Dr. Flavio Coelho, Vaccine Modeler

Recent research published in PLoS Computational Biology explored the dynamics of how people within a population decide whether or not to vaccinate themselves against an emergent disease threat.  A summary of that article, “Drivers of the Decision to Vaccinate“, was published Monday, August 17th.  What follows is a transcript of my interview with Dr. Coelho, one of the authors of this research.  Dr. Flávio Codeço Coelho is affiliated with the Theoretical Epidemiology Group, Instituto Gulbenkian de Ciência, in Oeiras, Portugal and Dr. Claudia Codeco, his co-author on this study, is affiliated with the Scientific Computing Program, Oswaldo Cruz Foundation, Rio de Janeiro, in Rio de Janeiro, Brazil.  Dr. Coelho was allowed to edit the transcript of this interview for clarity; I attest that no substantial changes were made.

AH: Today is August 5th, 2009.  I’m sitting down with Dr. Coelho, who has recently published “Dynamic Modeling of Vaccinating Behavior as a Function of Individual Beliefs” in PLos Comp Bio.  He’s graciously agreed to answer some questions about epidemiology and his methods.  How are you this morning, Dr. Coelho?
Dr. FC: I’m fine.

AH: Allright, so shall we just get right to it?  Personally, what initially attracted you to the study of epidemiology?  Are there any particular concepts or parameters within the study of epidemiology that you find particularly captivating?

Dr. FC: My first interest in epidemiology, particularly epidemiology of infectious diseases, is to be able to understand the dynamics of spread of infectious disease in large human populations.  There is a number of interesting questions, some partially answered, some still unanswered, about those dynamics.  And in today’s globalized world where people travel quite a lot between countries and different areas of the planet, understanding how human behavior influences the spread of infectious diseases is a very fascinating and very important topic.  So that’s basically what attracted me to this field and as a mathematical modeler there is a lot that can be done by setting up mathematical models to represent those dynamics.

AH: One of the concepts that’s often given with vaccination and epidemiology is the concept of herd immunity, where enough people are vaccinated that those who are not vaccinated are protected by proxy effect.  So within epidemiology, does the percentage of population coverage necessary for effective herd immunity vary by pathogen or is it fairly constant?

Dr. FC: No, it does vary by pathogen because it depends on how transmissible is the disease, meaning how easy it is for the disease to go from one person to the other.  And then according to that and how fast that disease or how well it spreads you have to have a higher coverage in order to obtain herd immunity as compared to diseases that spread with less speed or efficiency.  So you cannot have one figure that will determine what is the best vaccination coverage for any disease.

AH: Allright, in the study of epidemiology, how can data from past epidemics inform us of the course of potential future epidemics when the world has changed so rapidly in terms of moving people and goods around?  In particular, what can the 1918 Spanish Influenza tell us about the potential course of the emerging swine flu?

Dr. FC: Well, the study of past epidemics is very important because, even though the dynamics of human population movements are not the same, to a certain extent on some certain smaller geographical scales there is a lot to learn about how a certain disease spreads in a reasonably homogeneous population.  For instance, in a large city like London or New York where you have certain population density and depending on how many and what fraction of the people there are susceptible to a disease, meaning they have not been exposed to it before, as you examine the dynamics of past epidemics you’re able to determine certain key parameters about the mechanisms of transmission and the speed and things like that about particular diseases.  Even though diseases and pathogens evolve through time, this evolution is not sufficiently fast that you cannot really extrapolate from past epidemics certain facts that help you understand present epidemics.  And in the case of influenza and the current swine flu, this is also true. As you know, flu is one of the fastest-evolving pathogens that we have as a common infectious disease and it’s a seasonal disease because it evolves so fast, that every year we have a slightly different strain that we’re not longer immune to or not completely immune to, but even so the patterns of the transmission of flu don’t change much.  Flu is still spreading pretty much the same way now that it was spreading during the Spanish Flu in the beginning of the 20th century.  So the patterns of transmission don’t change a lot.  Pathogens change, but their modes of transmission tend to be more conserved in general.  Of course it may change for certain diseases but it’s a rarer event that the mode of transmission is changed.

AH:  In particular, within this frame, in the Western media there was a lot of initial very urgent media coverage of the emergence of swine flu and it has kind of faded to footnotes even as the epidemic has grown.  Based upon the results of your research, do you regard this reduced media coverage as beneficial for the emerging vaccination campaign, if a vaccine is developed?

Dr. FC: I my paper I was mainly concerned with media coverage of side-effects of vaccination having a negative influence in voluntary vaccine uptake. I think in the particular case of the swine flu the media coverage  has been a positive thing, with no ill effects on the prospective vaccine image because initially we had no vaccine. So the media did a good job of creating awareness about the disease and the behavioral changes which can minimize transmission. Compared to other vaccines available on the market today, there’s not a lot of resistance towards influenza vaccines as compared to some other vaccines that are compulsive in some countries, like MMR,Measles, Mumps and Rubella.  Some of those find some resistance among people who think they impose more risks than protection.  But in the case of influenza there is not much resistance towards the idea of getting immunized, so I think that the fact that the media is giving coverage to the swine flu pandemic, it will help us have a good acceptance and good vaccination coverage when the vaccine is finally available.  On the flip side of the story, there’s another problem, which is that vaccine production is pretty expensive and resource consuming and that there is no way we can produce in the few months of vaccine production enough vaccine doses to cover everybody that is at risk in the world.  So a big rush to vaccination, probably a big irrational rush in people who are not really in threat zones and zones where there is higher risk—this big rush could mean that people who need it most will not get it, particularly in developing countries.  So that’s the flip side of the high attention that the disease is getting.  It’s inevitable, but if the media coverage is of high quality that people understand what their risks are and who really needs to take the vaccine, that may help lessen that side effect that I just mentioned.

AH: I hope so.

Dr. FC: Yeah.

AH: In your opinion, has the rise of instant international communication helped manage and mitigate epidemics or has the deluge of news diluted people’s trusts in their governments and thereby weakened the effectiveness of vaccination campaigns?
Dr. FC: Well, as I mention in my paper, the relationship of information availability and decision-making in epidemic scenarios it’s a very complex issue and it depends on many factors, for instance, it’s highly sensitive to what the general public’s relation of trust with their own government.  There is, and I cite in the paper the example of Brazil and the yellow fever scare of an epidemic that never was, that scare was derived pretty much because the population did not trust that the government would tell them the truth, that they would try to cover up a real epidemic in order to avoid the political impact and thus the population decided to protect themselves no matter what the official recommendations from the Ministry of Health were.  So you see that it depends on what kind of trust relationships that is built over the handling of different similar scenarios in how the population will react.  It’s kind of hard to tell, but I think that in all, in each country the public health officials should have a pretty good feeling of what level of trust they have, what kind of relationship they have with their particular audience.  They should take that into account when they plan their interventions for those epidemic or pandemic threats.

AH:  On the same note, in terms of media coverage, recently, in the past couple years in the West, there have been anti-vaccination activists citing a discredited study that claimed a link between the MMR vaccine and the development of autism. As one who is interested in vaccines and epidemiology, do you think there any basis to this argument?

Dr. FC: I have to say that there’s not hard evidence to back that up.  There are some cases that might be associated.  It came up, if I’m not mistaken, during some campaign in the UK and it gained public attention but there is not exactly hard evidence to back this fear up.  Now, fears of vaccination stem from many directions.  There are some people who don’t vaccinate for religious issues, so it’s really a weird thing, it’s like, some people don’t believe vaccination as they don’t believe evolution and it’s not really a rational or data- or evidence-based behavior.  So you cannot treat resistance to the concept of vaccination as a homogeneous thing.  What we can do about, you know, to preserve the public image of vaccines is mostly to explain to people who are willing to accept evidence presented by the media or the government as real and honest evidence.  But there will always be a large, or, I don’t know, probably not a large, but a fraction of the population that will not be moved by hard evidence.  They have other reasons to fear vaccines and they will use fake associations between vaccines and ill effects to create their own scare and their own small communities, so that’s inevitable.  But what I try to point in my paper was that even, you know, you trust vaccines as a general fact, it’s also a fact that some vaccines have a small risk of adverse effects and they do occur.  So that should be taken into account in the way people react to vaccines.

AH:  In a similar vein, one of the arguments that the anti-vaccination campaigns often invoke is that Western children receive far too many vaccines too soon after birth and that this overloads their immune systems due to alum adjuvants, chemical preservatives, or trace amounts of mercury in the vaccines.  Is there any basis whatsoever to that?

Dr. FC: Well, I think that, clearly, there’s a lot of exaggeration in those claims but no matter what you might be able to argue scientifically against the number of vaccinations given up to a certain age, I don’t personally know of any hard evidence saying that that is really something bad.  But the reason for that is some real threat of childhood diseases and the real epidemic threat that they present to the entire population because if you stop vaccination for something as simple as chicken pox, for instance, in children, in a country like the US, or like Brazil right now, it can be a major epidemiological problem because a lot of the older population is no longer immune because the vaccine effects wane and people have not been exposed enough to the virus.  And chicken pox at a later age can be a serious illness and it can even lead to death or neurological problems—the disease, not the vaccine.  So for certain diseases, you have to keep them out of the population as best you can.  And the regular point of entry of any disease in a population that has some level of immunity is through infants because infants are born without immunity to those diseases because they’ve never been exposed before.  So unfortunately the only way we can guarantee that diseases will not get in is to get them be vaccinated against those diseases early on before they have a chance to be exposed.  So I don’t think personally that it’s a lot of vaccines that we give to children.  And the good it has done to humanity is much larger than any ill effects and there is no evidence for that, actually, I don’t believe that there’s hard evidence for this being a bad thing.

AH: In some of the seminars I have attended recently, the guest speakers have been talking about things such as DNA vaccines or transdermal vaccines.  How would you hope the development of these effective vaccines, if they can be effective, would impact consumers’ choices about vaccination trust?

Dr. FC: Naturally, the emergence of, especially DNA vaccines, they will certainly offer the ability of providing protection from diseases that we have so far been unable to have immunization tools for.  But, as with any new technology or any new thing, it has to be, there’s bound to be some risks that are still not known and as with any vaccine development or drug development, careful clinical trials have to be performed to understand [what] the possible side effects of such vaccines are.  We’re at the dawn of the DNA vaccines, which have a mechanism that’s completely different from traditional vaccines, which in any case mimic our natural immune behavior, so DNA vaccines are not only not the same as traditional vaccines, but also not anything like we’ve ever had naturally.  So it’s very hard to predict what possible associations it may have with other aspects of our health.  So that has to be looked at carefully—it’s not to stop research or anything, but it has to be looked at.  Of course, there will be risks that will have to be addressed and I’m sure that the technique and the science of it will evolve considerably in the next few years.  Now as for transdermal vaccines, I’m really not very familiar with what is being proposed so far so I would rather not comment.

AH: What I understand is that it’s primarily a push for vaccines that don’t require refrigeration, have longer shelf life, and don’t require needles for administration, particularly for use in rural areas of the developing world.  But to move on to the next question then.

Dr. FC: OK

AH: So, within the paper, from what I understood, the model was based on log pooling of Bay inference of individual choices.  So how was this developed and are there any particular caveats of the model in fitting this behavior to real life historical events?

Dr. FC: The model used logarithmic pooling to combine beliefs, represented as probability distributions, and Bayesian inference to update the pooled beliefs based on available information. So the model tries to represent our decision about vaccination as derived from our beliefs in how safe it is to vaccinate.  Because every decision in a person’s day to day life is based on what we know.  So if I’m going out on any given day I’ll take my umbrella if I think I’m on a rainy season if has rained a lot and there’s clouds in the sky.  So past experiences drive how we take from the simplest decision to the most complex ones.  So if we want to model something as complex as vaccine uptake in a population, we cannot start as has classically been done that people will vaccinate just because someone told them, or that they will vaccinate just as a fixed fraction of the population in a given situation.  It’s not as simple as that.  Another thing is that, to be able, another key contribution of this paper, was to not only make the decision derived from what people believed about it but also to allow that belief to vary throughout the epidemic according to the unfolding of events of a regular epidemic.  So that’s the paper’s core contribution: to use beliefs to drive decisions, and allow beliefs to vary in time in a dependent way with the facts unfolding during an epidemic.

AH: With these conclusions from the paper, it seemed from what I read of it, I mean I read it but I may not be understanding everything, but from what I got out of it, it seems early coverage of adverse events significantly affects a population’s willingness to vaccinate regardless of the gravity of the disease scare whereas if there’s a disease scare and only later reports of adverse vaccine events, coverage becomes much greater more quickly.  So with those conclusions, where do you hope to go next with you research?

Dr. FC: I think that the natural extension to my conclusions and the methods developed so far in this paper, the next natural step would be to bring the model closer to reality. There are two different avenues I can follow right now.  One would be to better parameterize the model from some field measure of what people believe and how they behave subsequently, but I have no expectations as to when I’ll be able to do that because that would require the opportunity to do fairly large-scale field work and interviewing people and then following up whether they vaccinate for a given disease or not.   The other, more theoretical, follow-up to this would be to complexify a little bit the belief model that I used, not in order to simply make it more complete, but to make it more realistic in a sense, because in any situation we don’t take a single fact or a single issue as a source of our decision.  Normally we combine different things in order to make a decision.  For instance, in the umbrella example I gave, I’ll take the umbrella not only depending on my beliefs on whether it’s going to rain or not but also if I expect to be outside walking on the streets for a long time or just taking a car ride.  So it’s a multivariate problem, so that multivariate approach to belief modeling connected to the epidemiological modeling is very important.  So another thing is that part of this model that could be further extended is the media amplification.  I introduced in this model the concept of media amplification factor and that’s a very important one, which is that media tends to latch on to things that sell and things that sell normally are the bad things.  Like sometimes if someone took a vaccine and were safe, it’s not news material, but if someone took a vaccine and died or went to the hospital, that’s news.  And it’s a fact that the news, the media industry, is biased towards things that are a lot more radical.  It introduces a bias in what people are exposed to, the kind of information that people are exposed to.  So I’d like to explore the concept of media amplification, the sources of bias that it could introduce.  The model that I had in this paper was a very simple thing.  I just said that vaccination ill effects were some number of times more likely to appear in the news than positive effects, and that’s basically a very simplistic way of talking about media amplification.  But I think that deserves some more careful and complex understanding.

AH:  So, out of curiosity, when developing this model, did you start with the MMR vaccine scare and the yellow fever disease scare?  So did you start from past historical examples, or did you develop the model first and replicate the behavior in those examples.

Dr. FC: I started with, my first motivation was to get belief, behavior, and epidemiological dynamics together.  And then I went out to look for historical examples, past situations where that might have been a key issue, where that might have played the role in concrete situations.  And then I gradually pulled together the elements that composed my approach.

AH: Given that PLoS is Open Science, why did you choose to go with PLoS Computational Biology?

Dr. FC: That’s a good question.  Despite the fact that PLoS is open source, PLoS Computational Biology is also a very respected journal and I first of all thought that it was a good venue for the idea I was trying to convey.  But on the other hand I’m a big believer in open sources for scientific information because scientific information is funded by the public and the information should be available without delay to the public, so open source, open access publications are key to that.  Moreover, PLoS is introducing some very interesting social networking technology to make the way people interact with the published papers more rewarding to the authors.  If you go to my article page, you can rate my article, you can select pieces of text and add comments, and that’s unheard of in traditional pubs and I think that when people start to get accustomed to talking back to researchers and adding comments and using all those tools that are so widespread in the way that people interact throughout the Internet nowadays, it will be a big way of society interacting with science and it can only be a good thing.  And that’s the moral reason, if you will, why I chose PLoS journals.  And I intend to keep publishing in PLoS whenever I can because I think that they’re pretty much the only guys out there offering this.

AH:  OK, thank you very much for taking the time to sit down with me.