Population Health Outcomes

Is Population Health Finally Coming into Its Own?

Why am I starting a blog on issues and perspectives in the field of population health?  It is because I believe we are at a “tipping point” in which a convergence of scholarship, policy, and practice initiatives seem poised to bring an overdue population health perspective to thinking about and acting on health and health improvement. The challenge of the next decade will be to find practical ways in which new approaches to financial and non-financial incentives and multi-sectoral partnerships can be applied to improve population health outcomes here and elsewhere.

Although the roots of broad population health thinking go back for centuries, in the United States, the second half of the 20th century was dominated by the rise of biomedical science and clinical medicine. Many in public and population health (how are these different?) have been critics of the dominance of health care emphasis and investment, but these advances have been critical and will continue to be so. The end of the century has seen additional growing understanding and support that such behaviors as tobacco use, diet, and exercise also make substantial contributions to the health of individuals and populations. My own population health epiphany came with my exposure in the early 1990s to the thinking of Bob Evans and Greg Stoddart and their population health “field model” from the Canadian Institutes of Advanced Research (1). At once simple and profound, it captured the basic population health principles:

• that health outcomes were more than the absence of disease;
• that these outcomes were produced by complex interactions of multiple determinants (health care, behaviors, genetics, the social environment, the physical environment); and
• that in a resource-limited world, the relative cost effectiveness of these determinants was critical for policymakers.

While this Canadian work has provided an important framework for international population health scholarship and policy development, the last 15 years in the U.S. have been dominated by such issues as health care access and costs and pressing immediate public health issues like emerging communicable diseases (e.g., HIV-AIDS and H1N1). Periodic activity and reports from government agencies and foundations have addressed parts of the broad issue of outcomes and determinants, but not in visible and sustainable ways. Notable exceptions have been (A) the Institute of Medicine (IOM) and Centers for Disease Control and Prevention’s (CDC) emphasis on a broad public health “system,” (B) the creation of the Robert Wood Johnson Foundation’s (RWJF) Health and Society Scholars Program to grow a new generation of multidisciplinary scholars for this emerging field, (C) the RWJF Commission to Build a Healthier America, and (D) the WHO Europe’s 2003 report on the social determinants of health. (I will return in a future post to the challenge the social determinants of health pose for policymakers.) In 2010, however, this emerging field has still not matured in either scholarship or policy. However, the last several years have seen growing attention from a policy perspective. Here are some examples:

• Decades of work from the Dartmouth group demonstrated how communities that spend the most on health care may not have the highest quality or outcomes.
• The groundbreaking documentary Unnatural Causes provided wide exposure to the multiple determinants of health.
• The California Endowment is making a major long term investment in the broad health of 10 communities in California.
• The State of Minnesota has invested $43M in its State Health Improvement Plan to create “accountable health communities” that address obesity and tobacco through policy, systems, and environmental changes.
• The Institute for Health Improvement has endorsed population health improvement as one of the three legs of its Triple Aim strategy and is considering moving to a regional strategy
• CDC has funded a cooperative agreement with the National Business Coalition on Health to work with business-led health coalitions and the business sector on building the capacity of members to be leaders of health reform and advance value based health and health care.
• Concerns about obesity have underscored the complexity of addressing this critical issue (look for a future post on the tension between free will and the social context with regard to behavior choices).
• Obama administration officials have been discussing a “place-based” approach for social investments as well as a major Community Health Data Initiative.
• The National Quality Forum and the National Priorities Partnership are discussing the need for new broad measurement tools such as a national index of health.
• Our own Robert Wood Johnson Foundation MATCH project in February released the first ever National County Health Rankings of 3,014 counties in all 50 states according to a broad population health model that has been used in Wisconsin since 2003.

When I published Purchasing Population Health: Paying for Results in 1996, this list of activities was much shorter. Now every day we learn of some new policy, practice, or research finding that is aimed at population health improvement. Keeping track of and commenting on these exciting developments is the purpose of this blog. We plan to post on a weekly basis, with content from our own team as well as guest posts from prominent policy and practice leaders.The University of Wisconsin Robert Wood Johnson Health and Society Scholars Program will each month identify several recently published papers that make important contributions to policy, practice, or scholarship. All posts will in some way relate to our underlying population health model, including key concepts of outcomes and factors.

Reference:

1. Evans R. Stoddart G. (1990). Producing health, consuming health care. Social Science and Medicine, 31(12), 1347-63.

Assessing Today’s Health… and Tomorrow’s

Today I’m starting a new series on population health measurement, motivated by the idea that “you can’t manage what you can’t measure.” While there is no shortage of measures and rankings, there’s actually little consensus on what metrics are best suited to assess progress in population health improvement. Last week I urged Triple Aim architects to clarify how population health is defined in their model, but they’re not the only ones with definitional issues.

Metrics can assess a single item, such as infant mortality rates, or be bundled into groups to create summary measures. America’s Health Rankings (AHR) and the County Health Rankings (CHR) both use summary measures to annually rank, respectively, the health of states and counties within states. AHR has been ranking the health of states since 1990 and released its most recent rankings in December 2010. This important national resource informed work by the University of Wisconsin Population Health Institute to rank the health of Wisconsin counties beginning in 2003 and, for the first time in 2010, counties in every state. Dr. Pat Remington and I are proud to serve on AHR’s Scientific Advisory Committee.

The AHR and CHR share a common goal of population health improvement and both projects take pride in their deliberate, systematic, and transparent approaches. Both projects create summary measures for health outcomes (using metrics such as premature death, quality of life, and poor birth outcomes) and health factors or determinants (using metrics that capture health behaviors, clinical care, and social/economic/physical/policy environments). However, while CHR uses the health outcomes summary measure to identify the healthiest counties, AHR creates its overall rankings based on both outcomes and determinants.

So what does this look like in action? From 2009 to 2010, Wisconsin dropped from 11 to 18 in AHR overall rankings. But, a look at separate outcomes and determinants rankings over the same time period tells a slightly different story. Our determinants ranking is identical to the overall ranking (we dropped from 11 to 18). However, our health outcomes ranking improved slightly, from 16 to 15.

Why is this important? In general, we think of outcomes as a reflection of our current health and determinants as a predictor of our future health. The two summary measures often move in the same direction, but not always. When we launched the national CHR in 2010 with support from the Robert Wood Johnson Foundation, some wanted to know why our rankings didn’t have an overall score like AHR. The answer has to do with how the results are communicated and their implications for action. We felt strongly in the value of the two different measures to communicate different but equally important concepts about population health, particularly for policy makers.

Places with determinants better than outcomes are on the right track toward health improvement. But, for those such as Wisconsin with declining determinants, unless we act to reverse the trend, there is cause for concern about potential declines in health outcomes. It is the outcomes we are trying to improve, and the determinants that will get us there.

Which Outcomes Should We Improve?’

recently argued on the importance of keeping population health outcomes and determinants separate, since the former represents today’s health and the latter tomorrows’ health. This week and next, I’ll be digging a bit deeper to look first at outcomes and then later at determinants.

Health outcomes have been defined as “all the possible results that may stem from exposure to a causal factor from preventive or therapeutic interventions.” As Gib Parrish noted in his 2010 award-winning article, there are many different ways to measure health outcomes. These include:

• Life expectancy from birth
• Age-adjusted or age –specific mortality rates
• Condition-specific changes in life expectancy and mortality rates
• Self-reports such as general level of health

Parrish notes that outcome metrics should present both the overall level of health of a population and the distribution of health among different geographic, economic, and demographic groups in the population.

The MATCH population health model underpinning this blog reflects much of this perspective. The 2X2 outcome diagram below includes both mortality and non-mortality (i.e., health related quality of life) components, as well as both population mean and population disparity metrics. This makes sense conceptually, but poses issues and choices in practice.

Of course, it is possible to separately track outcome measures in these four (or more) spaces. This is what many health planning exercises do, such as the recently released Healthy People 2020 (HP2020). In this important national process, a large number of individual measures of general health status as well as health related quality of life and health disparities will be monitored and reported on periodically.

One HP2020 indicator is “Healthy Life Expectancy,” which will be assessed through three distinct measures:

• expected years of life in good or better health
• expected years of life free of limitation of activity
• expected years of life free of selected chronic diseases

These three metrics are “summary” health outcome measures, combining mortality and health related quality of life together. In a similar approach, our County Health Rankings’ summary outcome measure assigns 50% to years of potential life lost before age 75 and 50% to 4 non-mortality measures. These weights reflect our interpretation and judgement of the relative importance of the components, but we recognize that not everyone would agree with our choices. There is no empiric “right” answer to the question “Which Outcomes Should We Improve?”. This is a question of values; decisions weighting length of life versus quality of life should be made by individuals and communities.

Why does this matter? Different outcome choices require different patterns of investment in the determinants of such health outcomes. I’d like to see a web-based tool that allows communities to explore a broad set of outcomes and that leads them through a process of selecting outcome priorities that have local meaning and relevance. Having a clear sense of place-based outcome goals should help establish policy priorities and guide resource allocation.

As readers of this blog know, we don’t yet precisely know which programs and policies are most cost effective for overall population heath improvement – let alone those that reduce disparities (stay tuned for more soon on this topic). But we are hopelessly lost if we aren’t clear on where we are going, and what our targets are. Clear and transparent choices about outcomes have tremendous promise, serving as a sort of population health compass to guide us step by step toward a healthier tomorrow.

Is Chronic Disease Burden a Population Health Outcome?

Responding to last week’s post on which outcomes we should improve, Matt Stiefel from Kaiser Permanente asked about using the incidence or prevalence of chronic illness as a population health outcome measure. Matt’s important question prompts me to continue the conversation this week.

The Institute of Medicine’s 2008 State of the USA Health (SUSA) Indicators report (I served on this committee), called for an index of chronic disease prevalence among its seven other health outcome measures that “reflect the overall health of the nation and the efficiency and efficacy of U.S. health systems” such as life expectancy at birth, infant mortality, and unhealthy days (physical and mental). To my knowledge, such an index has not yet been developed. As a member of the IOM SUSA committee, I am concerned with the classification of chronic disease prevalence (and other similar factors) as outcomes per se, because improving these factors is the means, not the ultimate end that we seek (i.e., living longer healthier lives). This opens the door for the means to become the end.One strategy might be to add a third category of “intermediate outcomes” to our model of outcomes and determinants/factors. These “intermediate outcomes” could capture those factors (like smoking rates or primary medical care or burden of chronic disease) that, if improved, are likely to directly and significantly improve our ultimate outcomes over time. (Others call these “proximal determinants” in contrast to the more “distal determinants” such as income and education.) This category of “intermediate outcomes” could be very useful for guiding population health monitoring efforts and policy priorities, and indeed may be required when we want to track or reward short term improvement.

We don’t have to look far to see the potential value of this third category. The National Research Council’s recent report highlights diverging paths in longevity among high-income countries. The report points to smoking rates as one main reason why life expectancy in the U.S. is increasing at a slower rate than might be expected. Thirty years ago, smoking rates were much higher among U.S. adults than they are today (37% vs. 21%); what we’re seeing now is the time-delayed ripple of these “intermediate outcomes” or determinants affecting long-term health outcomes.

Thinking about smoking rates as an “intermediate outcome” can be helpful from a policy perspective as well. Later this month, New York City’s Mayor Bloomberg is expected to sign a bill that will ban smoking in 1,700 city parks and along 14 miles of beaches. According to the New York Times, the new policy will represent the “most significant expansion of antismoking laws” since Mayor Bloomberg’s 2002 push to prohibit smoking in bars and restaurants. I hope the National Research Council will be able to credit these and similar policies to declining mortality in 2040.

So, while I’d like to reserve the population health term “outcomes” for our ultimate goals of increasing the length and quality of lives, a third category of “intermediate outcomes” deserves greater attention by both policymakers and scholars. Developing consensus around which indicators would best fit into such a category – as well as what programs and policies could most cost-effectively improve these numbers – would be a great place to start.

Doing Well or Doing Better

Our focus for the past couple of weeks has been on outcomes, both long-term and intermediate. Most of the examples mentioned have been discussed in the context of absolute achievement, such as being #1 on state or county rankings, or achieving the lowest infant mortality rate among all nations.

But is being the best the only thing we want to measure and reward? What about communities and states that are improving most rapidly, particularly from poor baselines? This issue is front and center in education policy: should we reward or incentivize schools or teachers who have the best student test scores, or those demonstrating greatest improvement.

I think both approaches are useful. Overall achievement should be measured and recognized, as with Olympic medals. But there’s also a case to be made for keeping track of improvement, particularly for purposes of allocating resources to communities with formidable social and economic challenges.

From a statistical perspective, measuring improvement is even more difficult than measuring achievement. For our County Health Rankings, we average three consecutive years of data to assess premature death as Years of Potential Life Lost (YPLL) before age 75 while for other measures, such as smoking rates, we use up to seven years of data. Using multiple years of data allows us to measure health in nearly every county in the US. But the need to use multiple years of data to get an estimate for a single point in time makes it difficult to measure change over time, particularly for smaller communities that have greater variability in their measures.

We should not let these analytic obstacles deter us from incentivizing doing better at the same time we honor doing well. On the data side, we need to oversample rural areas and seek out non-traditional data sources such as healthcare, employer and school records (with adequate individual privacy protection of course). For measuring improvement, we should focus attention on shorter-term metrics (such as rates of infant low birth weight) and closely track indicators that are strongly correlated with health outcomes (such as burden of chronic disease, income, and education). And we should experiment with both quantitative and qualitative metrics for assessing innovative and emerging intervention strategies.

Since no community has achieved the highest health outcomes that it can, each can aspire to do better and even be the best. The National Priorities Partnership of the National Quality Forum has set one goal that “the health of American communities will be improved according to a national index of health.” But to reduce inequalities across communities (a topic I’ll return to soon), those with worse outcomes need to improve at a faster rate to reduce the gap. As with our schools, learning what combination of metrics, mechanisms, and incentives maximize rates of community health improvement will be an important part of population health policy strategy in the coming decades.

Which Health Disparities Do We Want to Reduce?

Over the past few weeks, I have been blogging on population health measurement, under the headline that you can’t manage what you can’t measure. While I’ve given quite a bit of attention to overall outcomes, I haven’t directly addressed disparities.

Improving the length and/or quality of life overall doesn’t necessarily translate to improved health for all population groups. Policies and programs can have a differential affect on subpopulations that exacerbate rather than diminish disparities.

Deciding which outcomes to target to improve overall health involves value judgments – whether the goal is improving overall health or reducing disparities. There is not a “correct” formula for simultaneously improving overall health and reducing disparities. The 50:50 weighting suggested by the figure below represents just one possible approach but the ethical dimensions of such choices cannot be ignored; in a resource limited world, should we focus on raising the bar overall or narrowing the gaps?This is a value choice to be made by individuals and groups in light of their local, state or national situations and perspectives.

As the figure shows, disparities in outcomes exist across a spectrum of categories. These include race and ethnicity but also social and economic status, gender, and geography (and other dimensions as well). However, people often focus on disparities primarily in the race/ethnicity dimension. Last month the CDC released its first ever report on disparities and inequalities. Despite the report’s emphasis on the social determinants of health (the first section focuses on income and education), media coverage focused on the findings about racial disparities.

Here in Wisconsin, our ethnic minority populations are relatively small and geographically concentrated, but every county in the state has substantial health disparities by education level and income. In fact, the state earned a “C-” overall for health disparities in its 2010 Health of Wisconsin Report Card (which employs an innovative multidomain disparity index). The figure below from this report shows the distribution of death rates for various population segments both above and below the state average. It illustrates the point that racial disparities are significant but that there are also other critical differences across other population segments, such as those with different levels of education.

Chronic racism has been shown to have a biologic effect on health through stress pathways, and institutional racism has created many of the disparities in income, education, and health services that influence health outcomes. We clearly need to develop and support policies to address and eliminate racism on every level in our communities.

But focusing our attention too narrowly may make us think of obstacles instead of opportunities. To improve population health, we need to think beyond race to recognize the many factors that affect health differentially. In a country still battling racial stereotypes and prejudice, this broader perspective on disparities may help create a unified and constructive approach to addressing our collective health challenges.

Bending Health Disparity Curves