Visual representation is a big deal in public health. How we show what we know shapes how public health is known.
Geoffrey Rose wrote an article about the difference between causes of cases and causes of incidence that is de rigeur in public health.
Rose’s article is often—and justifiably—remembered for its appeal to population-based preventive interventions. But the lessons of Rose’s insights go further, even if he didn’t take them there himself.
The article gets straight to the point, identifying the common normative error in public health: “What is common is all right, we presume.” If everyone in a population is exposed to high levels of lead, then the entire population will have high rates of violence and poor school outcomes, and it will seem as if these scourges are just normal life.
Rose doesn’t emphasize it, but of course even worse interpretations are possible. Many people have succumbed to the temptation to assume that these scourges are somehow inherent in the population itself. As if perhaps high crime rates among African-Americans are an innate feature of personality, rather than a bug that has infected it from the outside.
Rose’s insight that what is common is all right, we presume, could be modified to suggest that what is common within a sub-population is natural for that sub-population, we presume.
The brilliance of this article is in the distinction it draws between the causes of cases and the causes of incidence. “I find it increasingly useful to distinguish two kinds of ætiological question. The first seeks the causes of cases, and the second seeks the causes of incidence.” Although Rose does not do so, the causes of incidence could be called forcing factors. Echoing the literature in climate change, forcing factors are factors external to the individual that cause the population average to shift.
Forcing factors have also been called “risk regulators” by Thomas Glass and Matthew McAtee and “fundamental causes” that place people “at risk for risks”, by Bruce Link and Jo Phelan. Those are both perfectly good monikers, but I like the term “forcing factor” because nature knows only forces, not causes. Causes are a human construct.
The visual representation of population health
Rose gives a graphical example of the distribution of systolic blood pressure among Kenyan nomads and among London civil servants. Both populations have a bell-shaped distribution, with a similar within-population variance. But the mean for the Kenyans is approximately one standard deviation lower than the mean for the Londoners.
The metric of standard deviations is not intuitive—very few policy-makers or even academics have a really good grasp of how meaningful it is for one distribution to be a standard-deviation higher or lower than another distribution. Rose’s graph visually displays the high stakes: about one-third of Kenyans have blood pressure lower than what would be expected if the distribution had been that of the Londoners. Yet this verbal definition isn’t very statistically precise, since the whole point is that the blood pressure of every Kenyan is below that of every Londoner at the same percentile in their relevant distributions. And this verbal difficulty is an enormous problem for population health. There is simply no succinct way to express differences between populations in words. We tend to fall back on averages, but we slide quickly from averages to typical experience to a debate over individuals. “The average blood pressure among Kenyans is lower than the average blood pressure among Londoners” easily becomes “The typical Kenyan has lower blood pressure than the typical Londoner”, which invites a comparison between individuals, which is beside the point.
Advancing a population-health perspective may require insisting on images of distributions.