr et al., 2008), Health Belief Model (HBM; Campbell et al., 2004; Ely et al., 2008; Hageman et al., 2014; Walker et al., 2009; Walker et al., 2010), Social Support Models (SSM; Campbell et al., 2004), and the Theory of Reasoned Action (TRA; Ely et al., 2008). In addition, three of these seven trials integrated multiple theories into intervention structures (Campbell et al., 2004; Carr et al., 2008; Ely et al., 2008; see Table 1).
Two of the eight trials were solely focused on increasing physical activity among rural adults (Carr et al., 2008; Peterson, Yates, Atwood, & Hertzog, 2005). In the remaining six trials, physical activity promotion was included as part of cancer prevention (Campbell et al., 2004), obesity management (Ely et al., 2008; Zoellner et al., 2013), heart health promotion (Folta et al., 2009), and overall lifestyle improvement (Hageman et al., 2014; Walker et al., 2009; Walker et al., 2010). All trials had multiple intervention contacts with participants (median = 14; range = 8-24 contacts). Intervention delivery methods included face-to-face (n = 5 studies), telephone (n = 1 trial), combination of email and telephone (n = 1 trial), and combination of email and print mail (n = 1 trial). The duration of interventions ranged from 3 to 12 months.
Only two of the eight trials measured objective physical activity outcomes (i.e., average daily steps monitored by pedometers; Carr et al., 2008; Folta et al., 2009), while the remaining trials utilized self-reported physical activity outcomes. Subjective physical activity measures included general physical activity level (Campbell et al., 2004), moderate and/or vigorous physical activity (Ely et al., 2008; Hageman et al., 2014; Peterson et al., 2005; Walker et al., 2009; Walker et al., 2010), walking duration (Ely et al., 2008), calculation of energy expenditure based on self-reported physical activity (Hageman et al., 2014; Peterson et al., 2005; Walker et al., 2009; Walker et al., 2010), and stretching and strengthening exercise (Walker et al., 2009; Walker et al., 2010).
Risk of Bias Assessment
The results of the Cochrane Collaboration Risk of Bias assessment suggest that the included studies had high or potential risk of bias in the domains of sequence generation (i.e., randomization and recruitment), allocation, blinding, and other bias (see Figure 2). Only two studies have low risk in six of the eight assessment domains (Hageman et al., 2014; Zoellner et al., 2013). Five studies have two low risk domains, and the remaining one has only one low risk domain (Peterson et al., 2005). Simple randomization or randomizing by church or community was common in included studies (Campbell et al., 2004; Ely et al.,2008; Folta et al., 2009; Peterson et al., 2005; Walker et al., 2009; Walker et al., 2010). Carr et al. (2008) did not report the method of randomization. Most of the studies did not mention blinding participants or investigators, except one study by Hageman and colleagues (2014). However, on the topic of this review, it is typically impractical or impossible to blind participants or study personnel to intervention group (Higgins & Green, 2011) and therefore inappropriate to describe all such studies as of “low quality” by using quality assessment scales or checklist. In addition, blinding of outcome assessor staff is not considered a critical issue when all data are self-report, as it was in most of these studies (Higgins & Green, 2011). Most studies mentioned the existence of missing data and the analysis method addressing the missing data, with the exception of Peterson and colleagues (2005). Most studies reported dropout rates except one (Folta et al., 2009). All studies reported other potential bias. A major source of potential bias comes from small sample sizes or a small number of clusters (Campbell et al., 2004; Carr et al., 2008; Ely et al., 2008; Folta et al., 2009; Peterson et al., 2005).
Efficacy of Interventions for Increasing Physical Activity
Two of the eight trials reported a significant increase in physical activity between intervention and control groups (Campbell et al., 2004; Folta et al., 2009). These two trials targeted multiple health behaviors (i.e., physical activity, diet, and/or cancer screening). In Campbell et al.’s (2004) study (four-arm, N = 587), intervention strategies included (a) mail-delivered individualized newsletters and videos on health behavior change, (b) education sessions delivered by lay health advisors (LHA) on health behavior change, and (c) the combination of the intervention (a) and (b). Frequency and duration of moderate to vigorous exercise were measured by a modified questionnaire during the 9-month intervention and at 1-year follow-up. Physical activity in terms of metabolic equivalent task (MET) hours per week were calculated. Their study found significantly greater moderate to vigorous exercise levels in the mail-based group with a medium ES (p < .05, ES = 0.32). However, no significant findings were demonstrated in the LHA group (ES = 0.28) or the combination group (ES = 0.16).
In Folta et al.’s (2009) study (two-arm, n = 96), the intervention group was engaged in physical activity classes twice a week for 3 months. Class components included either dancing or walking outside. Findings showed a significant increase in average daily steps post-intervention monitored by pedometers in the intervention group with a large ES (p < .01, ES = 1.04). The control group was not exposed to the exercise classes.
Six of the eight trials did not yield significant differences between the groups with small to medium ESs or negative ESs (see Figure 3; Carr et al., 2008; Ely et al., 2008; Hageman et al., 2014; Walker et al., 2009; Walker et al., 2010; Zoellner et al., 2013). The main physical activity measure in these six trials was self-reported weekly duration of moderate to vigorous physical activity. Four of the six trials were multiple behavior change interventions (i.e., obesity prevention or management and healthy activity and eating promotion) for 4 months (Zoellner et al., 2013), 6 months (Ely et al., 2008), or 12 months (Hageman et al., 2014; Walker et al., 2009; Walker et al., 2010). The trial by Carr et al. (2008) was solely conducted for physical activity promotion and focused on obese adults (two-arm; n = 32). The intervention group reported a significant increase in daily steps at 4 months (p < .05); however, this increase was not significantly greater than that observed in the control group. The remaining trial by Peterson and colleagues (2005) aimed to increase physical activity among rural women by enhancing social support through an interactive booklet and weekly face-to-face contacts for 3 months.
Effect sizes and 95% confidence intervals for outcome differences between intervention and control groups.
Discussion
To our knowledge, this is the first systematic review to examine the physical activity outcomes of lifestyle interventions targeting rural adults. Overall, the evidence base for physical activity promotion in rural areas is weak. Of the eight studies that met the inclusion criteria, no consistent outcomes of physical activity measures were reported. Significant increases in physical activity levels were found in interventions targeting multiple health behaviors with medium to high ESs (Campbell et al., 2004; Folta et al., 2009). In those trials with no significant findings, there were also some trends in physical activity increase in intervention groups with moderate ESs (Hageman et al., 2014; Peterson et al., 2005; Carr et al., 2008; Walker et al., 2009; Zoellner et al., 2013).
In this review, the majority of participants were young or midlife rural females. The inclusion of males or older adults was lacking in the included trials. Most trials were guided by social and health behavior theories. In addition, it was common that more than one theory was applied in interventions. Intervention delivery modes included traditional face-to-face contacts and mail-, telephone-, or Internet-based contacts. Physical activity promotion strategies mainly included (a) enhancing social support and motivation, (b) engaging in group-based exercise classes and education, and (c) mail- or Internet-based activity promotion materials. Only one third of the trials had relatively long study duration (9-12 months) and had follow-up measurements post-intervention.
The two trials that demonstrated efficacy targeted multiple health behaviors and included distinct components (Campbell et al., 2004; Folta et al., 2009). Folta and colleagues’ (2009) trial was the second most intense intervention which included 24 1-hr long group sessions. Campbell’s (2004) trial included intervention material that was very tailored based on focus group responses from the target audience. This intervention included four personalized computer-tailored newsletters and four targeted videotapes. Both of these trials were theory-based and included sedentary adults with no other reported chronic diseases besides overweight and/or obesity. Findings of this review are consistent with previous reviews which indicated that interventions targeting multiple health behaviors are effective for short-term behavior change (Marcus et al., 2006) and that group-based interventions appear to be an efficient and effective intervention delivery mode (National Institute for Health and Clinical Excellence, 2012).
The findings in this review should be interpreted with caution in consideration of the risk of bias. A major source of potential bias comes from sampling and randomization. Simple randomization by church or community with no minimization (i.e., a statistical method to minimize the imbalance between sample sizes of groups) rather than randomizing participants to intervention groups could affect the study findings as well. In addition, no studies reported blinding research personnel, and this could potentially be a source of bias in study outcomes. However, the self-report nature of most of the data could minimize this impact of bias but introduce participant self-report bias.
In trials with negative findings, there were many other potential biases that might moderate the intervention effects, such as imbalanced baseline characteristics between intervention and control groups, high level of baseline health behavior control (Ely et al., 2008), cross-intervention contamination (Walker et al., 2009; Walker et al., 2010), high attrition rate (Zoellner et al., 2013), and relatively short study duration. In addition, the measurement of physical activity by self-report might be biased by social desirability of response. For example, sedentary individuals in Hageman et al.’s study recalled a mean of more than 250 min of moderate intensity or greater activity per week at baseline, which is much higher than the national physical activity recommendation (150 min per week; PAGAC, 2008).
The generalizability of this reviews findings is limited. Generalizability is limited by study design of the selected studies with only half of the studies using a randomized controlled design. Furthermore, participants were not representative of rural populations overall. For example, some studies involved participants who were more educated with higher income (Folta et al., 2009). There was also inadequate evidence in minority population. In addition, most individuals included in this review were motivated to change health behavior. Such volunteer selection bias could also limit the generalization of the findings.
Inherent to review papers, there are limitations worth noting. This systematic review is limited by database search strategies. We used rural as a main keyword to identify potential studies. However, we may not have detected those studies that were truly conducted in rural areas but did not specifically state the setting of the study, although ancestry searches were conducted to complement our search. Environmental influences could be the barriers for the activities of rural population, such as the availability of safe neighborhoods, walking trails, and local activity resources or programs. However, this systematic review only focused on studies that had some interaction between investigators and participants, rather than community-based environmental improvement for physical activity promotion.
Despite the mixed findings, this review provides a valuable view of present physical activity promotion strategies in U.S. rural areas, as well as study features that may benefit future research. Physical activity promotion alone or integrated into multiple behavior change programs could be effective. More rigorously designed studies are warranted to reveal successful promotion strategies. Future research could be of more value if investigators provide careful control over sampling (e.g., baseline conditions, sample size), randomization (e.g., randomized or simple randomization by sites with minimization), intervention fidelity (e.g., compliance, cross-intervention contamination), and attrition rates.
Many different intervention delivery modes have been used in previous studies (Marcus et al., 2006). Face-to-face contact may still be a good initial intervention mode and may be followed by other more convenient and less expensive delivery modes, such as mail-, telephone-, and/or Internet-based contacts. In the included studies, smartphone-based interaction programs were not studied. Previous studies had found high acceptance of smartphone-based health promotion for rural communities (Heckman & Carlson, 2007; Kurti, Logan, Manini, & Dallery, 2015). Given that rural population live in broad geographic areas, non-face-to-face intervention contacts might increase intervention retention rate, decrease intervention expenditure, and be able to reach rural people from larger regions. However, face-to-face contact may be necessary initially to help support and motivate behavior change.
The reviewed studies mostly included younger or mid-life rural female participants. To increase the representativeness of intervention findings, future studies should target the least active subgroups of rural residents, such as older females. Recruitment strategies could be facilitated through existing county extension programs. Furthermore, as discussed above, self-reported physical activity level could be unreliable. Objective measurement of physical activity using accelerometers or pedometers is encouraged in future studies that could enhance the measurement about insufficiently active rural individuals. To the best of our knowledge, there is no cost-effectiveness data on physical activity promotion in rural areas. Future studies with cost-effectiveness data are warranted.
In conclusion, the current available evidence suggests that some physical activity interventions and multiple health behavior interventions may be effective in increasing physical activity levels in rural adults. Based on findings from the included studies, face-to-face contact may be needed to support initial behavior change. In addition, self-monitoring through the use of pedometers appears to be an important factor in intervention success. However, the evidence is preliminary and it is still unclear which intervention strategy and delivery mode may be most effective in rural population. More rigorously designed studies are need in future research.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.