Numerous epidemiological studies indicate that diet plays a major role in premature morbidity and mortality in the United States (1). Poor diet and a lack of physical activity may eventually overtake smoking as the leading preventable causes of death (2). Campaigns promoting positive changes in dietary behavior have great potential for improving the public's health.

Milk is a good choice for use in a community campaign to improve health behaviors because it is consumed by so many people and plays an important role in health and the diet. High-fat milk contributes significant amounts of excess calories and saturated fat to the American diet, is the third-leading source of saturated fat in the diet of American adults, and is the leading source of saturated fat in the diets of children aged older than 2 years (3-6). Saturated fats raise blood cholesterol levels and increase the risk of coronary heart disease (1,7). In contrast, skim milk has 40% fewer calories and 5 fewer grams of saturated fat per cup than whole milk. Only six types of food contribute about half of the saturated fat consumed by American adults (4), so we chose one of the items — milk — as the focus of our campaign. (A more recent study found that seven types of food contribute half of the saturated fat consumed by the average American [3].)

Mass-media campaigns may be an effective way to address diet-related population health behaviors, and paid mass-media campaigns have been shown to be a useful way to deliver a public health message to numerous people (8). Use of paid television, radio, and newspaper advertising combined with effective media relations have been shown to significantly affect health behaviors when a high level of market penetration is achieved (i.e., when the target audience is repeatedly exposed to the campaign message) (9).

In 1995, The Center for Science in the Public Interest began developing a mass-media community campaign — 1% Or Less — to encourage one important behavioral change: switching from high-fat (whole or 2%) to low-fat (1% or skim) milk (10) — to encourage one important behavioral change: switching from high-fat (whole or 2%) to low-fat (1% or skim) milk (10). The campaign involved three basic components: 1) paid advertising, 2) media relations, and 3) community-based educational programs. The campaign was implemented in numerous communities nationwide, and the results of the pilot campaigns have been published (11-13).

In this study, we assess the cost-effectiveness of various combinations of the 1% Or Less campaign components in four individual West Virginia communities. We compare the cost, exposure, and outcomes to the campaign message of the four different types of campaign combinations: 1) paid advertising, media relations, and community-based educational activities; 2) paid advertising and media relations; 3) media relations and community-based educational activities; and 4) paid advertising alone. The paid advertising component consisted of professionally produced, strategically placed television, radio, and newspaper advertising. Media relations comprised events designed and implemented to generate coverage by the local news media. The community-based educational activities included events such as blind milk taste tests in grocery stores, point-of-purchase signs about the program, school activities such as poster-design contests, and nutrition seminars conducted by trained speakers at work sites and for various organizations.

The health goal of the campaigns was to encourage community members (older than 2 years of age) to switch from high-fat to low-fat milk. The health communications research goal was to understand how the various combinations of the three 1% Or Less components work in community-based campaigns. The methodology was approved by the West Virginia University Institutional Review Board.

A quasi-experimental design was used in communities with populations ranging from 18,000 to 35,000. Each of the four communities received an intensive 6- to 8-week campaign, and each campaign involved a different combination of the three 1% Or Less components (Table 1). The campaigns had a rolling field experiment design. All campaigns were conducted during February and March for 3 consecutive years beginning in 1996 (Table 1). A matched comparison community for each campaign community was observed and received none of the campaign messages. None of the communities had overlapping media markets and thus were not exposed to the other communities' 1% Or Less campaign messages. All four communities had similar demographics, and their campaigns had similar budgets (11-13).

The Clarksburg, Beckley, and Wheeling campaigns all incorporated the paid advertising component and delivered the 1% Or Less message to television viewers (and therefore to significantly more people than those actually living in the cities for which data were collected). The television message was delivered to approximately 278,000 (Clarksburg), 363,000 (Beckley), and 418,000 (Wheeling) viewers (11-13). The Parkersburg campaign, which did not incorporate the paid advertising component but had extensive local newspaper coverage, delivered the 1% Or Less message to approximately 22,500 community members, which is the number of people who subscribed to the local daily newspaper.

The first trial was conducted in Clarksburg and included all three 1% Or Less components (paid advertising, media relations, and community-based educational activities) (11). The campaign in Wheeling consisted of paid advertising and media relations (12), and the Parkersburg trial involved a combination of media relations and community-based educational activities (13). The community trial in Beckley involved paid advertising only (13). The Parkersburg and Beckley campaigns were conducted simultaneously and were compared with the same comparison community.

In each campaign and comparison community, we conducted random-digit–dial telephone surveys of milk purchasing and consumption habits for approximately 400 adults immediately before the beginning of the campaign (11-13). A panel design was used. We called the baseline respondents again immediately after the campaign and were able to reinterview 69% of the respondents in Clarksburg, 73% in Wheeling, 67% in Parkersburg, and 67% in Beckley. We collected 1 month of fluid milk sales data from all supermarkets in the campaign and comparison communities for the month immediately before and the month immediately after (i.e., beginning the day after), 6 months after, and 12 months after the campaigns ended. In addition, 2 years after the campaign ended in Wheeling, we collected fluid milk sales data for 1 month. Communities were not randomly assigned to campaign or control.

The Clarksburg campaign included paid advertising, media relations, and community-based educational activities, which increased low-fat milk sales from 18% to 41% (Table 2). The low-fat milk sales were still higher (33%) 1 year after the campaign ended. Wheeling's campaign involved paid advertising and media relations and increased the low-fat milk sales from 29% to 46%, a change that was sustained at 42% 2 years after the campaign ended. Analysis shows that the increases in low-fat milk sales were statistically significant in the Clarksburg and Wheeling campaigns  (Table 3) (Table 4). Smaller increases in low-fat milk sales were documented after the Parkersburg campaign, which used media relations and community-based educational activities, and after the Beckley campaign, which only used paid advertising. The increases were not significant.

Tables 3 and 4 provide the descriptive statistics for low-fat milk sales in each of the four communities. The treatment effect for the precampaign to postcampaign (the day after the campaign ended) increase in low-fat milk sales (expressed as r) ranged from 0.01 (for the paid advertising campaign in Beckley) to 0.64 (for the Clarksburg campaign, which used paid advertising, media relations, and community-based educational activities) (Table 3), with a mean of 0.41. The average d effect size was 1.01 (z = 2.89, P = .002), which is a large effect. Increases in sales of low-fat milk in campaign communities were an average of 1 standard deviation larger than in comparison communities. No statistically significant between-group heterogeneity was found (χ² <1.00); that is, no statistically significant differences were found between campaigns, in spite of the large differences in effect sizes. However, given the wide range of effect size results (0.02 in Beckley to 1.67 in Clarksburg), the lack of statistically significant heterogeneity may be a result of the small sample size.

The precampaign to 6-month postcampaign effect (r) for low-fat milk sales ranged from 0.01 (Beckley) to 0.59 (Clarksburg) (Table 4). The average d effect was 0.84 (z = 2.89, P = .002), which is also a large effect. Increases in low-fat milk sales in campaign communities were an average of 80% of a standard deviation larger than in comparison communities. Assessment of between-group heterogeneity revealed no differences (χ ² <1.00).

The treatment effect across the four 1% Or Less campaigns as measured by survey respondents who reported switching from high- to low-fat milk ranged from 13% (Beckley) to 38% (Clarksburg) (Table 5). In Clarksburg, 38% of respondents who reported consuming whole or 2% milk before the campaign reporting drinking 1% or fat-free milk immediately after the campaign (P <.001). In Wheeling, 34% switched from high- to low-fat milk after the campaign (P <.001). In Parkersburg, 20% (P <.001) switched, and in Beckley, 13% (P <.001) switched.

Expressed as an r effect, the switching rates range from 0.10 (Beckley) to 0.39 (Wheeling), with a mean of 0.25. The average weighted d effect size was 0.53 (z = 3.03, P <.001), a medium effect (Table 6). These results indicate that self-reported switching in campaign communities was approximately half a standard deviation greater than in the comparison communities. A test for heterogeneity indicated significant heterogeneity between the different campaigns (χ ² ₃ = 8.606; P = .02) measured at the survey sample. Although the average effect was significant, results suggest a discernable difference in the effectiveness of the various campaigns.

Campaign exposure was assessed by asking telephone survey respondents about their awareness of the 1% Or Less message. The last column of Table 5 shows percentages of self-reported exposure to the campaigns in the four campaigns. We computed a linear contrast among the four effects (weights of 3, 1, −1, and −3), producing a significant effect (z = 3.161; P = .003). We also explored the possibility of nonlinear effects (weights 1, −1, −1, 1) and found a smaller but still significant effect (z = 2.12, P = .02).

We used several comparisons to explore patterns of nonlinearity. The strongest pattern suggests that campaigns in Clarksburg and Wheeling resulted in greater changes in low-fat milk consumption than in Parkersburg and Beckley (z = 2.72, P = .003) (Table 6). No reliable difference between the effects of the Clarksburg and Wheeling campaigns (z <1.00) was found, nor was a reliable difference found between the Parkersburg and Beckley results.

The cost of each campaign is shown in Table 7. Overall costs of each campaign were similar, ranging from $43,000 (Wheeling) to $61,000 (Clarksburg). In contrast, the number of people exposed to each campaign varied widely. As a result, the approximate cost per person exposed to each campaign ranged from $0.10 (Wheeling) to $2.27 (Parkersburg).

We estimated how much it cost in each campaign to cause one person to switch from high- to low-fat milk. The wide variation in the number of people exposed coupled with differences in switching rates among the campaigns contributed to a wide range in cost. The cost to cause one person to switch from high- to low-fat milk ranged from $0.57 (through the paid advertising and media relations campaign in Wheeling) to $11.85 (through the community-based educational activities and media relations campaign in Parkersburg). The combination of paid advertising and media relations in Wheeling cost approximately $0.10 per person exposed, whereas the Parkersburg campaign (which involved media relations and community-based educational programs) cost $2.27 per person exposed.