|
||||||||
|
Volume 2: No. 4, October 2005
ORIGINAL RESEARCH
|
Take a tour of selected print materials from the Walk Missouri campaign | |
The Get Movin’ St. Joe coalition organizers participated in implementing the Walk Missouri campaign. Local walking resources and activities organized by Get Movin’ St. Joe were incorporated into Walk Missouri newspaper and radio advertisements. For example, the Get Movin’ St. Joe logo was incorporated into the Walk Missouri campaign advertisements. Local collaborators increased visibility of the Walk Missouri campaign by distributing and displaying Walk Missouri campaign posters in community centers, businesses associated with the worksite wellness program, and other locations across the town. In this way, the Walk Missouri media effort helped to advertise community-sponsored walking activities and resources while capitalizing on local efforts to expand Walk Missouri campaign reach.
Phase 4: Impact assessment
The objectives of the Walk Missouri campaign were as follows:
The figure shows the conceptual framework of the evaluation, including the hypothesis that exposure to the Walk Missouri campaign could achieve a direct effect on walking behaviors, indicated by the arrow linking exposure to behaviors. Alternately, there might be an indirect effect of exposure on behaviors, mediated by pro-walking beliefs, indicated by the arrows linking exposure with beliefs and beliefs with behaviors. In testing for either of these effects, we controlled for likely moderating factors.
Figure. Conceptual framework for Walk Missouri media campaign evaluation, St Joseph, Mo, 2003.
The Saint Louis University Institutional Review Board approved this study. A postcampaign-only design was used: phone numbers for residents living within the city of St Joseph were purchased from a market research firm, and a random-digit–dial telephone survey was conducted. Individuals were eligible to participate if they identified themselves as adult (aged 18 years or older) residents of St Joseph. Trained callers conducted the interviews between July 31 and October 31, 2003. The survey required an average of 15 minutes to complete. Individual numbers were dialed numerous times before being eliminated from the survey. A total of 297 interviews were completed with the funds available for evaluation.
Measures
Exposure. As in the evaluation of other media campaigns (31), various campaign exposure measures were used to evaluate the Walk Missouri campaign, including campaign-exposure questions, media-type–exposure questions, and dose-exposure questions. Both prompted and unprompted questions were asked. (The Appendix provides all survey items used in the analysis.) To discern media-type dose exposure, individuals were first asked if they had been exposed to any campaign advertisements through billboards, radio, or newspapers or if they had seen any campaign posters or news stories about the campaign. (News stories were initiated by local newspapers in response to the press conference and the campaign.) Individuals who answered in the affirmative for a media type were then asked how many times they had been exposed to that type. For example, respondents who answered yes for billboards were asked in how many locations they had noticed billboards sponsored by the campaign (with answers ranging from one to six billboards). Respondents who answered yes for radio were asked how many times they had heard radio advertisements sponsored by the campaign (with survey items offering ranges of 1 to 5, 6 to 10, 11 to 20, 21 to 50, 51 to 100, or more than 100 times).
Two variables were developed for analysis of exposure: a four-level dose-exposure scale and a dichotomous variable (exposed and unexposed). The four-level dose-exposure scale summed the five media-type dose-exposure items in the survey. A higher value on this scale signifies either more types of media through which the campaign was seen or heard or a greater number of messages seen or heard through fewer types of media. Because the scale was highly skewed toward no exposure, the scale was recoded as a four-category variable (none, low, medium, and high exposure) to make coefficients more stable. A value of 1 (low) signifies that the respondent reported seeing one billboard, newspaper advertisement, or newspaper story; heard only five or fewer radio advertisements; or saw only five or fewer posters. A value of 2 (medium) signifies that the respondent reported exposure to the campaign through two or three types of media or exposure to more advertisements through one or two types of media. A value of 3 (high) signifies that the respondent reported exposure on four or more types of media or exposure to more advertisements on fewer types of media. Because of varying ranges within survey items and different kinds of exposure to messages on different types of media, it was not practicable to convert the scale into levels of frequency of exposure. The four-level scale was used to assess associations of amount of exposure with study outcomes; it was also recoded into a new dichotomous variable (exposed and unexposed) to test for group differences.
Beliefs. The survey asked participants 12 questions about their opinions of exercise using a 5-point Likert scale. Four subscales were computed from nine survey items to measure theoretical belief constructs. Despite a small number of items in each subscale (only two or three), the Cronbach a calculated for each subscale was near or higher than minimum desired level of a = .70 for social benefits (a = .66), pleasure benefits (a = .58), health benefits (a = .73), and social support (a = .60). (A fifth subscale for overcoming barriers was dropped from the analysis because of an unacceptably low Cronbach a of .46.) Subscales were computed using belief items that were recoded to three levels because the individual items and the resulting scales had normal distributions and would provide more stable results in statistical analyses. Strongly disagree, disagree, and neutral were consolidated as one value coded as 1; agree was coded as 2; and strongly agree was coded as 3. To make them comparable, each subscale was computed to three value scales by dividing the summed scale by the original number of items. To test for a mediating effect of beliefs on the association between exposure and behaviors, a single all-beliefs scale was computed from all 12 belief items in the survey (Cronbach a = .75).
Behavior measures. The survey asked six walking-behavior questions. Three dichotomous (yes or no) measures inquired about walking and wellness activities sponsored by Get Movin’ St. Joe. One dichotomous (yes or no) and two continuous measures of walking behavior were adapted from physical activity measures in the 2000 Behavioral Risk Factor Surveillance System (BRFSS).
Moderators. Various measures were included to control for possible alternative explanations for evaluation results. Demographic measures included sex, age, race, and level of education. Health-related measures included health status, medical diagnosis of chronic disease or overweight, medical advice to walk more, and recent injury. Perceived safety of the participant’s walking environment was assessed using six Likert items. Cronbach a for the scale was .63.
Analysis. Our conceptual framework (Figure) offers the hypothesis that exposed individuals are more likely to hold beliefs consistent with campaign themes and more likely to engage in walking activities than individuals not exposed to the campaign and controlling for likely alternative explanations. In the multivariate analyses, we controlled for moderators that were significantly associated with outcomes. In addition, we hypothesized that beliefs had a mediating effect on the association between exposure and behaviors.
Group differences for the exposed and unexposed portions of the sample were assessed using two-tailed t tests for ordinal and continuous outcomes and chi-square tests for dichotomous outcomes. Associations between amount of exposure and outcomes were assessed using the Spearman rank correlation (ρ). For multivariate analyses, linear regression was used for ordinal and continuous outcomes, and logistic regression was used for dichotomous outcomes. When beliefs were significantly associated with behaviors, stepwise regression was used to test a mediating effect of beliefs on any associations between exposure and behaviors.
During data collection, 4668 phone numbers were used, 2866 of which were out of scope (e.g., businesses, out-of-service numbers, numbers failing to be answered after multiple calls). Of the remaining 1802 numbers, 1461 refused participation before we were able to determine eligibility. Of the remaining 341, five respondents were aged younger than 18 years, bringing our total number of eligible respondents to 336. The total number of completed interviews was 297, resulting in a cooperation rate of 88% (297/336). Compared with the Council of American Survey Research Organization (CASRO) response rates of 54.6% found for the BRFSS Missouri, which include estimates of eligible households among households for which eligibility was not determined, our response rate was low at 17% (32). However, our response rate proved to be better than the rate of 9.1% provided for random-digit–dial surveys tracked by the Market Research Association (33). Low response rates for random-digit–dial surveys are increasingly a problem for evaluators of health promotion interventions (34).
Sample characteristics and exposure levels
The sample had more women and was older, better educated, and more diverse than the U.S. census indicates for this area (Table 4). Thirty-two percent of the sample reported exposure to the campaign through news and advertising media. Exposure levels by type of media ranged from 7% (newspaper advertisements) to 13% (newspaper articles and billboards). Among respondents reporting exposure to the campaign through different types of media, the median number of advertisements reported for each type of media was one to five posters, two newspaper advertisements, two newspaper stories, one billboard, and six to 10 radio advertisements. The exposed respondents (32%) were distributed evenly to low (11%), medium (10%), and high (11%) levels within the dose-exposure scale. A separate analysis found no demographic differences between exposed and unexposed groups, nor was any association found between dose exposure and demographic characteristics.
Beliefs
On a scale of 1 to 3, with 2 indicating agree, survey participants rated all beliefs as approximately 2 (Table 5). The mean for the all-beliefs scale was 4, also equivalent to agree, on the 5-point scale. The exposed group reported greater agreement with two of the four belief subscales (social benefits and pleasure benefits) than the unexposed group at a statistically significant level. Amount of exposure was associated with three of four subscales (social, pleasure, and health benefits) and the all-beliefs scale at a statistically significant level.
Behaviors
The exposed group reported a greater level of participation in three of six wellness or walking behaviors than the unexposed group at a statistically significant level. Amount of exposure was associated with the same three behaviors at a statistically significant level. Two of the outcomes were wellness behaviors: participation in a community-sponsored walk or participation in a health fair. The third outcome was a general walking behavior: the number of days per week the respondent walked at least 10 minutes.
Association of exposures, beliefs, and behaviors
Beliefs and behaviors associated with campaign exposure at the bivariate level were selected for multivariate analysis, controlling for variables associated with the dependent variable in bivariate analyses. Campaign-dose exposure remained associated with two of the four belief subscales (social benefits and pleasure benefits) when controlling for likely confounding factors (Table 6). The association of campaign-dose exposure with the health benefits subscale and all-beliefs subscale was not statistically significant when controlling for other factors.
Campaign-dose exposure remained associated with participation in a community-sponsored walk at a statistically significant level when controlling for educational level. In this analysis, the odds ratio of dose exposure was 2.14 (confidence interval, 1.04–4.41; P = .04); exposed respondents were more than twice as likely to participate in the community-sponsored walks than unexposed respondents. In another multivariate analysis (not shown), dose exposure was not associated with participation in community-sponsored health fairs when controlling for other factors.
Campaign-dose exposure was associated with the number of days per week walking at a statistically significant level when controlling for age and health status (Table 7). However, when the all-beliefs scale was introduced in the second step of the linear regression, the coefficient for campaign exposure lost statistical significance.
Impact assessment of media campaigns seeks to answer the question “Did exposure to the campaign lead to changes in beliefs and behavioral outcomes?” The evidence presented here shows that exposure to the Walk Missouri campaign had limited effects, producing small increases in positive walking beliefs and behaviors among residents of St Joseph. Effect sizes were small, with Spearman ρs of between 0.13 and 0.18 for statistically significant associations of beliefs and behaviors with campaign-dose exposure. Among exposed respondents, 4.3% reported participation in community-sponsored walks, compared with 0.5% of unexposed respondents. Exposed respondents reported walking for at least 10 minutes per day 5.2 days of the week, compared with 4.5 days per week for unexposed respondents.
When an external control community and baseline measures are not possible for assessing the impact of media campaigns, alternative approaches can offer evidence of effects when they show 1) moderate levels of campaign reach; 2) associations of exposure with behaviors, controlling for alternative explanations; 3) and confirmation of an a priori hypothesis positing mediation of exposure–behavior associations by beliefs promoted by the campaign (35). The Walk Missouri evaluation set out to provide such evidence. First, survey respondents reported a moderate level of exposure to the campaign, with about one in three respondents reporting some exposure. This level of exposure was near the 36% average for health-behavior media campaigns found in a recent meta-analysis (36). Second, exposure was significantly associated with three of six walking-behavior measures, and one association remained when controlling for several known predictors: demographic characteristics, health status, perceptions of the walking environment, and health beliefs.
Third, in addition to differences between exposed and unexposed groups, there was a dose-response relationship between exposure and outcomes, with higher levels of agreement on beliefs and positive walking behavior corresponding to higher levels of exposure.
Fourth, the association of exposure and number of days walking was mediated by health beliefs, providing evidence of a theoretically informed causal mechanism. For one wellness behavior — participation in community-sponsored walking activities — beliefs did not mediate the association with exposure. We conclude that campaign information on walking opportunities increased knowledge about these activities, leading to a slight increase in walking.
The single-site, postcampaign-only design limited the power of the study in several ways. Lack of an external control community and baseline measures may have weakened the study’s internal validity. We cannot rule out the possibility of reverse causal direction — that walking adherents were more likely to pay attention to and recall the campaign. Nor can we conclude that the associations we found were not the result of other unmeasured third factors. The low response rate for our random-digit–dial survey may have introduced selectivity bias into the sample and limited our ability to generalize even to the medium-size midwestern town of St Joseph. Self-report measures are vulnerable to socially desirable responses, although there is no evidence that this necessarily contributed to differences between groups.
Acknowledging these limitations and caveats, our study provides information for public health communication researchers and practitioners about the potential for media interventions to promote physical activity. The study elucidates how a media campaign can contribute to a community-sponsored effort to promote walking behavior.
Elements of this single-site, postcampaign-only, cross-sectional design support our claim of limited effect. Careful and multiple measurements of exposure allowed for the creation of an unexposed comparison group. Two measures of exposure confirmed both group differences and dose-response associations of exposure and outcomes. A dose-response relationship suggested a possible causal relationship between exposure to the media campaign and increased likelihood of undertaking walking behaviors (37). We account for likely alternative explanations for associations of exposure with outcomes by including moderating factors in multivariate analysis. Empirical support for a theoretical mechanism is established in a test of the mediating effect of pro-walking beliefs on the association of campaign exposure and walking behavior. Combined, these results strengthen our claim of limited effect by ruling out alternative explanations and supporting an a priori theoretical approach that underlies the campaign strategy and study design (35).
The association of exposure with social and pleasure benefits suggests that the campaign was most successful in communicating these ideas. Although health benefits, social support, and overcoming barriers were also included in the messages, they did not appear to have as much of an impact on the intended audience.
Originally envisioned and designed as a stand-alone media campaign, Walk Missouri was successfully integrated into local community-sponsored activities, consistent with recommendations from the literature (14-16,26), including the Community Guide, which strongly recommends programs that include informational promotional components. The local community coalition welcomed the opportunity to serve as the pilot community for the campaign, indicating that campaign themes and messages complemented their activities. It was not difficult to integrate information about local community activities and the coalition into Walk Missouri message materials, including scheduled walks and other wellness activities, lists of participating sponsors, and the Get Movin’ St. Joe logo. Adaptation of the Walk Missouri campaign materials provided practical, useful information grounded in local events and organizations and proved pertinent for residents of St Joseph.
The scale of the initiative was closer to the community-based initiative of Wheeling, WVa (25), than the national and statewide campaigns implemented in New Zealand and Australia (20,22), all of which included investments in community-based programs. The Walk Missouri campaign incorporated information about community-sponsored activities, but Walk Missouri did not otherwise fund or support local activities. Nonetheless, it can be argued that in actively linking with community-based initiatives, the campaign fit more closely within the category of community-wide activities than in the media-alone category suggested by the Community Guide (16).
The limited effect achieved by the Walk Missouri campaign is similar to the results of other media campaigns promoting general health behaviors as well as the few media campaigns that have promoted walking. The difference in wellness behaviors between exposed and unexposed groups was small (between 3 and 10 percentage points), consistent with research on the effects of health campaigns on behavior (36), and with previous efforts to promote physical activity through media interventions (20,22).
The study provides tentative evidence of independent and complementary effects of media campaigns on community-based interventions. We argue that the Walk Missouri media campaign expanded the reach of the local initiative, Get Movin’ St. Joe. Originating as a worksite-wellness effort and a community coalition, Get Movin’ St. Joe benefited from the visibility provided by the many posters, newspaper advertisements, billboards, and radio spots of Walk Missouri, and as a result, Get Movin’ St. Joe reached a wider audience.
The study design does not allow us to differentiate between effects of media exposure and exposure to other community-sponsored activities; however, the evidence does permit us to discern how the campaign affected walking behavior. The evidence suggests that the campaign achieved behavioral results in two ways. First, campaign exposure was associated with participation in community-sponsored walks, consistent with a direct effect of exposure on behavior. Integration of information about community-sponsored walking activities and resources into media messages rendered them practical and useful. In this way, the media campaign was complementary to the community-sponsored events and may well have boosted attendance.
Second, the association of exposure with number of days of walking was mediated by pro-walking beliefs, consistent with an indirect effect of exposure on behavior. This finding suggests an independent effect of the campaign on walking behavior. General walking behavior did not rely on coalition activities. More importantly, the pro-walking beliefs that mediated the exposure–behavior association were consistent with campaign themes highlighted in Walk Missouri. These results begin to distinguish the mechanisms by which communication elements of a community-wide campaign contribute to increases in physical activity directly by advertising events and indirectly through changed beliefs and attitudes about walking.
For a limited time, residents of St Joseph who were exposed to the campaign may have walked almost 1 day more per week than residents who were not exposed. More people may have attended coalition events because they heard about them through radio spots. Incremental increases in levels of physical activity at the population level contribute to major gains in public health (38).
By accounting for alternative explanations and substantiating a theory-based mechanism for impact, a simple study design with a small sample can provide persuasive evidence of campaign effects. The results of this small study with limited resources provide encouraging evidence that a media campaign can enhance the success of community-based efforts to promote positive walking beliefs and behaviors.
This publication was supported by award number U50/CCU721332-01 from the Centers for Disease Control and Prevention (CDC). The Walk Missouri project was a collaborative effort of the Health Communication Research Laboratory in the Saint Louis University School of Public Health (SLU–SPH) and the Heart Disease and Stroke Program of the Missouri Department of Health and Senior Services (MDHSS), with the support of the Preventive Health & Human Services (PHHS) Catalog of Federal Domestic Assistance 93.99, and the Division of Nutrition and Physical Activity, CDC. We acknowledge our state and county health departments and community collaborators for making the project and this assessment possible: the MDHSS, Buchanan County and Marion County health departments, the St Joseph chapter of the YMCA, and Heartland Health. Atlanta-based CDC colleagues at the Division of Nutrition and Physical Activity offered invaluable guidance for the analysis, and the paper benefited from the review of our SLU–SPH colleagues. The authors are responsible for any errors in analysis or interpretation. The contents of the paper are the sole responsibility of the authors and do not necessarily represent the views of the CDC.
Corresponding Author: Ricardo J. Wray, PhD, Saint Louis University School of Public Health, 3545 Lafayette Ave, St Louis, MO 63104. Telephone: 314-977-4075. E-mail: wray@slu.edu.
Author Affiliations: Keri Jupka, MPH, Health Communication Research Laboratory, Saint Louis University School of Public Health, St Louis, Mo; Cathy Ludwig-Bell, PhD, Southern Illinois University Edwardsville, Edwardsville, Ill.
|
Table 5. Telephone Survey Results by Level of Exposure to Walk Missouri Media Campaign, St Joseph, Mo, 2003
|
|
|
Table 6. Linear Regression Analysis for Belief Subscales, Walk Missouri Campaign, St Joseph, Mo, 2003
Table 7. Stepwise Linear Regression for Number of Days Walked per Week, Walk Missouri Campaign, St Joseph, Mo, 2003a
|
|
|
Appendix: Survey Questions for the Walk Missouri Media CampaignExposure questions1a. In the past 30 days, do you remember seeing any posters sponsored by the Walk Missouri campaign displayed in the St Joe area?
1b. In the last 30 days, in how many different locations have you noticed a poster sponsored by the Walk Missouri campaign in your community? (Read the ranges if respondent has trouble answering.)
2a. In the last 30 days, do you remember seeing any newspaper advertisements sponsored by the Walk Missouri campaign?
2b. In the last 30 days, how many times have you seen Walk Missouri advertisements in your local newspapers?
3a. In the last 30 days, do you remember seeing any local news stories about Walk Missouri or Get Movin’ St. Joe?
3b. How many news articles have your seen or read in the last 30 days that mention Walk Missouri or Get Movin’ St. Joe?
4a. In the last 30 days, do you remember hearing radio ads sponsored by Walk Missouri?
4b. In the last 30 days, how many times have you heard radio ads sponsored by Walk Missouri?
5a. In the last 30 days, do you remember seeing any billboard advertisements sponsored by Walk Missouri?
5b. In the last 30 days, in how many different locations have you noticed a billboard sponsored by the Walk Missouri campaign in your community?
Belief questionsI would now like to ask you a few questions about your opinions of exercise. Please indicate whether you
Behavior questions
Moderator questions: demographics, health status, and walking environment
|
|
|
|
|
|
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Privacy Policy | Accessibility This page last reviewed March 30, 2012
|
|