One of the first things I learned about in epidemiology was the importance of study design. The type of study that is conducted, and the type of data available affects the conclusions we can draw from the results.
A few weeks ago, I came across this startling headline:
“Breathing dirty air can make you sick. But according to new research, it can also make you more aggressive.”
But, when I clicked on the original study, I found that researchers hadn’t studied humans at all – they had studied American counties – and even though there is nothing wrong with that on its own, using this type of data to draw conclusions about relationships in people is problematic (1).
In my opinion, one of the greatest strengths of epidemiology is that it is one of the few scientific disciplines where the unit of analysis is a person. No extrapolating from mice, or cells in a petri-dish – we go directly to the source. This means that the exposure, outcome, and any other variables are directly measured on a person.
For example, here’s a few rows of hypothetical data below:
Table 1. Individual Level Data
Note that each row above represents an observation on a person. This dataset would have required information collected on three different people.
The ecologic study design is an exception to the typical epidemiologic study design because the unit of analysis is not a person, but a geographic unit.
Table 2. Ecological Data
In the dataset above, each row represents a county, not a person. This dataset required information on three different counties.
The ecological fallacy is an important concept applicable to ecological studies. It occurs when data collected using an ecological unit (like a county) is used to draw conclusions about relationships at the individual level.
Using a textbook example, we can see why this can be problematic:
An ecological studied revealed that the percentage of Protestants in communities and suicide rates were positively correlated in Prussian communities. Using just this ecological data, someone might conclude that being Protestant leads to a higher suicide risk. But, it is possible that Catholic individuals, who, when in the minority in largely Protestant communities may be at a higher risk of suicide due to feeling socially isolated (2). In this case, it’s clear why using ecological data to draw conclusions at the individual level can be dangerous.
Ecological studies can be very valuable: they can be completed relatively quickly, and results from ecological studies can help us generate hypotheses that can be tested using individual-level human data. But, standing on their own, results from ecological studies need to be carefully interpreted so that we don’t draw inappropriate conclusions regarding exposure-outcome relationships in humans.
ScienceDaily shouldn’t have used language implying individual level data in their article.
The fact that the authors were not able to provide any biological basis for how air pollution could lead to increased crime also makes me skeptical of the relationship described in the paper. The authors also conducted an economic analysis calculating possible cost savings from crime reduction if air pollution levels were reduced in America. These types of analyses should be reserved for when we have stronger evidence for causality, and not just correlations.
Although this study shouldn’t be used to draw conclusions about a causal relationship between air pollution and crime, there is a growing interest in the neurological effects of air pollution, and this study could add to this growing body of research as a hypothesis-generating paper. There is some up and coming research using individual level data that suggests that air pollution may be linked to neurological effects such as memory loss, fatigue, cognitive function, and Alzheimer’s disease (3). A recently published Canadian study has also linked air pollution to brain cancer. But, it is important to remember that this is an emerging research topic, and that many more studies will need to be conducted in order for us to understand the relationship between air pollution and neurological effects (4).
The main takeaway from this post: The unit of analysis is an important aspect of any study and shouldn’t be overlooked in scientific articles!
References:
1. Burkhardt, J., Bayham, J., Wilson, A., Carter, E., Berman, J. D., O'Dell, K., ... & Pierce, J. R. (2019). The effect of pollution on crime: Evidence from data on particulate matter and ozone. Journal of Environmental Economics and Management, 98, 102267.
2. Szklo, M., & Nieto, F. J. (2014). Epidemiology: beyond the basics. Jones & Bartlett Publishers.
3. Xu, X., Ha, S. U., & Basnet, R. (2016). A review of epidemiological research on adverse neurological effects of exposure to ambient air pollution. Frontiers in public health, 4, 157.
4. Weichenthal, S., Olaniyan, T., Christidis, T., Lavigne, E., Hatzopoulou, M., Van, K. R., ... & Burnett, R. (2019). Within-City Spatial Variations in Ambient Ultrafine Particle Concentrations and Incident Brain Tumors in Adults. Epidemiology (Cambridge, Mass.).