Did Trump Win the "Sick Vote"? Doc-to-Doc with Jason Wasfy, MD

With election-night results that defied pollsters expectations, the political science community has been scrambling to find an explanation for Donald Trump's surprise victory.

According to Jason Wasfy MD, senior author of a recent paper analyzing voting shifts which appeared in Plos One, the "secret sauce" of Trump's victory may have been a republican-shift among those with poorer health status. I discussed the provocative article with Dr. Wasfy, and asked what it means for both parties moving forward.

Transcript:

Perry Wilson, MD: The 2016 Presidential Election defied expectations from political scientists and pundits on both sides of the aisle. Since that time, researchers have been looking for explanations of the surprising results. We've seen reports saying that votes were driven by race, driven by class, or driven by certain ideologies. But one paper appearing in PLOS One brings healthcare and health itself to center stage. That paper was authored by Dr. Jason Wasfy and his team. Dr. Wasfy is a cardiologist and the Director of Quality and Analytics at Massachusetts General Hospital. Dr. Wasfy, thanks for joining me today on Doc-to-Doc.

Jason Wasfy, MD: Oh, I'm delighted to be here, and I really appreciate your interest in my research.

Wilson: Your paper links community health status to votes for Trump in the last election. Tell us a bit about what you found.

Wasfy: I think there's a lot that's known in political science about areas that vote for republicans and areas that vote for democrats, but what's less understood is how areas shift. Especially in this election, there appeared to be a political re-alignment that although the overall shift was much more republican in 2016 than it was in 2012, there was also shifting. The state of Wisconsin went republican for the first time in 32 years, yet the state of Texas went relatively more for the democratic candidate, Hillary Clinton, than any year in the past 20 years. There seemed to be a political realignment. A lot of this had been discussed in the popular media, for example, and associated with age and race and different things. We wanted to see if health was related to this shift.

Wilson: So you looked at these counties that changed their voting percentage by varying magnitudes and tied that to the level of health at the county level. Can you go into that a little bit? What factors dictate how healthy a county is?

Wasfy: Sure. It's a hard concept to measure. What we had to do is aggregate a lot of health status variables that were available, things like physical health, things like the mortality rate, violent crimes, all sorts of different health indicators. So the problem is -- and it's a familiar problem for those of us who do health services research -- that the variables are often colinear. So areas where there are more diabetes, there's more obesity.

To deal with that in a regression model, what we did is we reduced it to a single measure of health. We reduced all the variables we could get our hands on to a single measure of health, and then performed a regression on the health variable with the shifting of votes towards Donald Trump in the 2016 Presidential Election. In that way, the shift would be greater if the difference between Trump votes in 2016 and Romney votes in 2012 was greater. So some of these areas were actually very republican areas, but they shifted more relatively towards Trump in 2016.

Wilson: Just to make sure I have the direction correct. The less healthy an area was, as you sort of aggregate all this data, the more likely they were to shift sort of farther red, further towards Trump than they had been, even if they were voting republican originally. Is that right?

Wasfy: Yeah, that's exactly correct. Obviously with linear aggression, it's an overall effect. There were counties that did all sorts of different things, but in general, counties that were relatively more unhealthy were the counties that shifted relatively more towards Trump. So if they had voted for Romney, they went more for Trump, in general, than they did for Romney. Or even some of the counties that both times went for the democratic candidate went for Hillary Clinton relatively less than Barack Obama in 2012. So what we were measuring was shifting of votes within counties, not the overall vote.

Wilson: Put this in context for us. Healthcare was a major issue in the last election, obviously. We had both candidates talking about healthcare. Does this mean that the Trump message that healthcare was failing, that Obamacare was a disaster, really resonated in those counties? Was this effective messaging?

Wasfy: I think it's extraordinarily difficult to say why these areas shifted votes. I think the thing that we can say is that the areas that were more unhealthy, the counties that were more unhealthy, shifted relatively more towards Trump. The mechanism of that is very unclear. That's important to emphasize for several reasons.

We don't know that poor health caused this shift. For example, they could have been associated with other factors like social distress that were not included in the model because they're very hard to measure. That's one thing to be very clear about, that we can't impute causality from this, and we certainly can't say what these voters were thinking. I'm trying to stick very hard to the data and not sort of go beyond of what I can say, but I will say this. I think it is true that any policy that has its effect to decrease the proportion of Americans who are insured will relatively hurt these communities more, these communities that shifted towards Trump in the 2016 election.

Wilson: This was a multivariable model, so in addition to your health status variable, you accounted for socioeconomic status, race, age, a variety of other factors. Can you give us a sense of how much health mattered compared to those other things? We've heard so much about, for instance, certain demographics, white males without a college degree, or things like that, that people have told us drove this election. Can you give us a sense of the magnitude of health compared to some of those other factors that you included in your model?

Wasfy: It's a fantastic question. As we talked about before, we reduced a lot of the health variables, basically all of the health variables, to one unhealthy variable because of this issue with multi-colinearity and then included it in our regression model with some of the factors that you had mentioned, education, race, age, and so forth.

The problem, of course, is that your question is about relative magnitude, so it's a little bit challenging to say, "Because of the way that these different quantities were measured, if education or race, for example, was more important or less important." But I will say this. Even after risk standardizing for things like education, race, wealth, age, the health effect seemed very, very strong. It was particularly strong in the states where the electoral college shifted.

Again, regression is not a statistical technique that can impute causality. It never can in observational data, but I'll emphasize that this effect seemed very significant and was particularly significant in the states that switched their electoral college vote. Again, we can't impute causality from the analysis that we did, but certainly a causal relationship is possible given what we found.

Wilson: If I'm reading the paper correctly, you're looking at changing in voting as compared to health status in a community at a single point in time. I'm wondering if you examined the change in health status, in other words, were communities that became sicker more likely to vote one way or another or is it just sick overall?

Wasfy: It's a fantastic question and something we've thought about a lot. Strictly speaking, our analysis was a correlation of voting changes between 2016 and 2012 with health status that more or less correlated with that time interval. So you're right. You're seeing a time effect with voting, but you're not seeing changes over time in health status put in our analysis.

I think especially given our findings, it would be fascinating as a next step to understand how changes in health status affect changes in voting, because strictly speaking, you're absolutely correct. We did not look at changes in health status over time. I think the interdigitation of the health of communities over time and the way that they vote would be a fascinating next step, especially given how strong the effect we found had been in 2016.

Wilson: Given the findings in the swing states where health status was even more profoundly associated with change in voting direction, I think both political parties are going to be paying very close attention to your results come 2020 or 2018, even. What would you tell them? What do politicians need to focus on to sort of swing those votes either further or back the way they were before?

Wasfy: I think politicians and leaders throughout America need to recognize that health matters, that the public health of communities matters to the people who live there and may have a causal relationship with voting choices. We can't know, again, whether health is causing these sorts of effects or a marker of distress that causes these effects, but communities that are unhealthy are going to vote in a way that reflects their distress. I think that that shouldn't be shocking, but it clearly happened in 2016.

I think for a variety of a reasons -- and it's not only about our voting analysis, it's also the right thing to do is that -- when Americans are suffering that we all need to pay attention. I think that the public health of communities is one way that the suffering of communities can be manifest.

Wilson: Dr. Wasfy, thank you, once again, for joining me today.