A Public Health of Consequence: Review of the June 2016 Issue of AJPH

This month, in our invited editorial, Westreich et al.¹ provide an important illustration of the implications that a public health of consequence lens has for causal thinking in the population health sciences. Centrally, Westreich et al. make two points. First, they note that thinking about the consequences of our work for population health elevates the importance of external validity. Simply put, our work is unlikely to be consequential if we cannot use it to guide inference about the populations whose health we aim to improve. This point may seem simple at face value, but it both challenges conventional wisdom in the field² and upends our currently established hierarchy of study designs. As Westreich et al. and others continue to note, randomized clinical trials (RCTs) depend on stringent inclusion criteria, aimed at improving these studies’ internal validity. This, however, frequently limits the extent to which RCTs are representative of broader populations, limiting the inferences we can extend from the findings of these studies to these same populations. This observation is readily borne out by the abundant examples of RCT results³ coming in conflict with findings from equally well-done observational studies.⁴ This challenges our notion that RCTs represent a gold standard. While they might represent an outstanding study design to improve internal validity, they are less well suited to external validity and generalizable inference and may well be less informative for the goals of a public health of consequence. This is not intended to obviate the role of RCTs in population health research. It does, however, serve as a sobering reminder of the balance we must strike between internal validity and external validity concerns, and of how a public health of consequence lens helps us evaluate the scholarship that is of highest priority, toward the goal of informing efforts that aim to improve the health of populations. Second, Westreich et al. make the case to treat population intervention effects as a means to making population health science findings more accessible.⁵ This provides an equally important reminder of the need for accessible analytic representations that help us realize the promise of translation of population health scholarship.

With Westreich et al’s words as framework, we highlight two articles in this month's AJPH that address the two central points made in this framework; first we discuss an article that uses a population-based sample to conduct informative science that can guide population-level intervention, and second, an article that presents interesting modeling data, accessibly, in a way that can plausibly inform, inflect, and influence an important public conversation.

First, Keyes et al.⁶ demonstrate a link between early adolescent use of tobacco and marijuana and cocaine use by 12th grade. They suggest that each percentage-point decrease in the prevalence of smoking in the 8th and 10th grade is associated with an 8% decrease in the prevalence of later marijuana use and a 14% to 23% decrease in later cocaine use. The Monitoring the Future data are nationally representative data, and the sheer number of people to whom this applies puts the importance of these findings in context. Nearly 10 million adolescents will have tried marijuana by the 12th grade, and one million will have tried cocaine. Importantly therefore, this article suggests a logical way forward for any effort by public health to reduce marijuana and cocaine use by millions later in life. As the authors note,

[P]ublic health campaigns to reduce the burden of drug use among adolescents should focus on the early stages of adolescence when drug use habits are forming, and that prevention of cigarette smoking, and use of tobacco products more generally, may be a crucial component of a public health strategy.^6(p1148)

This article provides us with an elegant example of how a population-based sample can illuminate the tremendous potential of interventions that target ubiquitous factors (like early adolescent smoking) toward the creation of healthier populations. It seems to us that the logical next step in this thinking would be inquiry about the most effective population-based approaches that successfully reduce smoking in these target populations, and that do so without widening intergroup differences and introducing health inequities. As a start, Cobb et al.⁷ present early evidence of the potential utility of a Facebook-based smoking cessation intervention, showing that those initially enrolled in an online smoking cessation program can act as viral agents and enroll others to participate at no additional cost. Whether this is effective in reducing smoking among adolescents remains to be seen.

Second, Tsao et al.⁸ make an excellent contribution to the literature through an article that models the potential impact of a $15 per hour minimum wage on preventable mortality in the New York City population. Their analyses suggest that this minimum wage could have prevented 2800 to 5500 premature deaths in New York City, principally among populations of color living in low-income neighborhoods. Any such modeling exercise must always be approached with caution, recognizing that findings from this work are only as robust as the assumptions that were used to inform and parametrize the model; to this end Tsao et al. do an admirable job of presenting, comprehensively, the limitations to their work. However, notwithstanding these limitations, their presentation of these findings, as the authors note, “adds to a growing body of work by health departments to resurrect the centrality of minimum wages to population health.”^8(p1039) Importantly, it both does so and presents important data that are accessible to the nonspecialist and can contribute meaningfully to the public discussion about the minimum wage. As this article was going to press, Governor Cuomo had proposed a $15 per hour minimum wage in New York State.⁹ Further analyses demonstrating how this may influence premature death in the whole state would be welcome.

It is in part the premise behind this monthly commentary that the lenses we adopt to inform our work are important: they shape how we think, the questions we ask, the studies we design, and how we analyze our data. An approach to population health science that prioritizes the potential public health consequences of our work will definitionally ask questions that matter to populations, draw inference that can inform how we might intervene to improve these populations’ health, and present results in such a way as to inform those who can make change happen, locally and globally. These articles in this month’s issue provide a compelling illustration of these principles.