Nearly 21 million people in the United States have diabetes, and prevalence is particularly high among racial/ethnic minorities and populations with low socioeconomic status (1). Community health centers (CHCs) serve more than 15 million patients annually, 64% of whom are from racial/ethnic minority populations and 92% of whom have incomes below 200% of federal poverty level (2). CHCs are thus ideal settings for monitoring and improving the quality of diabetes care and reducing diabetes-related health disparities. In April 2006, the Massachusetts League of Community Health Centers initiated the Massachusetts Diabetes Health Disparities Collaborative, a state-based quality improvement initiative for 17 CHCs (3). For 13 months, participating CHCs focused on achieving system change in the delivery of primary health care for patients with diabetes by applying the Chronic Care Model (4). The 6 components of the Chronic Care Model are self-management, decision support, delivery system design, clinical information, organization, and community. When fully functioning, these components constitute a system that encourages high-quality chronic disease management, productive interactions with patients who take an active part in their care, and support for providers with resources and expertise (5).

Although several studies (6-11) report on the success of methods used by collaboratives, as measured by outcomes data, few (6,12,13) have reported specific, detailed strategies for achieving success and overcoming challenges. Wang et al (12) informally interviewed collaborative team leaders at 2 of 4 CHCs participating in the North Carolina Diabetes Collaborative and found that senior leadership support, physician champions, multidisciplinary teams, and making the collaborative a priority for the health care team were factors for success. Barriers were staff turnover, lack of senior leadership and physician support, and low priority in the organization's strategic planning. Chin et al (6) conducted semistructured telephone interviews with representatives of CHCs in the Midwest that participated in the National Health Disparities Collaborative. The findings were similar to those of Wang et al (12). Major perceived successes included development of patient registries, improved diabetes care, improved adherence to standards, and increased patient and provider awareness. Building on these findings, in 2008 Chin et al (13) reported the results of a self-administered questionnaire completed by 1,006 participants at 165 health centers in 21 states. The barriers reported by respondents supported findings from the earlier study, including time, funding, and lack of staff.

We sought to identify strategies that contributed to the CHCs' successes and challenges encountered in the Massachusetts collaborative and to frame them in terms of the components of the Chronic Care Model. Our objective was to find specific, detailed strategies for improvement.

Of the 52 CHCs in Massachusetts, 17 participated in the collaborative after receiving information and an invitation from the Massachusetts League of Community Health Centers. The 17 represented CHCs from across the state and both rural (18%) and urban (82%) settings. Each CHC assembled a team to participate in the collaborative. At a minimum, team members consisted of a provider champion (a primary care provider); a team leader, responsible for coordinating and reporting activities; and a data entry person, responsible for establishing, maintaining, and updating the electronic registries of diabetes patients. Teams implemented the Chronic Care Model using the provider champion's panel of diabetic patients, called the population of focus. We contacted team leaders at the end of the initiative in May 2007 and interviewed them using a site visit interview tool developed for this project. The tool is based on the ecological systems theory (14), which recognizes nested environmental systems and the interactions within and between the systems. Members of the collaborative's evaluation team reviewed questions for appropriateness, relevance, and comprehension and revised questions accordingly (Appendix). Study protocols were approved by the University of Massachusetts Medical School Committee for the Protection of Human Subjects in Research.

Two experienced interviewers collected data from all 17 sites. They interviewed team leaders and other team members who were available at the time of the interview during a site visit (9 sites) or by telephone. Interviews lasted 30 to 45 minutes. Interviewers took notes and transcribed them after each interview.

We used Microsoft Excel (Microsoft Corporation, Redmond, Washington) to summarize the data and facilitate content analysis (15,16). The 6 components of the Chronic Care Model formed the initial framework for the coding scheme. Responses were coded independently by 2 investigators, and intercoder agreement was 96%. Disputed responses were reviewed until 100% agreement was reached. We developed a secondary coding scheme to identify emergent subthemes for each Chronic Care Model component by using answers to the original interview questions and spontaneous comments. Subtheme responses were considered dominant if they were reported by multiple CHC teams. Two investigators working independently conducted a second round of coding using the subtheme coding tool. Intercoder agreement was 88%. A third investigator reviewed disputed responses until 100% agreement was reached.

Participating CHCs uploaded monthly reports and data summaries to a password-protected Internet database. These reports included information about plan-do-study-act (PDSA) cycles, a quality improvement model. PDSA cycles are useful for rapidly testing a proposed change in procedure or practice by developing a plan to test the change (plan), testing the change (do), evaluating the tested change (study), and determining what adjustments are needed (act) (17). We audited these monthly reports and collected information pertaining to the number of PDSA cycles and the Chronic Care Model component they addressed.

Selected characteristics of patients in participating CHCs compared with those in the general state population (Table 1) suggest that CHCs serve a more ethnically diverse population (18) with a higher prevalence of uninsurance (19). Interviewees were team leaders from all 17 CHCs, 2 provider champions, and 7 team members. Each component of the Chronic Care Model was represented in the findings (Table 2).

This study confirms and expands findings from previous research (6,12,13) regarding the challenges encountered by CHCs. We identified several facilitators for achieving success in a health disparities collaborative: a paradigm shift to involve the patient in goal setting and decision making, learning sessions that deliver information about guidelines, IT support to develop and use patient registries, support from key leaders in the organizations, and linking with other CHCs to share resources. Teams noted that persistence was essential to achieve success across all areas of the model.

Teams concurred that CHCs may need to change their approach to health care delivery to increase the percentage of patients with documented self-management goals. Helping patients to become engaged in their own care, and have clear self-management goals, requires that the health care team understands the patient's priorities and concerns. Making self-management goal setting a team priority complements this change in approach because all members of the team must help patients to set realistic and achievable goals and provide encouragement and positive reinforcement when goals are met.

CHC staff consistently described their patient data registry as both a barrier to and a facilitator of success. Despite recognizing the benefits of having patient data to examine outcomes, track progress, and engage providers, almost all sites struggled with their patient registry. Chin et al (6) described similar findings. Teams said they needed training before starting and needed ongoing IT support during the course of the collaborative. For organizations considering convening a health disparities collaborative, recognizing this need and providing IT training and resources before implementation could improve progress toward collaborative goals. Ideally, staff at CHCs would then have more time to use data from the registry to tailor interventions for improvements rather than struggling during the first few months to populate their registries.

CHCs rarely made changes related to the community component of the Chronic Care Model (Figure). Two previous studies (10,20) explored the number and intensity of collaborative activities as measured by PDSA cycles. One study (20) reported findings similar to ours; the mean number of PDSA activities in the community component of the Chronic Care Model ranked lowest. In contrast, Grossman et al (10) reported a high number of PDSA activities related to community, but the activities were low intensity, as measured by the likelihood of improvement as an outcome of the intervention. Although it is unclear why connection with the community was the component with the least reported activity in our study, CHCs have a long history of community partnership and involvement. Thus, they may not attempt to form new partnerships because they are already engaged with their communities. Teams reported that time was a barrier to participation in the collaborative, and establishing new community relationships can be time-consuming. Moreover, engaging CHCs in community partnerships may be beyond the scope of work for team members and may need to be addressed by senior management. Future collaboratives may wish to encourage participating CHCs to re-establish community partnerships or conduct outreach to new organizations.

These findings must be considered in light of the limitations of the study. First, the interviews were conducted with CHCs in a single state, so the findings may not be generalizable to all states. However, the CHCs that participated in the collaborative represented diverse patient populations from both urban and rural locations. Second, CHCs that participated in the collaborative chose to do so and, therefore, may differ from nonparticipants in terms of resources, staff, and patients.

We used a rigorous qualitative method to describe and categorize the experiences of CHCs that participated in a health disparities collaborative. Our taxonomy, based on the Chronic Care Model, proved successful, as determined by intercoder agreement, and may be useful in qualitative measurement of other quality improvement efforts.