Further Discussion on the Measles Outbreak

What’s really going on here?

In my first article on the measles outbreaks in the Pacific Northwest, I argued that measles vaccination rates have declined little since 1995. Here I’d like to expand on that idea, and comment on specific claims made by various sources.

In my first article on the measles outbreaks in the Pacific Northwest, I argued that measles vaccination rates have declined little since 1995. Here I’d like to expand on that idea, and comment on specific claims made by various sources.

Is the Washington Post Right?

Twitter seems to have stored a clip that cannot be found in the actual article being linked to in Dr. Pan’s tweet, so I am including it here. According to the summary, “lax state laws have helped drive down vaccination rates across the pacific Northwest.” What the article does state is the following.

The Pacific Northwest is home to some of the nation’s most vocal and organized anti-vaccination activists. That movement has helped drive down child immunizations in Washington, as well as in neighboring Oregon and Idaho, to some of the lowest rates in the country, with as many as 10.5 percent of kindergartners statewide in Idaho unvaccinated for measles. That is almost double the median rate nationally.

So is the article being truthful? Well, I’d say it’s an “alternative fact.” Consider the following graphs.

Washington State Vaccination Rates

I gave an overall analysis of immunization rates at the HHS region and state levels, in my first article. But I want to go into a little more detail about Washington State’s vaccination rates. First off, in the last couple of years, there’s been a rise in volatility, but it’s hard to tell whether this rise is out of the ordinary. When people cite such a short length of time and argue that there’s been a “decline” in vaccination rates, it’s no better than when people look at a few years of cold winters and say “look, the Earth is cooling!” A proper analysis does not look at a few data points. It looks at how those data points compare to the overall known data set.

This graph shows the volatility in the last few years. And perhaps there’s something in the 2012 blip, however it doesn’t seem to be enough to explain the timing of the outbreak, especially since overall vaccination rates are a an average of annual rates. That means that year-by-year variations are smoothed over multiple years.

Better Data Needed

In order to make any more conclusions, a lot more data will be needed. Pockets may be on the county level, or the nodes may be even smaller. If that’s the case, there could still be a change in distribution that my analyses aren’t capturing. But if that’s the case, then it doesn’t seem reasonable that they’d be having such an impact on larger scale outbreaks. We can think of very small communities, within a very large population, almost as individuals.

A Serendipitous Experiment has come out of these recent outbreaks. According to Ars Technica, the recent spike in cases, and widespread attention and concern, has caused a massive increase in vaccinations. So not only should we see an increase in vaccination rates in the coming data sets, but we should see a decline in pockets as well.

Now, the issue is that even if source of the recent outbreaks is something other than the antivax sentiment, we should see a decline in cases, because the vaccine does a good job at preventing symptoms. And that’s why we should ensure that vaccination rates are high. But this result also will cloud any data. A proper analysis would therefore have to adjust for changes in vaccination, if the goal is to see if perhaps the pathogen is becoming more virulent.

Asymptomatic Infection Analysis

Especially because of the recent outbreak and rapid changes in vaccination rates among certain populations, now is the time to take action in analyzing asymptomatic infections. As I’ve mentioned in my paper on whooping cough, asymptomatic infections are poorly understood and studied. While there are certainly differences between whooping cough and measles, and their associated vaccinations, there’s still something we should see, if we took a random sample of the population: infection rates should be significantly lower among vaccinated individuals than in unvaccinated individuals. And how different they are can give us greater insight into how well the current vaccine is working to stop the measles infection.

To give some insight into what we could find, I recently came across a paper studying measles vaccination in Taiwan, presented to me as justification for current understanding of measles vaccination. What the person failed to notice is that based on a random sampling of the population, rather than a survey of cases, it seems that the basic reproduction rate is not driven below one, suggesting an inability to generate herd immunity. One reason that I use “seems” is that I’m currently trying to double check with the authors to see if this was an estimate in the case of 100% vaccination or not.

The disconnect between the apparent “eradication” of measles in the United States, and the general recognition by the medical community that a 95% vaccination rate should generate herd immunity, may suggest that asymptomatic infection is indeed a problem. And that’s why we really need to do a wide-scale analysis in the United States.

Antivaxers who Vaccinate?

Does it make sense that antivax sentiment would reduce second and third doses, but not initial doses of the MMR vaccine?

One of the counter arguments to my initial analysis of measles vaccination habits in the United States is that the data set isn’t broken down by how many doses are received. It only a measure of one or more does. I made an assumption in my initial analysis: while the data I had measured the fraction of children who had received at least one dose of the MMR vaccine, we can track changes in this measure to estimate changes in vaccine hesitancy.

What’s the justification for this assumption? Occam’s razor, and the lack of alternative answers to questions of why an anti-vaxxer would give their child one dose of the vaccine, but not the remaining doses, especially when the later doses are given after the time when most children are diagnosed with having autism.

Graph Coarseness

Here’s another point that I brought up in my discussions. The argument, time and time again, is that even though vaccination rates haven’t changed on large geographic scales, there are pockets of low vaccination rates. The argument is that these pockets are very compact, on the community level. Well, the issue with making this statement is that we would need community level data to justify the claim that vaccination habits have changed.

But there’s another issue. If we’re looking at tight knit communities of people with low vaccination rates, made up of family units and neighborhood units that stay together, then we can look at the population graph as a coarser one, where each of these communities becomes a single vertex on the graph.

Whether or not this simplification works depends on whether the pockets are transient or persistent. If they are transient and break apart frequently, then we can’t use the simplification. But the argument is that these aren’t transient communities. And so we can use the simplification.

Now that we can treat each community as a single vertex on the graph, we can look at the vaccination rate of this coarser graph, as I have done, and again we are back to a lack of change in vaccination rates on that scale.

We Have County Data

One counter argument is that we do have county level data. Well, it’s true that we have the data for some years for some counties, but even if you wanted to just look at the vaccination habits in a single state, we would need the data for every county in the state, over the course of multiple years, in order to perform a proper analysis of if and how vaccination habits have changed. We cannot just say “hey look here’s a county where vaccination rates dropped.” Why not? Because it doesn’t tell us if the decrease in vaccination rates in one county weren’t matched with an increase in another. We need to conduct an analysis across every county within a given geographic region, and over many years, in order to make sense of the data, otherwise we’re relying on anecdotal evidence.

Anti-Vax Clusters and Epidemic Burnout

There’s another issue. The claim is again that the antivax community is a tight knit community where a bunch of people with this mentality live in close proximity and interact with each other often. Well, if that’s the case, then the measles virus should burn itself out in such populations.

When you have a highly connected population, and a pathogen that upon infection does result in persistent immunity against future infections — something that is viewed to be the case with measles — then the pathogen would rapidly spread through the entire population, until the only susceptible people left are newborns. But at that point, the fraction of the population that’s immune would be near 100%, which would result in herd immunity! In other words, these populations should become almost universally immune to the pathogen, after an initial outbreak.

Non-Medical Exceptions

Another issue is that people are relying on the existence of non-medical exceptions in order to estimate vaccination rates. But is this estimation justified? Consider the following tweet:

So JoeWV suggested that it was the strict vaccine laws in West Virginia that protects the state from recent outbreaks, as opposed to PA and OH which had relatively lax vaccine laws. There’s just one problem. WV has had consistently lower vaccination rates than PA. OH has generally had lower vaccination rates, but not by much, and in many years, the vaccination rates are higher.

So it doesn’t seem like we can just rely on the lax nature of these laws, or NMEs, to measure actual vaccination habits.

Shifting Burden of Proof

< Public Health Index|

Adapted from an article published at The Spiritual Anthropologist blog on 2/13/2019.