Poor health status, rapidly escalating health care costs, and seemingly little association between investments in health care and health outcomes have prompted a call for a "pay-for-performance" system to improve population health (1). The goal is to link structure and process to outcomes in the health system, which is the set of institutions and actors that affect people's health, such as organizations that deliver care, health plans, educational systems, and city and county governments. Linking these organizations will contribute to the control of health care costs, improve the health of the US population relative to the health of other developed nations (2), and reduce disparities by region, race, ethnicity, and educational attainment (3).

The lack of tools to measure the effect of clinical services on US population health is rooted in the historical development of the American clinical health care system, which evolved to respond to the acute care needs of the individual: relief of pain and suffering through diagnosis, therapeutic intervention, and reassurance (4). Responsibility for population health needs was with the public health sector alone, and the effect on health of social policies related to education, work, transportation, and other factors was neglected. The Centers for Disease Control and Prevention might be considered the national population health agency, and many state health agencies monitor population health, but these agencies do not have regulatory authority over the health care delivery system. Many local public health agencies are mostly safety net providers. Notions of accountability for population health are underdeveloped at all levels.

Although clinical care accounts for only a small portion of the population health determinants (5), clinical service providers and health plans can contribute to population health initiatives by promoting healthy behaviors and providing clinical preventive services. At a population level, the behaviors that most powerfully affect health are physical inactivity, unhealthy diets, tobacco use, and excessive alcohol consumption (6,7). These behaviors can shorten life expectancy by 10 or more years (8,9). Behavioral support, when delivered with sufficient intensity in settings such as work sites, increases people's odds of adopting and maintaining a healthy lifestyle (10,11). Behavioral and social support is necessary to increase the prevalence of healthy lifestyles because, even when presented with the opportunity to adopt a healthy lifestyle, people still must choose a healthy lifestyle. They are unlikely to do so in a physical and social environment that encourages poor health habits.

Properly selected clinical preventive services also improve population health (12). People are more likely to receive appropriate preventive services when quality assessment systems ensure that they are informed about the benefits of the services and invited to accept the services.

Clinical indicators can identify gaps in access to care — an indicator of quality — and guide the application of incentives to close the gaps. Reporting clinical indicators of population health may also increase the salience of health incentive programs to stakeholders such as clinicians or purchasers of health services, who might be more focused on clinical performance than on long-term mortality trends. The level of clinical indicators can change more rapidly than death rates and longevity, and thus, may give more immediate feedback about the effectiveness of intervention programs. For example, feedback can be provided about positive changes in smoking rates and physical activity rates long before the effect on mortality can be observed.

A list of access and quality indicators related to population health cannot be developed without asking what intervention strategies will improve population health. We believe that 4 clinical care system strategies are strong candidates. The first is to increase rates of healthy behaviors and the delivery of preventive services in traditional settings of health services delivery. The second is to support healthy lifestyles and increase access to health care by extending the clinical setting beyond the doctor's office, for example, by providing wellness and health promotion services in work sites, the dominant social environment in the United States. The third is to develop a system by which clinical care organizations collaborate among themselves to coordinate care and reduce the illnesses and deaths that result from poor communication (13,14). The fourth is to offer clinical service providers and health plans incentives to participate with other sectors in the community (eg, education, transportation, housing, food) to develop policies and programs to improve population health. We describe indicators that are available to promote the first strategy and suggest a set of indicators that could be developed to promote the other strategies.

Although nearly 3 dozen indicators are considered current and valid by the Institute of Medicine's Committee on the State of the USA Health Indicators (15), nearly all of them are limited in the health domains they assess or the populations they cover. For example, the Healthcare Effectiveness Data and Information Set (HEDIS) indicators, produced by the National Committee for Quality Assurance (NCQA), report the performance of participating health plans (16). In 2008, however, health plans covered only half of Americans, and only half of all health plans reported these indicators. Another organization, the Institute for Healthcare Improvement, has proposed mortality and whole system indicators for health care systems, but these are yet to be implemented. A third example is Minnesota HealthScores, a community-wide program that includes nearly all Minnesota payers. It uses HEDIS and composite indicators of quality of care that bundle many aspects of care performance by condition (17). However, Minnesota HealthScores reports quality of care indicators only for depression, diabetes, and vascular disease, and only for Minnesota. With few exceptions the valid indicators that apply to the entire US population are collected and reported only by the federal government. A notable exception is the Commonwealth Fund, which reports health care system performance and offers international comparisons for some indicators (18).

To inform discussion by policy makers and the public about population health, the State of the USA Project commissioned the Institute of Medicine in 2008 to convene an expert committee to recommend 20 county-level indicators of the health of the United States (15). The committee selected indicators with attention to the availability of data that could be used to report rates at the county level and to make comparisons with other countries. Six indicators of health behavior — smoking, physical activity, excessive drinking, nutrition, obesity, and condom use by sexually active youth in grades 9 through 12 — were selected. Another 6 indicators were selected to characterize the health care systems: health care expenditures; insurance coverage; unmet medical, dental, and prescription drug needs; preventive services; childhood vaccination; and preventable hospitalizations.

We recommend tracking 5 of the 6 health behaviors in a pay-for-population health initiative (Table 1). Although tracking body mass index as an intermediate outcome is useful, obesity is not a behavior per se; therefore, unlike the State of the USA report, we have not included it in our list. Regarding the indicators of health care system performance, we recommend tracking the 6 indicators recommended in the State of the USA report (15) (Table 2). To draw attention to the economic burden of health care and disparities in access to care, we also recommend 2 indicators that can be calculated from the indicators in the State of the USA report and federal data on per capita income, race, and ethnicity.

We suggest that 3 indicators of access and quality be developed for paying health plans or others to improve population health: provision of wellness and health promotion programs, participation in accountable care organizations, and participation in initiatives to benefit communities. Data and, in some cases, methods are not yet available to characterize counties with these indicators.

If a pay-for-population health initiative is to be implemented, criteria for most of our proposed indicators exist but they must be developed for 3 others. More difficult questions to answer are who would pay for the services and what organizations would be eligible to provide the services. The answer to the latter question is fairly clear for clinical preventive services, but it is less clear for work site wellness and health promotion programs. Would only health plans be eligible, or would any company that offered NCQA-accredited wellness and health promotion services be eligible? Would group purchasing allow small employers to offer wellness and health promotion programs? How would community-benefit initiatives that address the leading social determinants of health be evaluated? How would the contributions of the participating organizations be parsed? Might the agency purchasing population health write a performance contract? These questions can be answered only through an iterative process of negotiation among employers, purveyors of health promotion programs, health plans, communities, and the other stakeholders.

A fundamental requirement for any pay-for-population health initiative is performance data. Ideally, these data would be available to make comparisons at the county level, but the data are not available for some indicators. As part of the pay-for-population health system development, the appropriate federal agencies should be encouraged to collect data that can be reported at the county level.

Data are available to measure health care access and quality as reflected by 5 health behavior indicators and 8 health care indicators. Most of the data are collected by federal agencies and are available yearly at the state or county level. Additional indicators that should be developed include whether employees have access to wellness and health promotion services through the work site, whether a health care organization is a member of an ACO, and whether a health plan collaborates in community-benefit initiatives that address the leading social determinants of health. Data for these indicators should be collected at the county level by appropriate federal agencies.