POWER for Reproductive Health: Results from a Social Marketing Campaign Promoting Female and Male Condoms

Journal of Adolescent Health 43 (2008) 71–78

Original article

POWER for Reproductive Health: Results from a Social Marketing Campaign Promoting Female and Male Condoms Sheana S. Bull, Ph.D.a,b,*, Samuel F. Posner, Ph.D.c, Charlene Ortiza, Brenda Beaty, M.S.P.H.a, Kathryn Benton, M.S.P.H.a, Lillian Lin, Ph.D.c, Sherri L. Pals, Ph.D.c, and Tom Evansd a

Colorado Health Outcomes Program, University of Colorado Health Sciences Center, Aurora, Colorado b Department of Family Medicine, University of Colorado Health Sciences Center, Aurora, Colorado c Centers for Disease Control and Prevention, Atlanta, Georgia d Educational Message Services, Ventura, California Manuscript received April 4, 2007; manuscript accepted December 21, 2007

Abstract

Purpose: To evaluate effects of a 6-month social marketing campaign on awareness of, attitudes toward and use of female as well as male condoms for 15–25 year-old-women. Methods: Using a time-space sampling methodology, we conducted a cross-sectional survey of 3407 women at pre-campaign in 12 western U.S. neighborhoods on female and male condom awareness, attitudes, and use. Six of the 12 study neighborhoods were randomly selected to receive the POWER social marketing campaign designed to impact condom knowledge, attitudes, and use. The campaign was followed with another cross-sectional survey of 3,003 women in all 12 study neighborhoods on condom knowledge, attitudes, use and awareness of POWER materials. We compared pre-and post-campaign surveys to determine the efficacy of POWER and conducted post hoc analyses on post-campaign data to determine if exposure to POWER was related to higher levels of positive condom attitudes and norms and condom use. Results: We found no differences between neighborhoods with and without the POWER campaign with regard to our primary outcomes. To diagnose reasons for this null effect, we examined outcomes post hoc examining the influence of POWER exposure. Post hoc analyses show some evidence that exposure to POWER was associated with condom use. In the context of the nested trial, this raises concerns that post test only evaluations are limited. Conclusions: Establishing the efficacy of a social marketing campaign is challenging. This group randomized trial showed a null effect. Social marketing campaigns may need to have more media channels and saturation before they can show behavioral effects. Using a nested design with randomization at the community level and probability sampling introduces rigor not commonly seen in evaluations of social marketing campaigns. © 2008 Society for Adolescent Medicine. All rights reserved.

Keywords:

Female condoms; STD prevention; Pregnancy prevention

We describe here evaluation results of the POWER social marketing campaign (Prevention Options for Women Equals Rights), funded by The Association of Teachers of Preventive Medicine and the Centers for Disease Control

*Address correspondence to: Sheana S. Bull, Ph.D., M.P.H., Colorado Health Outcomes Program, University of Colorado Health Sciences Center, PO Box 6508, Mail Stop F-443, Aurora, CO 80045-0508. E-mail address: [email protected]

and Prevention (CDC). POWER was designed to increase awareness of female condoms, to change attitudes toward and intentions to use male or female condoms and to increase the use of both kinds of condoms among AfricanAmerican and English-speaking Latina women 15–25 years of age in neighborhoods in California and Nevada with high rates of sexually transmitted infections and teen births. Social marketing campaigns apply marketing principles to address health or social concerns and persuade people to

1054-139X/08/$ – see front matter © 2008 Society for Adolescent Medicine. All rights reserved. doi:10.1016/j.jadohealth.2007.12.009

72

S.S. Bull et al. / Journal of Adolescent Health 43 (2008) 71–78

adopt healthy behaviors through the use of strategies used to market other goods in our society, e.g. food, cars, music, entertainment [1–3]. The application of social marketing techniques combines price, placement, and promotion of a product with the intention of changing behaviors [4,5]. A 10-year review of such campaigns reveals that narrowly focused campaigns using principles of design such as identifying communication channels, extensive pretesting for message accuracy, acceptability, and credibility, combined with an assessment of target audience knowledge and attitudes, have shown effects in changing behaviors [2]. Campaigns have demonstrated effects in increasing awareness about condoms and changing attitudes toward them [6], and have shown that increasing the presence of condoms in the environment indirectly affects individual attitudes about condoms and increases use by implying that condoms are acceptable, thereby influencing social norms [4,7]. Researchers have shown social marketing campaigns to be effective for increasing condom use in multiple settings [2,7–10]. POWER (Prevention Options for Women Equals Rights) is a social marketing campaign funded by The Association of Teachers of Preventive Medicine and the Centers for Disease Control and Prevention (CDC). POWER was designed in an extensive formative process, following principles of effective campaign design [2,3] and pilot tested in

Denver Colorado in 1998 –1999. A theoretical framework to affect attitudes, knowledge, and beliefs about female as well as male condoms among the target audience using social marketing principles guided our campaign. POWER was developed in an iterative formative process testing message concepts and marketing plans in 12 focus groups with 89 African-American and Latina females 15–25 years of age. We used print media and gift incentives in the campaign in lieu of the more expensive radio and television ads. Further details about campaign development are presented elsewhere [11]. We used an extensive enumeration process to identify communication channels in community venues where women from the target audience congregate. Once identified, many of these places were selected as campaign sites. POWER included posters and take-away information cards placed in bathroom stalls and on bulletin boards about condom efficacy and use, as well as local rates of sexually transmitted disease (STD) and birth among 15–25-yearolds. Take-away cards contained group-specific information and graphics about female and male condom efficacy, instructions on use, and rates of STD and birth among 15– 25-year-olds in each campaign neighborhood. The takeaway card included a coupon redeemable at one of three to five sites in each campaign neighborhood for a silk carrying case containing two female and three male condoms, lubricant, and instructions for use (Figure 1). We monitored

Figure 1. Selected POWER campaign materials. (Upper) Sample study poster. (Lower) Front of unfolded take-away card.

S.S. Bull et al. / Journal of Adolescent Health 43 (2008) 71–78

uptake of take-away cards on a weekly basis throughout the 6-month campaign by visiting each campaign site, and moved to new venues and new locations within venues as warranted to increase uptake of materials. Methods The evaluation of POWER was completed using a randomized controlled trial, with the neighborhood as the unit of analysis, and we sampled our population using a time– space sampling method. We chose this design because we could not practically limit campaign exposure at the individual level and because it has the added rigor of controlling for correlations within social groups of commonly shared characteristics, improving on the issue of inflating the type I error rate [12]. This design also improves upon concerns with evaluations of other social marketing campaigns relying only on post-campaign assessments [2,13]. We completed a baseline survey in each of our study neighborhoods to ascertain knowledge of, attitudes toward, and use of female as well as male condoms before conducting the campaign. These neighborhoods and regions are identified in Table 1. We chose these sites because their demographics and rates of STD and teen births were similar to those in Colorado, where POWER was developed and pilot tested. We selected cities, zip codes, and census tracts with the highest rates of chlamydia, gonorrhea, and teen births for 15–25-year-old women as our targeted campaign area. We chose the age range of 15–25 years because it represents as the age range of highest susceptibility to STD for young women [14]. Using the nested design, we anticipated needing data from women in twelve communities to obtain adequate power to detect differences across campaign and comparison neighborhoods. Without previously published estimates of intraclass correlation coefficients for sexual risk behaviors, we conservatively anticipated changes in female or Table 1 POWER study neighborhoods in California and Nevada Neighborhood

Area

East Oakland West Oakland SF-Mission SF-Lakeview Inglewood East Los Angeles East Long Beach North Long Beach Cambridge North Las Vegas Oceanside San Diego

Oakland/SF Bay Area, Oakland/SF Bay Area, Oakland/SF Bay Area, Oakland/SF Bay Area, Los Angeles, CA Los Angeles, CA Los Angeles, CA Los Angeles, CA Las Vegas, NV Las Vegas, NV San Diego, CA San Diego, CA

Condition (assigned randomly after pre-campaign assessment) CA CA CA CA

Comparison Comparison Campaign Campaign Campaign Campaign Comparison Comparison Campaign Comparison Campaign Comparison

73

male condom use based on data from our Denver pilot study and assumed an intraclass correlation coefficient of .02 and sample size of 300 per neighborhood. At follow-up we reduced recruitment goals to 250 per neighborhood to facilitate more rapid accrual of the sample within a time period sufficient to detect campaign effects. Reducing sample size from 300 women per neighborhood to 250 per neighborhood has a negligible effect on power, because the number of women per neighborhood contributes to the total sample size but also to the variance inflation factor in a group-randomized trial. More data on the intraclass correlation coefficients is available elsewhere [15]. Sample selection We used a time–space sampling approach to identify and recruit women to complete pre-campaign and post-campaign assessments [16]. Our selection of venues for sampling women for our surveys also served as the mechanism for selecting our communication channels for the campaign. This approach assisted us in identifying community locations to maximize exposure to POWER materials. We visited community venues identified as popular among 15–25-year-old women by community leaders to complete a visual count of women who appeared to be in our target audience entering a predetermined area for 30 minutes and used this to estimate the number of persons from our target audience we could expect at the venue during a 2.5-hour period. Venue owners and managers identified the most popular days of the week or times of day in each venue to establish best days of the week and times of day (venue day time increments [VDTs]) for recruitment. This became our sampling frame. Survey recruitment At each VDT, one coordinator would “click” women entering a predetermined area that he or she thought might be eligible for the study based on observable characteristics. The coordinator would then signal a recruiter to approach women clicked to screen them for eligibility. Recruiters approached as many women clicked as they could, but the volume of women clicked would not always permit us to approach 100%. Those screened were asked their age (eligible: age 15–25 years); zip code (eligible: residence in one of multiple zip codes per neighborhood); and language (eligible: English speaking). Those eligible were invited to complete the interviewer administered informed consent and Health Insurance Portability and Accountability Act (HIPAA) authorization and then the self-administered survey. Women were offered a $10 coupon to a local store for participation. By using this approach, we recruited women from a probability sample of venues. This is an improvement over traditional convenience sampling that allows more reliable statistical inferences about our findings to the population of

74

S.S. Bull et al. / Journal of Adolescent Health 43 (2008) 71–78

women who attend the type of venues in our sampling frame. We were also able to use the most frequented venues as campaign sites, increasing our confidence that these would be strong communication channels for our low-budget intervention. Following the pre-campaign survey, study neighborhoods were stratified within regions included in the study, to ensure adequate comparability between campaign and comparison neighborhoods. Stratification before randomization can help equate groups before the start of an intervention where simple randomization may not. The six campaign neighborhoods (Table 1) were then selected at random using a computer-generated program. Between September 2004 and March of 2005, we implemented a social marketing campaign in our six randomly selected campaign neighborhoods. The campaign included 561 displays with small take-away information cards in 400 sites across the six communities. Intervention neighborhoods had between 51 and 125 displays across campaign venues and 46,602 takeaway cards were distributed during the campaign, with 1,641 (3.5% of the card distributed) of these cards redeemed for purses with condoms inside. The post-campaign surveys were implemented immediately following the campaign period, between April and July 2005. The study protocol for all study phases was reviewed and approved in 2003 by the Colorado Multiple Institution Review Board and the Institutional Review Boards at the Centers for Disease Control and Prevention, with a waiver of parental consent for participants ⬍18 years of age. Procedures Data were collected with a self-administered, structured, confidential survey, which took 20 minutes to complete. The survey covered modules on male and female condom awareness as well as modules on attitudes toward condom use that covered both male and female condoms. Only women who had heard of female condoms were asked to answer the questions related to female condoms. Measures Background characteristics reported on here include age, race and ethnicity. We measured positive and negative attitudes toward condoms, condom use norms, and intentions to use condoms. These measures were developed and used in the POWER pilot. They are based on measures developed and used in multiple other studies, and were shown to have strong reliability [17] and in other research also to have influence on behavior [18,19]. Our primary outcomes included ever having used female or male condoms for vaginal or anal sex; having used female or male condoms at last vaginal or anal sex; and the proportion of protected vaginal or anal sex acts, measured as

the number of vaginal or anal sex acts in the past 90 days protected by condoms divided by the total number of vaginal or anal sex acts in 90 days. Exposure to POWER We measured exposure to POWER by asking women about their familiarity with unique aspects of the campaign. We asked women to identify campaign elements from among a list of possible choices. Only one choice represented the correct POWER campaign element. Women who could identify that they were familiar with promotions on male or female condoms and who had seen posters or displays related to female condoms and who could identify any of the following four elements of the campaign were considered exposed: (1) The name POWER; (2) the availability of take-away cards with local statistics and instructions on condom use; (3) the availability of free silk purses; (4) the availability of free male and female condoms. In a separate analysis, we also considered women exposed if they said they were familiar with one of our campaign posters; only those saying they had seen the POWER poster, from among a multiple choice of posters, were identified as exposed. Data analyses The primary outcome analysis was conducted using permutation tests. Permutation tests are performed by referencing the observed difference between study conditions against the distribution of differences between study conditions resulting from all possible assignments of groups to study conditions. For POWER, 12 groups within four strata were assigned to two study conditions, resulting in 144 possible arrangements of groups to conditions. Cross-sectional samples were nested within each group at each of two time points, so permutation tests were performed on group-level difference scores (post-campaign–pre-campaign). Permutation tests were conducted for all primary outcome variables including measures of awareness, attitudes, and use of male and female condoms. For the post hoc secondary analyses, we analyzed data using SAS Version 9.1.3, PROC MI, and PROC MIANALYZE (SAS Institute, Cary, NC). Missing data for some study variables (e.g., items measuring attitudes) were imputed with a Markov chain Monte Carlo method generating a random sample of values based on the mean and covariance of observed values [20]. Ten imputed data sets were produced without rounding of dichotomous predictors, with point and variance estimates combined [21]. We conducted mixedmodel logistic regression analyses on imputed datasets, adjusting standard errors and p values for correlation within neighborhoods, and combined [22] the imputation-specific estimates.

S.S. Bull et al. / Journal of Adolescent Health 43 (2008) 71–78 Table 2 Demographic characteristics of the POWER study samples in California and Nevada at baseline (collected in 2004) and follow-up (collected in 2006) Characteristic Age, y 15–17 18–19 20–25 Missing Race/Ethnicity African American Latina Other Missing Ever had sex Yes No Missing Had sex in past 90 days

Baseline % (N ⫽ 3,407)

Follow-up % (N ⫽ 3,003)

41.9 (1428) 19.5 (663) 38.1 (1299) 0.5 (17)

41.9 (1259) 18.5 (555) 39.5 (1185) 0.1 (4)

33.0 (1124) 41.7 (1420) 23.1 (788) 2.2 (75)

32.3 (970) 35.4 (1064) 30.1 (904) 2.2 (65)

68.7 (2342) 29.8 (1014) 1.5 (51) 54.4 (1853)

66.8 (2005) 32.5 (977) 0.7 (21) 50.0 (1500)

Results Pre-campaign sample We identified 363 venues and 621 unique VDTs for data collection across the 12 study neighborhoods. During these 621 VDTs we identified 16,478 women who appeared to be eligible for the study. Of these women, 10,136 were approached, and 6,122 (60%) agreed to complete our eligibility questions. The remaining 4,014 women refused and we were unable to determine their eligibility. Of the 6,122 women screened, 4,032 were eligible and 3,437 (85%) completed a survey. After reviewing data we removed 30 surveys from analyses because of irreconcilable inconsistencies (e.g., participants said they had never had sex but reported having had 30 sex partners and having used a condom 45 times in the past 90 days). The final pre-campaign data set contains data from 3,407 participants.

75

Post-campaign sample Post-campaign, we visually identified 12,183 women who appeared eligible and approached 6,682 of them (55%). Of the 6,682 approached, 4,228 agreed to screening (64% of those approached) and 3,920 of those screened were eligible (78%). Of these 3,920 women, 3,036 agreed to complete a survey (92% of those eligible) and 3,030 completed a survey (99% of those agreeing to survey completion). We eliminated 23 women from analyses because of data inconsistencies; thus our final post-campaign data set includes 3,003 women. Background and some behavioral characteristics are shown in Table 2. Similar proportions of women in each sample were 15–17 and ⬎19 years (⬃40%), with the remainder 18 –19 years. Women were primarily African American or Latina. Over half the women in both surveys were sexually active, and had had sex in the past 90 days. The main study outcomes related to female condom use are shown in Table 3. Of 3,003 women post-campaign, 317 (10.5%) said they were familiar with one of our POWER posters; 193 of 1,498 women were surveyed in intervention neighborhoods saw a poster, compared with 126 of 1,505 women in comparison neighborhoods (12.8% vs. 8.3%, p ⫽ .0002). We observed no differences in condom use when comparing pre-campaign to post-campaign surveys in campaign and comparison neighborhoods. Estimates of female condom use in study neighborhoods (Table 3) range from 3.4 –15.7% at pretest to 2.8 – 13.1% at post-test. Effects were similar for condom knowledge and attitudes; that is, there were no changes in knowledge or attitudes related to male or female condoms when comparing campaign to comparison communities or when comparing pre- to postcampaign samples (data not shown). These results showed that the campaign had a null effect. However we did see evidence that there had been contamination across study neighborhoods, where women in comparison communities were able to define unique elements of

Table 3 POWER pre-and post-campaign assessments in 2004 and 2006 for proportion ever using a female condom by study neighborhoods in California and Nevada Neighborhood

Area

Condition

Pre (%)

N

Post (%)

N

East Oakland West Oakland SF-Mission SF-Lakeview Inglewood East Los Angeles East Long Beach North Long Beach Cambridge North Las Vegas Oceanside San Diego

Oakland/SF Oakland/SF Oakland/SF Oakland/SF Los Angeles Los Angeles Los Angeles Los Angeles Las Vegas Las Vegas San Diego San Diego

Comparison Comparison Campaign Campaign Campaign Campaign Comparison Comparison Campaign Comparison Campaign Comparison

15.7 11.4 7.3 13.4 8.1 4.6 9.4 7.3 7.7 6.1 3.4 9.6

229 272 284 282 270 301 296 298 285 292 293 289

10.2 4.7 12.7 12.2 9.0 9.2 8.2 4.7 6.5 6.0 2.8 13.1

244 255 244 246 255 250 243 258 248 254 248 244

Effect: 0.01941, p-value (2-tailed): 0.34722 from two-tailed permutation test on the difference scores (post-pre) averaged across strata.

76

S.S. Bull et al. / Journal of Adolescent Health 43 (2008) 71–78

Table 4 POWER 2006 post-campaign California and Nevada condom knowledge and use by level of awareness of POWER campaign elements Awareness of POWER campaign (0 ⫽ none, 4⫹ ⫽ most)

0

1

2

3

4⫹

Total N*

N % % % % % % % % % % %

1,556 30.0 8.5 8.3 13.2 2.1 2.0 90.8 72.9 60.8 80.9 77.9

395 100.0 64.6 64.6 69.9 14.3 4.8 97.0 90.1 86.1 92.7 86.0

517 100.0 70.4 74.1 72.5 20.2 5.9 96.7 92.3 88.8 91.9 87.8

275 100.0 81.5 85.1 82.5 22.4 4.3 98.6 94.2 93.8 94.5 91.0

260 100.0 85.8 88.8 84.6 27.3 12.7 97.7 96.5 95.4 94.6 90.7

3,003 1,912 1,198 1,232 1,304 244 90 2,821 2,477 2,251 2,605 1,678

Heard of a female condom Seen any information about female condom Read information about female condom Seen a female condom Ever used a female condom* Used female condom at last sex* Heard of male condom Seen any information on male condom Read information about male condom Seen a male condom Ever used a male condom*

Note: Awareness was scored by assigning one point each time a woman identified an element unique to the POWER campaign (e.g. Take-away cards with STI data; Posters in Bathrooms about STI and/or Pregnancy; silk purse with condoms). * Among those who have had sex (N ⫽ 2,005).

the POWER campaign, suggesting that they had been exposed to POWER (e.g., 39% of the 87 women saying they received a silk purse with condoms at follow-up were from control neighborhoods). Because we learned that many women across all study communities had exposure to POWER, we decided to conduct a post hoc analysis of the post-campaign data for all women, regardless of neighborhood, to learn whether those exposed to POWER had different awareness, attitudes, and condom use compared with women not exposed. Results from these analyses are presented in Tables 4 and 5. We explored whether having had exposure to multiple campaign elements (i.e. posters, take-away cards, campaign name, silk purses, and free condoms) had any greater effect on female condom attitudes, norms, intentions and use

Table 5 POWER 2006 post-campaign California and Nevada predictors of having used a condom at last sex Total (n) in model Number of women (n) using condom last sex

2,005 1,132

Variable

Odds ratio

95% confidence interval

1.50 Ref 0.82 0.56 1.98 1.24 1.30

(1.07, 2.11)

Seen Posters Age

Intentions to Use Condoms# Condom Use Norms # Positive Attitudes#

Category

15–17 18–19 20–25

(0.60, 1.13) (0.43, 0.73) (1.74, 2.26) (1.08, 1.43) (1.16, 1.46)

Note: Values derived from a mixed-model logistic regression analysis. # Higher scores indicate higher positive attitudes, a three item scale ranging from 0 –15; lower negative attitudes on a three item scale ranging from 0 –15; Higher scores indicate greater intentions to use condoms on a three item scale ranging from 0 –15; and higher scores indicate more positive norms towards condoms on a five item scale ranging from 0 –25. These measures are described in detail elsewhere.2

among women who already had heard of a female condom. Of the 3,003 women surveyed at post-campaign, 1,556 (52%) had no exposure whatsoever to POWER, and 48% had exposure to between one and four campaign elements. There were 395 (13%) who had seen at least one campaign element, 517 (17%) who had seen two elements, 275 (9%) who had seen three, and 260 (8%) who had seen four. It appears that with more frequent exposure to unique POWER elements, women had more often seen female condoms, read information about them, and used them (Table 4). In an effort to control for other factors that could influence female condom use, we conducted a logistic regression analysis to determine what influence exposure to the POWER posters might have on female condom use. Data in Table 5 show that women exposed to one of our POWER posters were 1.50 (CI ⫽ 1.10 –2.11) times as likely to have used a condom at last sex. Other predictors of having used female condoms included age; individuals 15–17 years of age (our referent group) were more likely to have used them compared with individuals ⱖ18 years of age, and women with greater positive attitudes toward condoms, greater positive condom norms, and greater intentions to use condoms were also more likely to have used a female condom. We controlled for race/ethnicity in this analysis and saw no effect, and therefore did not include this in the final model.

Discussion We have identified several important aspects of this study to consider. By considering neighborhood as the unit of analysis and using a time–space sampling approach, we were able to conduct a rigorously controlled randomized trial of an environmental level intervention and improve our abilities to generalize our findings to a larger population. This type of intervention is frequently overlooked because of po-

S.S. Bull et al. / Journal of Adolescent Health 43 (2008) 71–78

tentially high cost or complexity, thus limiting our understanding of the efficacy of environmental level interventions. Our sampling method has improved generalizability that most cross-sectional studies or panel assessments do not have. Ultimately, however, the result of this more rigorous assessment was to show a null effect. There are several reasons that this could have occurred. It is likely that the campaign was simply not intensive enough and did not use enough communication channels to saturate the market. With a small media campaign we may have had limitations on effect without complete saturation of a neighborhood, that is, materials in most business and community venues, an approach that would likely not be replicable, particularly for campaign promoting condoms. We conducted an evaluation that included intercept points for participants that did not coincide with every campaign site; this was intentional to determine how well effects might have “spread” through neighborhoods. It is certainly possible that we intercepted women who never frequented the organizations displaying our materials. We may have been able to see greater exposure if we had limited our intercept points to only those sites where the campaign was displayed, but that would have limited our ability to generalize findings to the entire neighborhood. It is also possible that women may have been exposed to the campaign but that effects were short lived and not documented within a short enough time frame, as is suggested by other researchers [23]. Our decision to wait six months to evaluate the campaign was driven by venue owners who requested that we wait a longer period between pre-and post data collection periods. The challenge of community-based work in this case was to maintain good relationships with venue owners and not overload them with study activities in their places of business. It is possible that our campaign and comparison neighborhoods were situated too close to one another, and that this promoted contamination across sites. Finally, it is possible that we did not have enough statistical power to detect a change in condom use; our study was powered to detect changes of 20%, a fairly large effect. Of note is that our post hoc analyses show different findings. These analyses use methods similar to those of others evaluating social marketing campaigns [24]. This raises a question as to the importance of introducing more rigor into all evaluations of social marketing, to ensure that effects are not artifacts from baseline or nonrandomized comparison groups. The post hoc analyses do raise some issues for social marketing that are worth considering in future studies. Is it necessary, for example, to use multi-modal campaigns to increase the intensity and duration of the effect from a campaign? How can we achieve that in a public health environment that is relatively resource poor compared with other commercial industries? The work of Randolph and Viswanath [10] is instructive in considering these questions. In their examination of 52

77

social marketing campaigns, they determined that it is not necessary to have more money in a campaign to effect change, but that specific efforts to amplify and reinforce materials is critical. They advocate building partnerships within communities to leverage support for the issue and getting corporate sponsorship for public service announcements or other campaign elements. With POWER, the use of time–space sampling to enumerate women for evaluation also proved an effective approach to identifying key placement opportunities for the campaign. Future efforts could develop relationships with venue owners and managers where the target audience congregates to leverage more community support. In the absence of unambiguous advocacy that may exist for other public health concerns, such as seat belt use or tobacco prevention for youth, more intensive promotion of condoms with limited resources may require careful partnering in communities substantially affected by sexually transmitted infections and human immunodeficiency virus.

Acknowledgments This research was supported by Cooperative Agreement TS 781 from the Association of Teachers of Preventive Medicine and the Centers for Disease Control and Prevention. The findings and conclusions in this paper are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.

References [1] Maibach EW, Van Duyn MA, Bloodgood B. A marketing perspective on disseminating evidence-based approaches to disease prevention and health promotion. Prev Chron Dis 2006;3:A97. [2] Noar S. A 10-year retrospective of research in health mass media campaigns: Where do we go from here? J Health Commun 2006;11: 21– 42. [3] Grier S, Brant C. Social marketing in public health. Annu Rev Public Health 2005;26:319 –39. [4] Cohen DA, Farley TA, Bedimo-Etame JR, et al. Implementation of condom social marketing in Louisiana. Am J Public Health 1999;89: 204 – 8. [5] Lamptey PR, Price JE. Social marketing sexually transmitted disease and HIV prevention: A consumer-centered approach to achieving behavior change. AIDS 1998;12(Suppl 2):S1–9. [6] Keating J, Meekers D, Adewuyi A. Assessing effects of a media campaign on HIV/AIDS awareness and prevention in Nigeria: Results from the VISION Project. BMC Public Health 2006;6:123. [7] Kennedy MG, Mizuno Y, Seals BF, et al. Increasing condom use among adolescents with coalition-based social marketing. AIDS 2000;14:1809 –18. [8] Dubois-Arber F, Jeannin A, Konings E, Paccaud F. Increased condom use without other major changes in sexual behavior among the general population in Switzerland. Am J Public Health 1997;87:558 – 66. [9] Meekers D, Rahaim S. The importance of socio-economic context for social marketing models for improving reproductive health: Evidence from 555 years of program experience. BMC Public Health 2005;5:10.

78

S.S. Bull et al. / Journal of Adolescent Health 43 (2008) 71–78

[10] Randolph W, Viswanath K. Lessons learned from public health mass media campaigns: Marketing health in a crowded media world. Annu Rev Public Health 2004;25:419 –37. [11] Bull SS, Cohen J, Ortiz C, Evans T. The POWER campaign for promotion of female and male condoms: Audience research and campaign development. Health Commun 2002;14:475–91. [12] Zucker D, Wittes J, Schabenberger O, Brittain E. Internal pilot studies II: Comparison of various procedures. Stat Med 1999;18:3493–509. [13] Mhyre S, Flora J. HIV/AIDS communication campaigns: Progress and prospects. J Health Commun 2000;5(Suppl):29 – 45. [14] Centers for Disease Control and Prevention. Youth risk behavioral surveillance—2003. MMWR 2004;53:SS-02. [15] Pals S, Beaty B, Posner S, Bull S. Estimates of intraclass correlation for variables related to behavioral HIV/STD prevention in a predominantly African-American and Latino/Hispanic sample of young women. Evaluation (in press). [16] Muhib FB, Lin LS, Stueve A, et al. A venue-based method for sampling hard-to-reach populations. Public Health Rep 2001;116(Suppl 1): 216 –22. [17] Posner SF, Bull SS, Ortiz C, Evans T. Factors associated with condom use among young Denver inner city women. Am J Prev Med 2004;39:1227–33.

[18] Albarracin D, Gillette JC, Earl AN, et al. A test of major assumptions about behavior change: A comprehensive look at the effects of passive and active HIV-prevention interventions since the beginning of the epidemic. Psychol Bull 2005;131:856 –97. [19] Albarracin D, McNatt PS, Klein CT, et al. Persuasive communications to change actions: An analysis of behavioral and cognitive impact in HIV prevention. Health Psychol 2003;22:166 –77. [20] Rubin DB. Multiple Imputation for Nonresponse in Surveys. New York, NY: John Wiley & Sons; 1987. [21] Horton NJ, Lipsitz SR, Parzen M. A potential for bias when rounding in multiple imputation. Am Stat 2003;57:229 –32. [22] Little RJA, Rubin DB. Statistical Analysis with Missing Data. 2nd ed. Hoboken, NJ: Wiley; 2002. [23] Snyder LB, Hamilton MA. A meta-analysis of U.S. health campaign effects on behavior: Empahsis on enforcement, exposure, and new information, and beware the secular trend. In: Hornik RC, ed. Public Health Communication: Evidence for Behavior Change. Philadelphia: Lawrence Erlbaum Associates, Taylor and Francis Group; 2001: 357– 84. [24] Myhre S. Evaluation Report: Narrowcasting HIV Prevention Campaign. New York: Gay Men’s Health Crisis; 1996.