In response to the call of the Institute of Medicine for multisector partnerships (1), many new initiatives for population health improvement feature partnerships of leaders and organizations across multiple sectors of society. These partnerships typically include representatives and resources from various substantive issue areas — for example, education, economic development, transportation, agriculture, and health — and span the business, nonprofit, and governmental sectors. The purpose of this article is to review 1) the rationale for such partnerships as a tool for population health improvement, 2) key organizational and contextual factors that appear to be associated with effective multisector partnerships, and 3) the limited evidence regarding the effect of such partnerships on population health outcomes.

During the past 3 decades, efforts to improve population-wide health outcomes have moved toward community organizing and collaboration. Community organizing refers to the unit of analysis and action, shifting the focus from individuals to systems, rules, social norms, or laws to affect health behaviors and outcomes (2). This ecologic approach recognizes the connection between health and social institutions, surroundings, and social relationships (3).

Collaboration refers to the process of system change, shifting the focus from the responsibilities and effectiveness of individual institutions to their relationships and collective effect on population health. In particular, efforts have increased to involve many sectors of a community in pursuit of better health outcomes and the economic and social benefits thought to be associated with such outcomes. The rationale behind multisector partnerships is that, because no single organization or sector has full control over the determinants of population health, effective solutions require interorganizational coordination and collaboration (4). By pooling resources, talents, and strategies from a broad range of actors, each of these sectors can more effectively carry out its responsibilities as they affect population health (2). Researchers have advanced similar theories of collaboration to improve the effectiveness of initiatives on related issues such as poverty and community development.

Researchers have conceptualized partnerships for health improvement differently. Three dominant models of partnerships for health improvement have been described (4). In the first, public health agencies are primarily responsible for promoting activities and services that affect the health of the community. Their partnerships with other organizations exist primarily to extend the reach and capacity of governmental public health. In the second, many organizations play some role in promoting public health and so must be involved in health improvement. However, the focus remains primarily on the delivery of public health services. The third model focuses on the system of actors and actions that promote or threaten population health and includes activities in all sectors of community life (eg, education, business) (4). This last model, the most ecologic of the 3, has received increasing attention. However, the evidence to date suggests that these large-scale community health promotion projects have changed population health behaviors and outcomes only moderately (5).

In response to the mixed results of approaches based on the third model, some argue it is necessary to reconceptualize partnerships for health improvement (6). According to this argument, even the broadest partnerships have not shifted from an individual intervention paradigm to a true systems paradigm. Systems thinking focuses on the collective influence of a broad range of actors. It recognizes communities as networks of dynamic, nested relationships among individuals and organizations. These constantly evolving complex adaptive systems comprise diverse agents operating in various subsystems and suprasystems without centralized control (7). Although most partnerships adopt interventions targeting multiple levels within a system, they may fail to recognize the full scope and complexity of the system and miss opportunities to improve population health. Hawe and colleagues (6) argue that unique problems are associated with scaling up partnerships from the organizational level to the community level. They suggest that these partnerships learn from ecologic-systems perspectives that examine linkages, relationships, feedback loops, and interactions among systems. From this approach, multisector partnerships can be conceptualized as events within systems that either leave a lasting footprint or wash out, depending on how well the dynamic properties of the system are harnessed. The success of a partnership depends on activity settings, the social networks that connect people and settings, and time (6).

Recent work on social networking approaches to collaboration examines the importance of looking at the effect of a particular intervention rather than measuring the changes in a system over time. In network approaches, leaders focus not only on management challenges and opportunities at an organizational level but also on how to mobilize resources more broadly for the greatest social impact (8).

Drawing from these approaches, a fourth conceptualization of multisector partnership seems to emerge. This model focuses not only on the relationships among organizations in the partnership but also on the partnership's relationship to the context of the place it is trying to change. In some ways, this model is a continuation of the focus on neighborhood-based and community initiatives. However, it adds a new emphasis on considering the characteristics of context, including the timing of the intervention and past events, particularly earlier interventions that may have created networks. From this perspective, partnerships work to build capacity over time and consider the impact on the context itself as the primary outcome.

Extensive research has identified the qualities perceived as contributing to strong multisector partnerships in health and other issue areas. This section summarizes some of the lessons learned about the most important dimensions of partnerships.

Despite a common belief that multisector collaboration can improve population health, researchers seldom study the effect of such collaboration on population health outcomes. Evaluating the effect of multisector partnerships on population health outcomes is difficult. Some of the most-cited challenges are the short study period of evaluations, limited use of evidence-based logic models and theories of action to guide interventions, the difficulty of measuring the degree of individual exposure to interventions, and multiple or broad population indicators (21).

Researchers fail to agree on what factors are most closely linked to improved population health outcomes. Often these factors have been drawn from a broad review of literature from multiple disciplines, each defining efficacy differently (14). Even researchers who agree that a particular quality of a coalition is important may disagree about how to measure that quality (16).

In a review of hundreds of collaborations, Roussos and Fawcett (21) could identify only 34 evaluations of partnerships working locally to address community health that had a study design or logic model to guide their work. Of the 34 partnerships, 10 presented improved population-level outcomes that might be attributed to collaboration activities. The review found stronger support for the ability of collaborations to change behavior and systems. Of the 34 studies of partnerships, 15 included measures of behavior change, 14 of which indicated some shift in behavior. All 34 studies reported some sort of systems change in the form of new programs developed, funds generated, or other measures (21).

Another literature review (16) yielded similar results. The authors searched major databases for studies on partnerships that targeted local geographic areas to improve population-level health outcomes, and defined and measured both coalition effectiveness and coalition-building factors. The review noted that across studies, researchers have defined and operationalized coalition-building factors and effectiveness differently. Studies had different definitions of coalition functioning, often failed to connect coalition-building factors to coalition effectiveness, and yielded mixed results (16). One study concluded that multisector partnerships and interventions continue to be driven primarily by ideology and action rather than sound scientific design and evaluation (22).

Kreuter and Lezin (23) observe that justifications for collaborating to change health status and health systems fall into 2 major categories, conventional wisdom and evidence. Of the 2 justifications, conventional wisdom is vastly more common in the literature. The need for continued research and evaluation of broad-based initiatives to improve population health is clear, given the challenges of studying the influence of multisector partnerships in complex systems. Further research is needed to understand the circumstances in which formal multisector partnerships are likely to be formed, the extrinsic and intrinsic motivations of leaders and members, and how to increase the commitment of members through incentives and other means. In addition, further research is needed to identify whether and how multisector partnerships affect both the levels of population health and disparities within a population and to clarify what characteristics of partnerships and what contextual conditions are necessary for improved health outcomes. Finally, more research is needed to examine the comparative effectiveness of multisector partnerships and other strategies for improving population health, in particular, when the leadership and resources required to organize and maintain formal partnerships are not necessary to improve health outcomes or reduce health disparities.

General lessons are available: first, systems thinking is essential to organizing and sustaining efforts to improve population health, and to assessing their impact. The outcomes of partnership approaches depend on the social, economic, and political context of the community in which partnerships are formed and operate. Only by studying the varying contexts can researchers discern whether any form of partnership is sufficient for population health improvement.

Second, characteristics of partnerships — goals, sponsorship, membership, resources, leadership — do appear to matter, but this has been established primarily through studies based on perceptions of participants rather than objective measures of outcomes. Therefore, more research is needed on multisector partnership outcomes using longitudinal and prospective designs that include measurement of activities, social network development, and types of organizations involved and resources engaged. To aid this area of inquiry, better and more widely adopted measures of structure, process, and outcomes are needed to link partnership formation to community-wide impact. One step toward building a stronger evidence base of what works would be the adoption of common models or frameworks for defining different forms of public health partnerships — for example, the typology offered by Mays (4). Standard models, as well as more standard measures of partnership contexts, characteristics, and strategies, would improve the generalizability and replicability of research and accelerate learning.

Third, multisector partnerships almost certainly offer some value to individuals and communities apart from any direct and measurable effect on population health. The shared effort and communication that result from a health initiative may highlight problems, shift resources, or raise expectations for participation and performance in other areas of community life. Studies of multisector health partnerships should be alert to such catalytic changes and spillover effects as researchers pursue a clearer view of the connections between partnerships and population health improvement.