We implemented a series of steps designed to identify obesity-related policies in Missouri. Our first step was to establish community partnerships so we could contact key informants from our 7 target environments. We collaborated with staff of the Prevention Institute, a leader in the development of the ENACT database, to learn from their methods and approaches (12
). We also established an expert advisory group that reviewed MoNAP project goals, definitions, and data collection methods and provided initial contacts from each of our target environments. The advisory group consisted of educators, politicians, health care administrators, and members of state government recognized as leaders in obesity policy. Finally, we worked extensively with the Missouri Council of Activity and Nutrition (MoCAN), a coalition of representatives from groups interested in implementing the Missouri statewide obesity prevention plan. MoCAN has more than 46 active members from academia, business, health care, and other community groups. MoCAN members and workgroups were critical to educating the public about MoNAP and in securing extensive contact information in the target communities.
To ensure common language and consistency in assessing policies from each environment, we 1) identified a list of key terms, 2) conducted a literature search on commonly accepted definitions, 3) reviewed these definitions with key members of our team and advisory group, and 4) came to consensus about the meaning of our core constructs (12
). We defined obesity-related policies as written documents describing a strategy, plan, or objective related to carrying out a physical activity or nutrition-related agenda (2
). We used standard definitions and examples to guide the policy assessment of each of our target environments ().
Standard Definitions Used to Assess Environments — the Missouri Obesity, Nutrition, and Activity Policy Database, 2007-2009
Study design and sampling plan
This study took place between 2007 and 2009. The Washington University in St. Louis institutional review board approved the conduct of this study. This study employed a stratified nested cluster design. The primary unit of sampling was the county; we obtained policies for sampled counties. The study team stratified the sample by using the 5 health regions as defined by the Missouri Department of Health and Senior Services. These regions are northwest (28 counties), central (29 counties), eastern (11 counties), southwest (24 counties), and southeast (22 counties).
We characterized each county as urban (>75% of residents living in an urbanized area or urban cluster), mixed residence (25% to 75% living in an urbanized area or urban cluster), or rural (<25% living in an urbanized area or urban cluster) and by racial/ethnic composition within county type. We drew the sample of counties proportionally to the number of counties in each health region and to the number of counties at each level of urbanization in each health region. For each county type (rural, mixed, and urban), we stratified by the relative racial/ethnic composition of the county (≥5% or more African American, 1% to <5% African American, and <1% African American). We oversampled counties with higher percentages of African Americans.
We next drew a stratified sample of cities from each of the sampled counties. We divided the cities within a county into tertiles based on population size. For each sampled city, we obtained policies from the 7 target environments. This approach enabled us to determine the presence or absence of obesity-related policies in 2,356 environments in 89 Missouri counties.
The government environment included city governments and special districts. The 114 counties in the state of Missouri necessitated a sample of 89 counties. (County-level policies were too broad to include.) The 972 cities in the state of Missouri required a sample of 276 city policies. We sampled city policies from each of the sampled counties (one from the highest, middle, and lowest tertile of city sizes), yielding a sample of 267 cities. Large cities were defined as those in the highest tertile (n = 89). Additionally, we forced the largest and smallest cities of each of the 5 regions into the sample for an additional 10 cities, yielding a total city count of 277.
For the community environment (eg, church associations or community centers) we sampled policies from 2 locations per city, yielding 554 locations for review. We identified 108 hospitals from the 23 health care systems throughout Missouri that provided coverage for the sampled cities. It was beyond the scope of this project to enumerate the number of worksites in the state of Missouri; however, we ensured representativeness by sampling policies from 2 worksites per sampled city (1 public worksite such as a park district office and 1 private worksite) for a total of 554 contacts for review. We defined school environment as the 432 school districts in the state of Missouri, requiring a sample of 204 districts. Because several cities are served by the same school district, a final target sample size of 217 was reached. Because of overlap in the school environments, we obtained 1 private after-school program policy (eg, YMCA) from each city. Finally, we sampled only child care centers licensed by the state (not family homes or group homes). One child care center was sampled from each sampled city, and an additional center was drawn from each large city (n = 92), yielding a total of 369 locations for review.
We used the MoNAP Policy Checklist (MPC) to assess the content of collected obesity-related policies. The MPC was based on the ENACT checklist and was modified to have 4 sections: demographics, topics, status, and funding. Demographic data included general information about the organization and key informants (eg, key informant name and title, organization name, policy name, city and county of organization site). The topics section included policy focus (eg, physical activity, nutrition, or both) and the presence or absence of obesity-related content (eg, access to fresh foods, body mass index reporting, land use/planning/zoning, rails to trails). The status section collected information on the type of policy (eg, city plan, ordinance) and addressed whether the policy was proposed or adopted. The funding section assessed whether the policy allocated funds for implementation. Informants were specifically asked whether funding was available and the source of that funding. Finally, 4 open-ended questions addressed history of policy development, policy adoption and implementation challenges, policy enforcement, and methods of policy evaluation by the organization.
Data collection and analysis
We used 2 primary methods of data collection: key informant interviews and review of public documents. Key informant interviews were designed to generate a representative sample of policies and names of additional key informants from whom to gather policies. To identify these key informants, we worked with our advisory group and MoCAN to secure lists of contacts associated with each environment. For example, in the school environment, we identified people holding 1 of 4 positions: principals, physical education teachers, school nurses, and food service workers. We sent e-mails to these key informants explaining the purpose of the project and requesting samples of obesity-related policies. We followed the e-mails with a telephone call from project staff to review the project goals and ask key informants whether their organization had any written obesity-related policies and whether they would provide a copy of any written policy and names of other contacts. We made 3,666 contacts: government (n = 518), community (n = 444), health care (n = 299), worksite (n = 571), schools (n = 761), after school (n = 457), and child care (n = 616).
We also systematically reviewed public documents associated with each of the target environments. We conducted a Web-based search using various search tools (eg, Lexis, Nexis, Google). We identified relevant Web sites for the target environment (eg, schools) and key links to public policy-related documents (eg, school board meeting minutes), and collected any obesity-related policies or supporting documents. For example, for the government environment, we conducted Web-based searches to collect city council minutes, resolutions, and ordinances relevant to obesity-related policies.
We coded each site by information on policy status as policy available, no policy available, or participation declined. No policy available meant a verbal response of "no policy" or that we were unable to verify the presence of a policy. Participation declined meant a verbal response of "will not participate" or inability to contact the participant after at least 5 attempts (via telephone, e-mail, or both). Two project team policy analysts evaluated policies from "policy available" sites using the MPC. If there was disagreement in any MPC area, a third member of the project team assessed the policy and recommended category. There was 96% agreement among all policies collected. Most disagreements were due to omission as opposed to interpretation of content in each environment: government (24%), community (3%), health care (10%), worksite (2%), school (37%), after school (9%), and child care (15%). The difference in the rate of agreement for each environment was probably an artifact caused by capturing more policies in some environments than in others (eg, schools vs community). We used descriptive statistics to determine policy presence across each of the 7 environments.