Investigators from all of the 6 PRC CHIRTN sites shared leadership for the project, first to select the research topic and then to reach consensus on a core protocol. Local community teams or advisory committees participated to ensure relevance to partner communities and to assess the language and clarity of discussion questions and research materials for participants. Institutional review boards at each university reviewed and approved the protocols for that site. Investigators met monthly by telephone conference and in March 2008 in person at a national PRC CHIRTN meeting to develop the core protocol and then to track progress and resolve questions.

The protocol included a series of discussion questions for use at all sites and suggested demographic items to document the diversity of samples and allow comparisons between sites. A full list of questions and prompts is available from the corresponding author. Four categories framed the discussion questions:

The suggested set of demographic information included age, sex, race/ethnicity, education, geographic location (rural or urban), and length of residence in the United States (if an immigrant population). Several sites also collected information about participants' knowledge of their cardiovascular risk status or history.

Each site chose the appropriate qualitative method for its partner community and site-specific goals, which in some cases extended beyond this study's aims. Five sites (UCD, UIC, UNC, UW, and WVU) conducted small-group sessions of ethnically, geographically, or socioeconomically underserved groups, and 1 (UR) conducted in-depth interviews and neighborhood walking tours with families of urban school children. The facilitators or interviewers for all research sessions were experienced university-based researchers trained in qualitative methods. Following principles of community-based participatory research that promote collaborative partnerships in all phases of research (15), community advisory groups identified and recruited participants representative of each community of interest. On the basis of criteria reviewed and approved by local institutional review boards, participants provided oral or written consent.

Small-group sessions were conducted in English with 2 exceptions; new-immigrant Hispanic sessions were conducted in Spanish (UIC), and Asian immigrant groups were conducted in separate sessions in Mandarin, Cantonese, Vietnamese, or Korean (UW). Small-group sessions took place during the summer of 2007. All sessions were audio-recorded; researchers at each site transcribed (and if necessary translated into English) that site's recordings, removing any personal identifiers. Site-specific details about the groups are provided (Table).

Investigators from all sites assembled a "universal" codebook. The codebook consisted of a priori codes derived from the discussion questions and additional concepts that emerged during analysis. The major coding categories were the following:

Personnel at each site open coded (16) that site's transcripts; at least 2 researchers independently coded and then reconciled differences, using Atlas.ti (Atlas.ti Scientific Software Development, GmbH, Berlin, Germany) software at 4 sites and NVivo/NUD*IST (QSR International, Cambridge, Massachusetts) at 2 sites to facilitate analysis. Investigators then conducted axial coding, combining the original codes into categories by connecting them "in terms of conditions that give rise to them, properties that are common to them, strategies that guide them, and consequences they share" (16). After each site's investigators coded and identified themes, representatives from all sites met face-to-face and then by conference call to identify cross-cutting themes and differences across sites. Investigators then selected representative quotations from their sites to illustrate the research findings.

A total of 380 community members participated in qualitative projects across the 6 sites. Participants included 62 children (UR, WVU) with their parents and other interested adults. This analysis includes only the 318 adults (Table). Of the adults, 85% were women (some sites by design sampled only women); 22% were younger than 45 years, 32% were aged 45-64, and 29% were aged 65 or older (additionally, 9% reported age as 35 or older, 1% reported age as 40 or older, and 6% did not provide age information). Slightly more than half (54%) of the adult participants were married.

Recruitment achieved the desired oversampling of at-risk populations. Most of the participants (89%) were of Hispanic ethnicity or nonwhite race; 39% reported themselves as African American, 25% Hispanic, and 24% Asian. More than one-fourth (27%) of respondents had less than a high school education. Although only 2 sites collected information on insurance coverage (UCD, UIC), the rate of those with no insurance (43%) was almost twice the national average (23%) (17). Three sites collected information on health history; 48% of participants self-reported a family history of heart disease (UCD, UIC, UNC), and 28% self-reported a diagnosis of diabetes (UCD, UNC). Rural residents made up 33% of the sample. All small-group sessions at WVU and UCD took place in rural areas (small towns outside of Morgantown, West Virginia, and small towns in the rural San Luis Valley of Colorado), and 51% of UNC respondents identified themselves as rural residents.

Across varied sites and diverse populations, we found that participants who were not recent immigrants had basic general knowledge about CVD, risk factors, disease outcomes, and prevention but experienced locally specific challenges to putting knowledge into practice. Most participants generally recognized high blood pressure and high blood cholesterol levels as risk factors for heart disease, understood the roles of lifestyle and genetics in these risk factors, knew the numbers that indicate high blood pressure and high blood cholesterol levels, and recognized lifestyle and medications as components of reducing risk.

New immigrant populations (UW, UIC) had only rudimentary or incomplete knowledge about CVD, compared with the more established populations. For example, most participants in the UW small-group sessions knew high levels of blood pressure were bad but had difficulties describing what high blood pressure is. Participants in the UIC groups knew that CVD was related to lifestyle factors including diet but did not know the numbers defining high blood pressure and high blood cholesterol levels. Participants in these settings did not understand the importance of taking medication to treat high blood pressure and cholesterol levels.

Participants at all sites had less knowledge about stroke than about heart disease, and they reported fear of stroke. They expressed concern about the loss of independence associated with stroke, the resultant economic implications for families, and the burden on family members who provide care.

Across small-group sessions and interviews, common socioecological themes emerged as barriers to translating knowledge into healthy behaviors: multiple role demands, lack of economic resources, social and cultural issues that include lack of family and community support for healthy habits, and concerns about health care. Appendix A provides examples in participants' own words.

Particularly for women, their multiple roles as wage-earners, household managers, child care providers, providers of elder care, and people responsible for dealing with health care demands for all family members left them too exhausted to attend to their own health, without time to exercise, and too overextended to prepare nutritious meals at home as often as they liked or knew they should.

Many respondents lamented the high price of fresh fruits and vegetables, the cost of gas for transportation, and the cost of health insurance. Living in resource-poor communities contributed to their difficulties.

Respondents at most sites remarked on the number of fast-food restaurants, citing them as often the only source of meals outside the home. They observed that corner stores with limited fresh foods were more accessible than well-stocked supermarkets. Their neighborhoods often lacked safe walking environments or facilities such as parks, gyms, and playgrounds. Additionally, at some sites, a strong street culture (eg, drugs, violence) competed with families for the attention of their children and created barriers to healthy behaviors in the young. Even for respondents with health insurance, difficult access (eg, lack of transportation) created barriers to appropriate health care. Hispanic respondents noted the lack of providers who speak Spanish, and new immigrants had problems navigating the US health care system and integrating it with traditional beliefs and practices. Asian respondents preferred to use their own community physicians and pharmacists for advice and to rely on their own traditional remedies and products because they perceived that they had access to few other resources for information. Respondents spoke of weather and transportation as barriers to healthy physical activity.

Community and social context in some cases facilitated healthy behavior (Appendix B). Family and friends provided knowledge and services such as child care. Learning occurred best in a social context, with information in language that people use and understand. Social support — walking partners, family reinforcement, healthy behavior role models — improved the likelihood of adopting and maintaining healthy changes. On an individual level, several people spoke of the importance and the difficulty of motivation.

Respondents identified a number of information sources that included health care providers, family and friends, printed materials, and other media (eg, Internet, television, radio). Asian immigrants (UW) trust health information from physicians despite communication problems. They also trust information from other providers (eg, acupuncturists), family or friends, and community centers when offered in their native languages. Recent immigrant Hispanic respondents (UIC) reported receiving little CVD information in any form. Nonimmigrant groups (UR, UCD) preferred peer-to-peer discussion groups and family and friends for information and to generate strategies for putting knowledge into practice. Respondents from all sites found currently available print and other media materials of limited usefulness.

Respondents provided suggestions for disseminating health information into their communities. They identified the following characteristics of successful strategies: responding to the influences of intrapersonal (motivation), interpersonal (social), community (cultural), and institutional contexts. They cited interactive personal contact, use of preferred language ("just being able to present things in a way that people can understand and give them ideas, don't just say — 'here, eat healthy'" [UR]), and tailoring to cultural values and priorities of the local community (eg, promoting community gardens). Respondents recommended family-friendly group sessions ("a little workshop or a little women's retreat or something like . . . , if I experienced it and did some things I would be more likely to incorporate them in my life" [UCD]). They preferred to hear from informed family and friends, especially those who speak of their own experiences; trusted groups (eg, AARP); and health care providers, but not always ("I wouldn't really listen to a doctor. I would listen to a friend that has been through it." [UNC]). Asian and Hispanic immigrants (UIC, UW) expressed the need for educational materials in their own languages and for dissemination approaches sensitive to community and culture.

Small-group sessions and interviews may not adequately mirror the characteristics of the community or population that they represent. The diversity of settings we reached suggests that the results described here reflect similarities and differences across underserved groups with increased CVD risk. Although the research sites used different data collection methods, the development of common protocols, discussion questions, and coding schemes means that we spoke of the same issues in the same ways. Investigators from all sites collaborated to conduct the highest level of analysis and synthesis, which increased the rigor of the results and the implications drawn from them.

Public health practitioners and programs must reach people at high risk and engage them in prevention activities. The results of this study make it clear that we need to move beyond individual- and knowledge-based interventions to new approaches that involve social marketing; environmental change to improve access to nutritious food, physical activity, and health care; and strategies tailored to the context.

This study identified a number of common barriers to cardiovascular health across groups of underserved populations that community interventions may address. Although sensitivity to unique cultural settings must be considered, many similarities exist across groups concerning suggestions for approaches to improve knowledge and CVD prevention practice. The socioecological perspective provides a framework for creating multifaceted disease prevention interventions and related communication strategies that simultaneously target the different levels of influence and build on community strengths.