Ebola Ebola

Thomas Eric Duncan died this morning.

I spend a lot of time in my writing hiatus sending emails about ebola. I continue to spend a lot of time reading emails- and papers, and datasets, and my own models- about Ebola. I submit to you that Ebola matters not only because it is a horrible tragedy for a cluster of countries in West Africa, and a smaller tragedy for a handful of families elsewhere in the world, but also because it represents the first big, fast-moving outbreak since computer modeling reached its current level. SARS was a good test case, and should be cited more often in public discussions of disease (instead of, say, Spanish Flu, which happened in a completely different public health universe) but in the end, it is Ebola that has permitted everyone with a pet theory about epidemic disease to trot out their assumptions and try to prove them right. Let’s talk about Ebola. Let’s talk about what Ebola means.

First some facts

Ebola is a virus, it is extremely rare, it seems to have a non-human reservoir, presumed to be migratory bats, and it affects mammals other than humans. Humans are not a terminal host for Ebola, but not a great one either. There are several known strains of Ebola, effectively one per outbreak, and a genetic “tree” of sorts has been developed from the sequences of these strains, but the extent to which they differ clinically is both dramatic and poorly understood. The current outbreak is Ebola Zaire, the same strain as originally discovered by Piot’s team in Yambuku in 1976. The current outbreak in the Democratic Republic of Congo, which was once actually called Zaire, is also Ebola Zaire, but is not the same as the current West African strain.

Ebola is a hemorrhagic fever. It presents like most any other viral fever, with severe headaches, runny snot, puking, diarrhea, muscle aches, fatigue, and mental vagueness (“depersonalization”). It also causes intravascular hemolysis and DIC, which stands for disseminated intravascular coagulation and means, pretty much, “your blood clots randomly until it runs out of clotting stuff, then it never clots at all.” It is this profusion of blood clots and uncontrolled bleeding that leads to the rash (small specks of bleeding from capillaries under the skin), the black vomit (bleeding into the stomach) and the bloody stool. Ebola also causes nosebleeds, bloody spit, and bloody urine (from bleeding in the kidneys.) Ebola also causes profound hiccuping. Folks, I don’t make this stuff up.

The typical Ebola patient acts like the sickest you’ve ever been and then a bit sicker. They lie on the ground and moan, the get up and lean on something, and they are totally miserable. Sorry to anyone who was hoping for predatory zombies. PS you’re a jackass.

Ebola is extremely infectious and not very contagious. This means if you get any Ebola virus in you, you are extremely likely to get sick, however it is relatively hard to get the virus into you. Obviously Ebola spreads very rapidly and dramatically, but what it does not do (or what it seems not to do very often) is cause the kind of sub-clinical infection where your body fights it off before you realize you’re sick. In other words, everybody in your workplace got exposed to that cold last year, but only those six people ever had symptoms and only one had to stay home? Ebola is not like that.

To get Ebola, you need bodily fluids on a mucus membrane. That can be your eyes or mouth, and those fluids can include droplets propelled by coughing or power-washing (seriously!). There is a lot of chit-chat about “airborne” viruses, which I’ll talk about below. Short version, though, is this- WHAT THE HELL IS WRONG WITH YOU ISN’T THE DISEASE AS IT ACTUALLY EXISTS BAD ENOUGH?

Ebola outbreaks in the past have happened in rural areas, and have burned out fairly quickly. We’ve never seen an urban outbreak like this one before. The behavior of different strains, even different outbreaks of the “same” strain, has been pretty diverse. There are serious data holes in all epidemics, and predictive models, like all models, are all wrong and sometimes useful. I don’t know what’s going to happen and I’m not afraid to worry in public, so here are a few ideas explored.

Is this the end?

The 2014 Ebola outbreak is quite possibly a civilization-ending event, if the civilization in question is the stable post-civil war Liberia that was just getting back on its feet after killing off 5% of its population in the course of a few short decades. With, like, guns and machetes and stuff. President Ellen Sirleaf Johnson already had to dismiss her cabinet when they took the opportunity to flee the country at the beginning of the outbreak and refused to return when ordered. The collapse of the health system has been matched by distrust of health care workers, though nothing (yet) like the healthcare teams being attacked and murdered in neighboring Guinea. In a particularly nasty string of events, a “clinic” run by the government health service in a capital-area slum turned out to be an abandoned school with no health care staff and no running water, essentially a jail for the poor and febrile. The families of the “patients” (it would not be incorrect to say “inmates”) came to bring their parents and children home, it triggered a riot, leading to the quarantine of the entire district. The next few days saw armed guards beating and even shooting residents, and an attempted mass exodus of boats was turned back by the Liberian coast guard. USAID’s Famine Early Warning System is predicting Phase 2 Food Insecurity in Liberia for the next six months. Furthermore, for individual Liberians, the stigma of Ebola is likely to make the prospect of traveling to unaffected nearby countries for work much harder, which will deplete exchange credits and exacerbate social tensions at home. I’m not saying another civil war is certain, but the next ten years will likely not be kind to Liberians, even after the disease is gone.

For everyone else, there are undeniably models that predict widespread infection outside the ECOWAS area. The lack of reliable data makes assessing the relative likelihood of these models impossible. If we didn’t know how Ebola behaves in an urban African population, we sure as heck don’t know how it behaves in an urban industrialized high-development population. Many of the measures that are floated as parameters for these models, such as R0 (the number of people infected by a single person, on average) or doubling time, are historical observations and are highly sensitive to medical, cultural, political and even demographic changes. Grant Brown at GitHub has done a good job at tracking changes in R0 by time and by country (scroll down to “Empirical R0”) and the variation there should put any would-be predictor on watch. We will know more soon enough, I imagine, but this isn’t a 28-days-later kind of collapse. Even the worst affected countries today are more at risk from the social consequences of clumsy governance than from the population dying off.

Is it mutating?

The possibility (no, sorry, the certainty) of Ebola mutating to become “airborne” is the nightmare that seems to afflict Americans worse than the actual disease afflicts anyone else. I trace this, unscientifically, to Stephen King’s 1978 novel The Stand, which popularized an airborne, weaponized epidemic that escapes from a lab and kills almost everybody. (Life becomes very exciting for survivors, which has ensured the novel’s continued popularity- apocalypse is always more fun than employment) The presence of the Kenema Government Hospital and its partnership with USAMRIID have been a provocative focus for these speculations. Of course, USAMRIID is up to its elbows in creepy and always has been, but…

The facts aren’t really there to support the idea that Ebola has mutated to spread through the air. Just to be clear, virions traveling on air currents into your lungs, without the protective shell of a droplet of fluid, don’t spread the disease. That doesn’t mean that everyone who shares space with an Ebola patient who doesn’t touch them directly is safe. In fact, given how rapidly this disease is spreading without airborne transmission, its hard to see why so much emotional energy is caught up in this fairly technical distinction.

The best argument floated as “proof” that Ebola is airborne is the fact that health care workers are being infected, and surely they would know how to avoid contamination… unless they were being lied to! Images of MSF and others wearing full-body gowns and rubber aprons have created the impression that somehow the virus must be penetrating the very fabric of their protection. In fact, people are wearing PPE only when in contact with patients. The combination of heat, humidity, and the impermeability of the gowns mean that the human body can only tolerate about forty minutes without removing the suit to recover. I trained in Class A suits (slightly worse) in Texas (not as hot or humid) and after ten minutes I literally had standing water in my booties from sweating so much. Ken Brantly has talked about how even during the outbreak in Liberia, he still saw patients without PPE. The current case in Spain, a nurse who was infected while treating a returned missionary, has been blamed by hospital administrators on the nurse’s poor practices, but by her coworkers on gowns that were not impermeable and didn’t meet WHO standards.

A more sinister take on the question of mutation comes from Ebola discoverer Peter Piot, who suggests that in fact Ebola may mutate to become less deadly. Let me explain- diseases that are too deadly burn themselves out. People die before they can infect enough people or, if the disease is highly contagious as well, it infects and kills faster than people can get out of the way. You get a lot of bodies, fast, and in a small area, and then the disease is gone. For a long time, Ebola was assumed to be in this category, and every prior outbreak was in a rural region where mechanism #2 (nobody gets out in time) seemed to be in play. 1.8 is a very low R0. However, the more slowly it kills you, the further you go and the more people you see along the way, and R0 starts to climb. This is good for the virus, so it tends to be selected for. There’s an elaborate debate in infectious disease research about whether this has happened to syphilis, for instance, which when it first appeared in Europe killed everyone it infected within a few years. A slower Ebola would spread further, and be harder to contain, and hence would be a terrible mutation that is, unlike the “airborne” thing, well within the range of possibility.

Is it a colonial thing?

Paul Farmer is 100% right about malaria. We could stop it, but it would cost a lot and we don’t want to spend the money. This isn’t even controversial; the Gates foundation bases its malaria work on exactly this precept. I was surprised, however, to see Dr. Farmer claim that we could similarly stop Ebola if we weren’t facing a failure of will. Joel Achenbach, who interviewed Farmer, makes a few good points about the mission of the CDC, which is to prevent disease in the US, and how that doesn’t always translate to effective intervention in the rest of the world. And Farmer, of course, is right that more hospitals and more isolation wards could slow the spread of the disease (this is in fact the CDC’s sole strategy, but we’ll get to that in the How Does it End? section below). However, I’m inclined to see Dr. Farmer’s claims here as the leftist equivalent of the right-wing claim that “we” need to close off affected countries. It contains a germ of truth, but serves more as a misapplication of an attractive general idea (fighting poverty fights disease) to a situation in which it is perhaps not the best response. When epidemiologists say this outbreak is beyond control, they literally mean just that. It isn’t beyond existing controls, there are no possible measures on order that will place the future of Ebola in human hands.

Not that we shouldn’t be building hospitals and helping with nutrition etc in West Africa anyway.

And not that Africans, for their part, aren’t justified in suspecting USAMRIID (as above) or other players in the spread of Ebola. The list of diseases accidentally or deliberately spread by colonial powers in the third world is quite long, continuing right up to the appearance of a Nepalese strain of cholera in Haiti in 2009. Diseases are weapons, usually quite clumsy ones, and also show up where basic control over daily life has been stripped away by disasters, imprisonment, famine, or slavery. West Africa has suffered all of these, and anyone who knows who Leopold II was knows why West Africans remain suspicious.

We Need Anarchists! I mean data!

Experience in Haiti and after taught me that nobody lies like an entrepreneur, specifically, the guy who claimed he and his software introduced open-source systems to the Haitian earthquake relief effort. Every time there are large agencies involved in a crisis, and a less-than-immediate success, people start agitating for decentralized and open-source approaches to the problem. That’s great, but it tends to ignore the fact that most big NGOs are already working on an open-source basis. The UN had open-source crowd mapping going in Haiti as soon as they got the first tent up. The CDC released its predictive model two weeks ago (its in Excel! its in frigging Excel! what is WRONG with these people?). GLEaM is available for researchers (and uses approximately the same computational resources as a small climate model.) MSF has a fairly permissive data sharing policy (though they have nothing up right now, probably because they’re busy). There are at least two well-developed models on GitHub, by Grant Brown and Caitlyn Rivers. The public health systems of the affected countries are also making crap-tons of data available, which is being mirrored at github and elsewhere. The difficult transportation situation is being addressed by crowd-sourced mapping as well. For better or for worse, the open source revolution already happened, and at least in science the good guys won.

The fact is, the holes in the data- and there are huge, huge holes in the data, from unreported cases to unexplored population structure- have more to do with nobody really having the time to metatag everything in an epidemic. MSF, for instance, collects possible routes of exposure on some or all of their patients, but only using hand-filled paper records. Think anyone has time to put that online? It would be great to know how many people live in each household in the affected areas, but there’s a shortage of demographic anthropologists doncha know. Even “testing” for ebola is a fraught and bootstrapped process; no company has ever scaled up the manufacture of CLIA-waived “ebola testing kits” for the (obvious) reason that its a very rare disease. Testing is by rtPCR, and if you don’t know what that means, it means fuggedaboutit.

So How Does it End?

Of the predictive models so far, only the CDC’s terminates. When addressing an incurable disease, the goal is to force R0 below 1.0, so that each generation of the epidemic is smaller than the last. The generation time for this outbreak is believed to be about 15.3 days, and R0 ranges from 1.4 to 1.8, with outliers on both sides (especially in places like Nigeria and Senegal, which controlled their outbreaks quickly and effectively.) These are the outcomes I can come up with:

Total Infection

Everybody gets sick. 19% of the population survives (incidentally, there’s a bunch of BS about case fatality in the press- at any time in the course of this outbreak, the number of deaths is likely to be half the number of sick people. However, most of those sick people were infected too recently to have died or not, and the epidemic is growing. These factors combine to make it look less lethal than it is. The ratio of currently dead, to infected-sixteen-days-ago, has been 0.81 from the start of the epidemic, with very tight confidence intervals. This is likely the “true” CFR and closer to what we’ll see on the way down.) and everyone else dies. Almost certainly not going to happen; there’s too much else going on.

Herd Immunity

If the average person interacts with a hundred people over the course of their illness, and potentially infects two of them, what difference does it make if 50% of those people are immune? The R0Effective has to take this into account. This is why measles doesn’t kill every unvaccinated American- the high R0 for measles (over 20.0!) crashes into the high vaccination rate to produce an effective R0 below 1.0 For Ebola, a “raw” R0 of 1.4 will still die out if ~71% of people are immune. To get there without a vaccine, of course, 93% of your population has to get infected and 75% of your population dies- the remaining 18% who got sick and survived make up 72% of the remaining quarter. This isn’t much better than everyone getting sick, but it at least shows how not everybody has to die.

Structured Populations

This is why Zulus, for instance, didn’t die of the black death. Not everyone in a population interacts equally with everyone else. Within a household, Ebola spreads like fire, but between households not so much. Between villages, not so much at all. Between countries? Very, very, ineffectively. Eventually you reach a hierarchical structural level where the R0 is either less than 1.0, or adjusts to that accounting for immunity. There’s an argument to be made we’ve already reached that at the country level- there isn’t sweet fark all we can do for Liberia and Sierra Leone, but Senegal, Nigeria, Ghana, and the rest of the world are pretty much safe. Piot (see link above) worries about India, and they’ve had a few watchers, but by and large, structured populations are nearly impossible to model (see GLEaM above) and this is going to have to remain in the realm of guesswork.


The CDC’s Excel-based model terminates some time in 2015 (its a 300-day model) based on moving more and more of the population to hospitals. As noted above, transmission does occur in hospitals, but the R0 is less than 1.0, and far, far below the R0 in the wild. These hospitals don’t exist, but the US and the rest of the international community is building them like crazy. Without a vaccine, this is probably the most sensible response available. Remember that not everyone has to be hospitalized, only enough people that the overall transmission rate drops below replacement. Other guesses generally range from two to four years (Summer 2016-Fall 2018).

It Doesn’t

This is the scariest outcome of all, and the one that popped up in my own clumsy model. Essentially, Ebola reaches a stable island biogeography- an outbreak here, an outbreak there, none lasting more than a year, but never really burning up the planet, but never quite catching the last case or immunizing enough people to contain the spread altogether. Ebola becomes a fact of life, like HIV, which in some countries infects a quarter of the population (about four times the prevalence of depression- do you think about this often enough?) People die, superstitions and stereotypes emerge, the demographers notice a few dips in the population math, and the first world maintains an effective index-case containment system that leads powerful people to believe they deserve, or have earned the right to be Ebola-free. Ebola is subsumed into the mass of misfortunes of poverty for which those with the power to respond (treatment options for Ebola are scarce and expensive, but given three decades we could be on top of it if we tried) feel no responsibility. A convenient pretext to close inconvenient borders, seize mineral rights, dismiss religious movements, and otherwise opportunistically rob the world. That, my dears, is the zombie plague you should worry about.


It has been brought to my attention that I flubbed some of the numbers in the herd immunity section. For an R0 of 1.4 the cutoff is actually 29% immunity, meaning only 68% of your population needs to get sick, 55% die, and the recovered 13% are 29% of the total survivors. I don’t think that changes the import very much, but it does teach me not to use the microsoft calculator on the fly.


21 thoughts on “Ebola Ebola

  1. Pingback: Ebola, or: How Halloween arrived early this year | Exopermaculture

  2. That was a terrific summation of the current Ebola situation. Thanks very much. Some points to ponder….

    Ebola R_0 is pretty much determined by individual behavior at the critical moment. Patrick Sawyer’s expedition resulted in R12 but the runner to Port Harcourt R3 and so on. I think the fact that the mobile infective stage of Ebola is so short is the operative variable and the reason that disease transmission seems like a slow motion train wreck.

    I have read that we are entering the time of year when large numbers of male west african populations go on the move to new seasonal jobs in neighboring like harvesting Cacao and so on. On foot. The author feared that this would result in new outbreaks in currently unaffected surrounding countries. If so, this disease will be with us for a very long time and your scary outcome becomes more likely. Global pandemic and endemism. very bad.

    Lastly, we should consider the possibility of (and test for) new disease reservoirs/vectors like pigs, other human domestic animals and community associates like lice, bedbugs, etc..


      • Yes, OIE, supposedly. I don’t see much discussion of the topic anyplace though and it obviously has important implications for the spread of endemism and agriculture.

        The models you link to are already out of date and were grossly inaccurate in any case, as models with bad source data often tend to be. There is nothing inherently wrong with using Excel to model and nothing that requires more complexity in modeling disease outbreaks in order to predict outcomes more or less accurately. They also provide the unusual step of matching the model with empirical data on a continuing basis. Looks good to me. Especially considering that field data is said (by WHO & NGOs) to be 1/2 to 1/4 of the reality. Here is the link, if you want to take a look and run it yourself.


  3. A couple things:
    Measles doesnt kill everyone not primarily because of herd immunity but because the case fatality rate is no higher than 15%.
    Second, their are surely likely to be important population level differences in Ro for ebola are their not?


    • Oops. Yeah. Try “infect” rather than “kill.” Also, I think the CFR is under 15%, with SSPE being a greater risk, but I don’t feel like looking it up right now as I’m late for something already.

      As for “population-level-differences” yes, but I’m not sure what you mean. Behavior is a major mediator of contagion, which means that any time behavior changes (say, everyone gets nervous about shaking hands) the R0 changes. I’m sure there are population-level differences between how often people lean against cab windows, or use non-disposable hankies, or pretty much anything else. Then again, there are big inter-individual differences in the same thing, and the biggest effect on a single case R0 is probably random chance. Hence, its a post-facto historical calculation on large numbers of cases that gives you some guidance, but not much, about the future.

      I personally think the biggest data hole here is the effect of structuring- people interact differently within their family, with their friends and neighbors, etc. So far at least three countries have had single-index outbreaks with autochthonous transmission (Nigeria, US, Spain) and all seem to have controlled the spread far better than, say, a household in an endemic region would if Uncle Joe came to stay and got sick; this is just because countries are bigger and able to quarantine, whereas families are compelled to help each other. Again, this is really, really hard to model.


  4. Did I miss discussion of the importance of asymptomatic infections in this fine article? Some estimate suggest that as many as 70% of the infections are asymptomatic, and if so this will dramatically alter the course of the illness through much more rapid herd immunity.


    • Really? I’d like to see a source for that. The fellow who buries all the dead in Bong county, Liberia, is notable (i.e. made the ISID email list) precisely because he seems to be immune and has never been symptomatic. To my knowledge nobody has ever done an antibody screen on the never-ill in affected regions, but the general understanding seems to be that asymptomatic/subclinical infections almost never happen. I did hear speculation that the C.A.R. has been spared due to the high rate of immunity, but I don’t think any data was ever produced. Would be a game changer if it did though, that’s for sure.

      On another note, have you seen the evidence (here, or here, or here) for subclinical human infections with rabies? Holy cow!


      • Heffernan RT, Pambo B, Hatchett RJ,
        Leman PA, Swanepoel R, Ryder RW. Low
        seroprevalence of IgG antibodies to Ebola virus
        in an epidemic zone: Ogooué-Ivindo region,
        Northeastern Gabon, 1997. J Infect Dis 2005;
        191: 964–68.
        Leroy E, Baize S, Volchkov V. Human
        asymptomatic Ebola infection and strong
        inflammatory response. Lancet 2000;
        355: 2210–15.


      • Okay, fascinating. Thanks for these- they do, in fact, change everything. I think you may be a bit wrong about the numbers- the implication here is that a large number of people with antibodies were never symptomatic, but remember that most people who were symptomatic are now dead, so the fraction of infections that lead to subclinical immunity is still fairly low. I can run the numbers if you’d like but I think the CIs would be wide.

        Most interesting, though, is that they separately tested the different Ebola peptides for antibody response; it puts the genomics work of Sabeti’s team in an even stronger light. I wrote up a brief on the Gnirke et al paper that just came out (moment of silence for the five authors who died collecting data) but I’m probably not going to post it- this Ebola thing has gotten beyond my usual blog management abilities and I want to move on. Still, I recommend google scholaring it- really brilliant informatics work in the hardest of hard situations. I think its open access, or at least it should be.


  5. There was some field work done on Zaire Ebola seropresence in Pygmy populations of central Africa. I think the numbers were ~30% seropositive (significant longterm Ebola endemism?). And of course the Reston outbreak numbers were 100%.

    [Microbiological surveillance: viral hemorrhagic fever in Central African Republic: current serological data in man].
    Nakounné E, Selekon B, Morvan J.
    Bull Soc Pathol Exot. 2000 Jan;93(5):340-7. French.
    PMID: 11775321 [PubMed – indexed for MEDLINE]

    Ebola and Marburg virus antibody prevalence in selected populations of the Central African Republic.
    Gonzalez JP, Nakoune E, Slenczka W, Vidal P, Morvan JM.
    Microbes Infect. 2000 Jan;2(1):39-44.
    PMID: 10717539 [PubMed – indexed for MEDLINE]

    Filovirus activity among selected ethnic groups inhabiting the tropical forest of equatorial Africa.
    Johnson ED, Gonzalez JP, Georges A.
    Trans R Soc Trop Med Hyg. 1993 Sep-Oct;87(5):536-8.
    PMID: 8266403 [PubMed – indexed for MEDLINE]


  6. Pingback: Trying Again | More Crows than Eagles

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s