A new paper in JAMA Network Open takes using epidemiological statistics to support ideological goals to the next level. It suggests that since it seems to have happened in 2016, if a Republican even campaigns for President in 2020 Latina women will have more preterm births. 

Well, a Republican is guaranteed to campaign so while we can't prevent all that psychological damage if it's really going to happen, we can at least show this is more politicization of epidemiology. In their comparison of the Obama years (let the partisan dog whistling begin) to the beginning of the Trump administration, of the 32, 860,727 live births there were more 1,342 male and 995 female preterm births than the average in the nine-month period after the election of President Trump - that is 0.00007 more yet they claim to have statistical confidence enough to suggest Republicans are racist and that causes stress and that causes preterm births and correlated birth defects.

With any event, someone who wants to do so can use statistics to make a hyperbolic claim. After the tsunami wreaked havoc in Fukushima, Japan, including a reactor shutdown due to damage caused by the weather event, an American academic mapped differences in birth defects in the US west coast, noticed they were slightly higher, and suggested Fukushima radiation was causing birth defects in the US already.  It got nationwide attention. It didn't matter that actual Fukushima has less background radiation than Tokyo, epidemiologists often have no meaningful science training, all that mattered was that correlation could be created. In the right hands, coin flips can be shown to be not random at all, but completely biased against Latina women. You just have to run a simulation, find a section of data that excites you, and create the link.(1) 

There are numerous problems in this new paper and they start in the first sentence.

If you go to the link they cite as the reference for this, it starts

Is this an op-ed saying we should be more tolerant of each other, even if some people are breaking the law, or a science article?

This kind of mumbo-jumbo technique is ridiculed in the science community so often there is even a whole graphic about flowery language with no real meaning.

One confounder is that they don't know if the Latina women who had more preterm births immigrated here legally or illegally. They might even have been born here. The authors instead fall back on circumstantial data such as that foreign-born women generally have fewer pre-term births but there is a very clear socioeconomic factor in health outcomes, especially between radically disparate groups.

Legal immigrants, regardless of country of origin, tend to be healthier. Illegal immigrants are desperate enough to leave conditions that are harsh, which is already going to have health implications, they are going to endure stress and risk to get here, and then they have to avoid capture until the baby is born. All of those add up as epidemiologists should know. And they should know the Trump administration did not come in and replace immigration officials and policies, they used the policies and staff of the Obama administration. President Obama was derided as "deporter in chief" by the Latino community because he did many of the same things that only now get national media attention. President Obama's own secretary of Housing and Urban Development now says his boss went too far, but Julián Castro only seemed to acknowledge that after President Trump did the same thing and he needed to compete against Obama Vice-President Joe Biden for the Democratic party nomination.

Instead of recognizing all of that, the epidemiologists frame it as that the Trump campaign and election had immediate health consequences. Yet up until literally election day, every legitimate poll had Republicans losing, so we cannot be expected to believe some daisy chain of psychological events immediately led to adverse birth outcomes when the policies being enforced had been enforced the same way, under the same conditions, by President Obama.

Sometimes there is just too much wishful thinking.

What is ailing modern epidemiology in 280 characters. The truly sad part will be that legitimate epidemiologists will spend time objecting to this article and the tweet while never once criticizing NIEHS, IARC, or any of the academic groups that do this every single week. Because most epidemiologists want to work at one of those places and get their names in a story about a link of some chemical to cancer or eternal life. Source: Matthew Hankins

Latinos are 16.7 percent of the U.S. population so if the female numbers are around 50 percent, that would be <8.4 percent women, yet the cohort was 23.5 percent of births, which is impressive - and might explain why Democrats and Republicans are scrambling to lock in their votes. The data size is not a problem but the make-up of the data is. Why not include Muslim immigrants who would likewise have been anxious about deportation? It's a mystery - "the publicly available data we used lacked information." What? You didn't even create your own data set from the publicly available data? It may be clouded by their desire to bolster a claim by Krieger that the Latino community is so anxious about Trump that their health is failing and excluding other immigrants weakens that. There aren't anywhere near as many immigrants from North Africa or the mideast arriving illegally so using their data might boost the case for psychological anxiety if those women suffered the same way.

Instead, it devolves into speculation that immigration-related Executive Orders in January 2017 caused women so much stress they gave birth. Without asking a single Latina.

That is not science, it is not even epidemiology. It is statistical Gerrymandering.


 (1) Coin flips can be shown not random, but biased against Latina women so they get tails with statistical significance. Like this, using a coin flip random generator, if seven times out of 10 you get runs of five tails in a row, you can claim that coin flips are not random with statistical significance. Each “1” is a statistical false positive, a finding that would not be expected to replicate, like that Latina women are so stressed about Republicans they give birth early.

Credit: Drs. Stan Young and Henry Miller, Genetic Literacy Project