Let’s begin at the end—the final sentence of “Understanding Population Health Terminology,” a paper published by one of us, Kindig, in 2007: “The overriding population health question is, what is the optimal balance of investments (e.g., dollars, time, policies) in the multiple determinants of health (e.g., behavior, environment, socioeconomic status, medical care) over the life course that will maximize overall health outcomes and minimize health inequities at the population level? This is a significant challenge that will require decades of attention by scholars and policymakers.”

This idea derives from the 1990 Evans-Stoddart population field model that has been the intellectual underpinning of our field for decades. This paper and its final figure, exhibit 1, shows the evolution from the medical model shown in the health care and disease boxes on the right to the broader model with expanded concepts of outcomes and the addition of multiple determinants of health.

Exhibit 1: A Population Health Field Model

Source: Evans RG, Stoddart GL.. Producing health, consuming health care. In: What are some people healthy and others not? New York (NY): Routledge; 1994. p. 27-66.

This is certainly a complex model that one of its creators later called a “Fantasy Equation,” saying that “at present we but vaguely understand the relative magnitude of the coefficients on the independent variables that would inform specific policies rather than broad directions, even if we are beginning to see the variables themselves more clearly.” Robert Evans and Greg Stoddart again noted in 2003 that “most students of population health cannot confidently answer with precision the question: ‘Well, where would you put the money?’” That has not stopped us from calling for its solution over the past 25 years here and here but with little to show for it.

One of us, Kindig, presented this conundrum to a group of students during a guest lecture for the other’s, Mullahy’s, “Introduction to Population Health” course. At this point, Kindig asked: “How can this be? This can’t be as difficult as all the modelling and equations needed for landing on the moon, can it?”

Here are the answers we came up with on the class whiteboard.

It Is Harder

This is in the realm of social science, not physics and engineering. Causality is difficult to conceptualize and, even if well conceptualized, demonstrate empirically.

There Are Multiple Outcomes

With the model’s important expansion beyond disease into health and function and even well-being, the number of outcomes explode: overall mortality, morbidity, health-related quality of life, as well as disparities and inequities in all of them. Summary measures while sometimes useful add the complexity of weighting components. This seemingly constant instability led one student to question if the “Fantasy Equation” does exist, is it only applicable in a steady state, where variables of the systems or the process are unchanging in time? Since we live in a dynamic state, such a fixed solution to the “Fantasy Equation” likely does not exist and even if it did, it may not be applicable in a decade or two.

There Are Multiple Units Of Analysis

Another outstanding question is: Which population? Which is of primary interest and clinical or social policy relevance: individuals, communities, nations, the world, marginalized groups, separately or all together?

Many, Many Complex Empirical Issues

To speak of “a solution” to the fantasy equation is itself a fantasy. Its essential nature is that of a complex set of cause-effect relationships. For data to shed light on these relationships, not only must the specific causes and outcomes have clear definitions, but those definitions must find empirical counterparts in available data. What follows, therefore, is a litany of additional questions:

• What are the individual and/or population health metrics of interest?
• What are the specific determinants that are amenable to manipulation by policy interventions? (A reminder that, as is sometimes claimed in the causality literature, there is “No causation without manipulation.”)
• What imaginable policies can be designed or modified to bring about such manipulation?
• With what time lags do determinants and policies have their effects?

The empirical task at hand is hardly simplified when one recognizes that confounding and interactions among the determinants and among the policies at a point in time and over time are almost certainly of fundamental importance. Even if such interactions could be characterized conceptually, learning about them from existing data would be a formidable task.

Another student suggested that the “fantasy equation” is too complex, too fluid, and rife with too many unknowns, to ever be solved. Outside forces and tradeoffs add additional layers of complexity such that changing any one variable or coefficient will alter so many other variables that affect the downstream outcomes. 

Data Limitations 

We can only examine what we have data about. We know a lot about Medicare since it is a massive program in the public sector. Data on other determinants are more limited, and some issues such as gun violence can’t be fully understood because of political restrictions. Moreover, in the spirit of privacy protection various statistical agencies—such as the Census Bureau—are increasingly creating obstacles for researchers to access individual-level data.

At the end of the discussion, the majority of students agreed that the moon landing was much less complex.

Where Does This Leave Us?

One of the students asked, “How long do we weigh the pros and cons and argue about how much to invest and where? How long does an idea ruminate in a think tank before it becomes relevant to the very people it aims to help?”

We refuse to accept a policy scenario in which investment decisions are set on guesses, hunches, political whims, or opinions. New data sets and analytic approaches should yield more precision, and those efforts could potentially have impact worthy of a Nobel Prize in medicine or economics.

Despite the slow progress, we are asking the optimal-balance-of-investments question more often, and answers are beginning to emerge. New disciplines are attacking the problem from a systems science perspective. Bobby Milstein and colleagues, for example, asked “Which priorities for health and well?being stand out after accounting for tangled threats and costs?” and found that “poverty reduction and social support were the most highly ranked interventions for all outcomes in all counties. Interventions affecting smoking, addiction, routine care, health insurance, violent crime, and youth education also were important contributors to some outcomes.”

Following this class, we reached out to Gregory Stoddart and invited him to join us in writing this piece. He declined, citing his satisfying retirement from McMaster University, but did send this email message: “Although as you know I think that the fantasy equation may be unsolvable, that doesn’t mean that we don’t know which directions to reallocate resources. The concept of marginal returns can and should guide us here, even in rough orders of magnitude. We don’t require precision to help more people be healthy or to be more equitable.”

In other words, robust estimates of directions and orders of magnitude may be just as important in serving decision makers as precise but unreliable findings. In a clinical research context, John Mullahy and colleagues described this challenge this way: “If the considerable investment in turning discovery into health is to pay off, then understanding when research efforts do and do not yield complete discovery is essential. When research falls short of yielding complete discovery, the fact that it may partially identify magnitudes of interest should be celebrated, not bemoaned.”

With all that said, there remains an equally pressing step in solving “the” fantasy equation—whether fully or partially. That is to investigate what kinds of information about these cause-effect relationships are actually useful to know. A valuable concrete step in this direction would be to engage real-world decision makers to learn what sort of information about population health causes and effects would be most valuable in shaping policy and practice.

George Box wrote famously that “all models are wrong, but some are useful.” The task at hand is to determine decision makers’ willingness to trade off “right” for “useful.” We speculate that many will tolerate a reasonable degree of imprecision. Knowing this should usefully guide the next generation of population health research about the fantasy equation.

Authors’ Note

We appreciate the contributions of students from the fall 2021 class “Introduction to Population Health” PHS 795 University of Wisconsin Madison School of Medicine and Public Health.