When everything is a leading cause of death, nothing is
Is poverty the fourth leading cause of death?
I saw an op-ed published in The Guardian this week claiming that poverty is the fourth leading cause of death in the US. The source for this assertion is an April 2023 study in JAMA Internal Medicine. The study examined a representative, longitudinal dataset tracking individual income matched with vital records recording deaths. The researchers concluded that currently being poor was associated with a 42% higher risk of death and cumulative poverty (“being always in poverty vs never in poverty in the past 10 years”) was associated with a 71% higher risk of death.
It shouldn’t surprise anyone to learn that poverty is associated with worse health outcomes. In fact, it’s more than a mere association. Poverty limits access to healthy living and appropriate medical care. It exposes one to environmental and social dangers. It cuts off educational opportunity. Poverty can fairly be viewed as a fundamental cause of death.
This isn’t a new finding. So what the researchers have done to draw out the importance of poverty is directly compare it to other leading causes of death. As you can see from the figure below, the study estimates that its measure of “cumulative poverty” contributed to almost 300,000 deaths in 2019. Taking this chart at face value, poverty could be viewed as the fourth leading cause of death.
You shouldn’t, however, take this chart at face value. Charts like these conflate risk factors or upstream contributors to disease with the final, biological disease process. There were 2.85 million US deaths in 2019. Add up the numbers in this chart and you will reach greater than 100% of all deaths, even before counting minor causes not listed on the chart.
To see why this is a problem, consider “smoking,” which is positioned as the third leading cause of death in the figure. The CDC estimates that smoking contributes to 480,000 deaths, but smoking is not typically listed as an underlying or immediate cause on death certificates. Instead, there is a separate box where the physician or coroner can check off whether they believe that tobacco contributed to the death. To directly compare heart disease and cancer and respiratory disease to smoking is nonsensical; smoking causes death through heart disease, cancer, and respiratory disease. This is double counting.
Many upstream factors contribute to a biological disease process. If an older gentleman who is poor, overweight, and a smoker dies of a massive heart attack, which risk factor “caused” his heart attack: age, poverty, obesity, or tobacco? All of them, of course. But not equally. (The cardiologists would likely say that “lack of a statin drug” was also a major contributor.) It would be fair to compare the relative contribution of different social and biological risk factors to the disease process. These studies have been done elsewhere, but not here.
Pathologists and social epidemiologists understandably look at mortality through different perspectives and causal models. But a tally that directly compares smoking and obesity with cancer and heart disease does not present a coherent framework. The precise interactions between social factors are complex, unclear, and probably beyond the purview of a research letter in a medical journal. (Who thinks poverty, obesity, and smoking are fully independent?) Furthermore, an observational association present in one data set, even if some confounders are controlled for, does not establish a definitive effect size. In this case, I have no doubt that poverty’s effects on health are “real,” but calling it the fourth leading cause of death is unjustified precision for the sake of politics—and I likely agree with much of this politics.
This isn’t the first time the “trick” of combining risk factors and disease into the same figure has been used. A viral paper published in 2016 claiming that medical errors were the third leading cause of death made the same mistake, tacking on a risk factor for death to a tally of causes derived from death certificate data. There are many other problems with this approach, as well, which I have written about elsewhere.
The medical error article called for a specific entry on death certificates—in the same we do for tobacco—to acknowledge when a mistake contributed to the death. (New York City has incorporated “therapeutic complications” as a manner of death.) One could reasonably propose the same thing for poverty. But the exact contribution of these and other risk factors to an individual death cannot typically be discerned from physician judgement, which is what death certificates rely on. Furthermore, selectively listing risk factors gives them an inherent prominence in vital statistics. We’d only ever be capturing a small selection of risk factors, which will certainly lead to the misunderstanding that only these risk factors matter.
One way of thinking about why someone dies is to consider the counterfactual: would they be alive had that event not occurred? The counterfactual is pretty clear if someone dies of a gunshot wound. They would be alive were it not for the bullet. The early days of COVID also presented a clear counterfactual: the person dying of severe lung damage in an ICU would be alive had SARS-CoV-2 not infected them. Today, the situation is more complicated. Doctors like to say that viral illnesses often “tip over” already-sick people. This is evidenced by the fact that COVID is increasingly listed as a contributing rather than underlying cause on death certificates. (This does not mean the virus was incidental.) When it comes to chronic diseases, the causal chain is muddy. The day before an elderly, obese, poor, smoker dies of a heart attack, they were also elderly, obese, poor, and smoking—but they were alive.
How do we apply interconnected social influences like poverty to this “but-for” framework? Can we truly imagine a world where nothing changes except that no one is poor? Think of all of the downstream effects of having or not having money, and imagine what upstream factors—political institutions and economic markets, to name a few—would have to change for poverty to be eradicated.
During my pathology training, I was often asked to render a conclusion about cause of death with an autopsy. The doctors or family wanted a smoking gun: a clot in a coronary artery or a metastatic tumor disrupting a vital organ. There was often no smoking gun to be found. There are many medical abnormalities that cannot be detected at autopsy—cardiac arrhythmias and metabolic derangements being common ones. But there are also causes more fundamental than that, causes like poverty, which cannot be seen under a microscope or easily quantified in any particular case, yet are nevertheless real to anyone with eyes and ears and a heart.
Thanks for reading,
Ben