Quarantine!

by hilzoy

On October 4, President Bush said this at a press conference:

"The policy decisions for a President in dealing with an avian flu outbreak are difficult. One example: If we had an outbreak somewhere in the United States, do we not then quarantine that part of the country, and how do you then enforce a quarantine? When -- it's one thing to shut down airplanes; it's another thing to prevent people from coming in to get exposed to the avian flu. And who best to be able to effect a quarantine? One option is the use of a military that's able to plan and move. And so that's why I put it on the table. I think it's an important debate for Congress to have."

When he made that remark, I thought: the idea of using quarantines in the face of the threat of disease is exactly the sort of idea that might occur to some people for bad reasons, and perhaps be opposed by others for equally bad reasons, especially in the face of an emerging infectious disease. Just think back to the early 1980s, when AIDS first hit the news: there were all sorts of calls for quarantines; kids with AIDS were prevented from going to school or, in one case, allowed to sit in a glass box in the classroom; and so on and so forth. This was a completely inappropriate reaction to AIDS: for reasons that will become clear later, AIDS is a terrible candidate for quarantine. Nonetheless, they were very common then (and some people still advocate them to this day.)

It seemed to me that it would be a very good idea to write something about quarantines and the circumstances in which they can and should be used, so that as many people as possible outside fields like public health will already have thought about them before the need arises and emotions get heated. That way, there will be more people scattered about the general populace who can assess calls for quarantines if avian flu or some other new and dangerous infectious disease hits. And the more such people there are, the less likely we will be to do something stupid.

Definitions: When you restrict the movement of a person who already has a disease to prevent her from spreading the disease, e.g. by keeping her in a hospital isolation room, that's isolation. When you restrict the movement of people you suspect might have a disease to keep them from spreading it, that's quarantine. (The word comes from a practice of keeping ships offshore for forty days before allowing them to enter a port, so that any infectious disease would have a chance to reveal itself.)

The Law: The President has the legal authority to do more or less whatever he, in consultation with the Secretary of the Department of Health and Human Services and the Surgeon General, thinks necessary to prevent an epidemic. Moreover, most states have quarantine laws of their own.

CharleyCarp was kind enough to put some legal research on quarantines on his blog, for which I was and am incredibly grateful. Money quote: "the Fourteenth Amendment was not thought then, or previously, to prevent quarantine of someone actually infectious." He then provides very interesting cases.

The Rest: There is always something to be said against a quarantine. Quarantines deprive people of freedom of movement, after all, and that's not something you'd want to do without a very good reason. (Especially not if you're considering using the military to enforce a quarantine: shooting citizens is a very, very serious step, and it should only be taken for very, very good reasons.) The burden of proof, therefore, should always be on those who want to establish a quarantine.

Whether that burden is met depends, I think, on two things. First, how serious is the problem you're trying to avert? Second, how likely is it that a quarantine will actually avert it? As to the first: even a huge risk that some hypothetical form of infectious dandruff-causing disease might become established here would not justify a quarantine: dandruff is just not serious enough. But avian flu obviously is. If there is a good chance that a quarantine would contain the spread of avian flu in the US, then I think there would be a serious case for imposing one.

But this is ONLY true if there is a good chance that a quarantine would, in fact, work. If it wouldn't, then you incur all the considerable costs of imposing a quarantine without getting any of its benefits at all. And that would just be stupid: exactly like trying to stop an influenza pandemic by walking around saying "go away, you silly virus!", only with much, much greater costs.

So: would a quarantine against influenza work? How can we tell? The very, very short answer is: quarantines do best when several things are true. First, the disease is not all that infectious. Second, people become infectious only after they have developed symptoms. (That way it's easier to know who the infectious people are.) Third, you do a good job of containing all the people you want to contain. Finally, the disease progresses relatively slowly. When any one of these is false, it becomes a lot harder to contain a disease. When several of them are, it's almost impossible. Here's why.

One way to think about quarantines and when to use them is as follows. Epidemiologists have a useful concept: the reproductive rate of a disease, or R. R has several variants. R(0) is the 'basic reproductive rate': the average number of new cases of a disease that a given person with the disease will give rise to in a population of people who are all susceptible to that disease. (What epidemiologists call 'a virgin population'. Think of the introduction of smallpox to the Americas, or an outbreak of a new form of flu.) Similarly, R(t), the 'effective reproductive rate' at a given time, refers to the average number of new cases that each infected person gives rise to at a given time. (It differs from R(0) in that not all people are necessarily susceptible at that time, and also in that control measures against the disease may be taken.) Thus, if R(t) for a given disease at a given time is 2, then each person who gets the disease at that time, on average, will infect two new people with it.

[Update: If R(t) went on being 2 forever, that would obviously be really bad. One person gets the disease and infects two; each of them infects two and so we have four more; each of them infects two and so we have eight more; a little of this goes a long, long way. But in epidemics involving diseases that confer immunity, like measles or a given strain of flu, the more time passes, the more people are immune (or dead), and so R(t) will tend to drop of its own accord. If R(t) is 2 at the opening stage of an epidemic, that's not nearly as bad as it could be.]

R(t) is not the sort of thing that is fixed in stone for a given disease. It can be affected by a number of things, including:

• How many people the average person with the disease encounters in a way that can spread infection (e.g., for STDs, how many sexual partners the average infected person has; for flu, how many people the average infected person gets close enough to spread the virus to; etc.), for some unit of time
• How many of those people are susceptible to the disease (thus, people who are immune to the disease, because they have already had it or been vaccinated, are not susceptible to it.)
• How many of these people, on average, actually get infected (or: how efficient the virus/bacterium/etc. is at taking advantage of opportunities to infect others)
• How long the infection lasts.

So, for instance: when AIDS first hit IV drug users, its R(t) went up, because sharing needles is a pretty efficient way of spreading HIV. When a disease spread by coughing finds its way from, say, rural Kansas to New York subway users, its R(t) goes up, since the average number of people an infected person will have a chance to cough on is a lot higher when those people include New York subway users than when they include only Kansas wheat farmers. In an epidemic, the proportion of people (in a given population) who are susceptible to a disease that confers immunity goes down over time, since more and more people have already had the disease and are immune. If you vaccinate the entire country, R(t) in that country goes to zero, since nobody is susceptible anymore. (OK, not exactly zero, since there are borders, and people cross them, but you get the point.) And so on.

If you want to control an infectious disease, you want to bring R(t) below 1, since if it's below 1, then on average each person with the disease will infect fewer than one new person, and eventually the disease will die out. That's the goal. (And that's why I went on about R(0) and R(t): it converts the fairly vague problem 'controlling a disease' to a much more precise one: get R(t) down below 1.)

You can lower R(t) in any number of ways, by affecting any of the factors listed above. Something as simple as a successful campaign to get people to wash their hands a lot can bring down R(t), for instance, by affecting the efficiency with which a virus or bacterium takes advantage of contacts. (More handwashing, less efficient transmission.) Quarantine is a way of trying to bring down R(t) by bringing down the number of contacts infectious people have, by preventing them from moving about. (Also: by confining the infection within a limited area, you make it more likely that the people an infected person comes into contact with will already have, or have had, the disease, which lowers the proportion of susceptible people she comes in contact with.) So the question is: in what circumstances is it likely to work?

Basically, the answer is: it depends on what R would have been without the quarantine, and what proportion of the infected people your quarantine manages to catch. If, for instance, you are trying to stop the spread of a disease, and R is only slightly above 1 to start with, then you have room to miss quite a few people, and even if your quarantine is not very effective, it can still control the disease. (This is especially true given what this book describes as "the inconvenient fact that humans are born, infected, and die as integer units (rather than in a continuous and deterministic fashion)" (p. 20).) If, on the other hand, R is high, then you have to have a much more effective quarantine in place, since that quarantine has to get R down from a much higher number.

[Update: Imagine, for a moment, a disease whose baseline R(0) was something really scary, like 100,000: on average, each person who gets infected will infect 100,000 other people, and each of them will infect 100,000 more, and so on. You would need to have an absolutely 100% effective quarantine to do any good at all, since if you miss even one person, huge numbers of people will be infected. As I said before, the only case in which a quarantine is justified is if it works. Given a disease with a huge R, a quarantine is almost never justified.

On the other hand, imagine that R(0) is 2. As I said earlier, if you leave this disease unchecked, it will spread quickly: one person will infect two more; those two will infect four more; and so on, exponentially, until enough people become immune or die. But here's the good news: to get this disease under control, you only have to block a little more than 50% of new transmissions. That's a decent margin of error, and one that some quarantine might actually meet.]

As an example, consider pandemic flu. Its R(O) -- the reproductive rate in a population with no immunity to the disease -- is quite high. In this article from the Proceedings of the National Academy of Sciences (unfortunately, behind a subscription wall -- but hey, it's better than the Mirror!), it is estimated to be anywhere up to 25. (They don't estimate a lower bound, since they are talking about pandemic flu, which necessarily involves a whole new strain. Most sources I've read that guesstimate the lower bound put it at around 5.) That's very, very high. In order to bring R(t) down to zero, you would have to quarantine almost everyone who has the flu.

And that, it turns out, is almost impossible to do. One reason is just that quarantines are not 100% effective outside very special circumstances, like (for instance) quarantining the passengers of a plane. In that case, you already have all the people you want in an enclosed space, and keeping them there is fairly straightforward. Quarantining a city, by contrast, is really hard: there are lots of ways in and out, and it's really, really hard to police them all.

But there's a much more important reason, one that no amount of administrative skill will get around, and it has to do with the biology of influenza. It turns out that people can transmit the flu before they develop any symptoms. Not only can they do this: the aforementioned PNAS article estimates that 30-50% of transmission occurs before the person doing the transmitting develops any symptoms at all.

The appearance of someone with symptoms of a disease is generally what prompts people to say: hey, let's isolate this person, so that he or she cannot infect anyone else, or, alternately: hey, let's quarantine this person's neighborhood, city, or whatever, so that the disease can be contained. But if 30-50% of that person's transmission has already been done, then you are not going to catch anything like all the people that person infects by isolating him or her after she develops symptoms. Nor will you do so by quarantine: people travel, and by the time someone in a given city develops symptoms of the flu, she will have infected others, and those infected, asymptomatic people will have had a couple of days to disperse themselves and find whole new cities to infect.

This would not be such a problem if influenza's R(0) were low. After all, you only need to bring R(t) down to 1, and if it were only 1.5 to start with, catching 50% of new cases would be fine. (Each new case won't infect very many people anyways, so it's less of a problem when some get away.) But since R(0) is likely to be anywhere from 5 to 25, the fact that lots and lots of people are not going to be caught by a quarantine means that that quarantine won't work.

There's one more nasty feature of influenza that makes quarantines even less likely to work: it's very fast. The amount of time between someone's getting infected with a disease and that person's becoming infectious to others is, basically, the time you have to track them down before they start spreading the disease. If influenza had a nice long period in which people were not yet infectious, then it would spread a lot more slowly, and there would be more time for measures like: figuring out who each person with the flu had come in contact with, and trying to find and isolate those people. Unfortunately, it spreads very quickly, which means that it is likely to outrun most attempts to catch the people who have the flu but don't know it yet.

In general, the possibility of containing a disease through quarantine and isolation seems to depend a lot on several things: first, R(0) (the basic reproductive rate); second, the proportion of new infections produced by an average person that occur before symptoms appear; and third, how many of the people quarantine and isolation are trying to capture they actually succeed in capturing. (Where 'the people they are trying to capture' does not mean, e.g., 'all the infectious people' -- in that case this third point would include the second -- but rather something like: 'the people with symptoms, and all their previous contacts', or whatever similarly defined group you're aiming at.)

This is why quarantine and isolation did help to contain SARS. SARS had a much lower basic reproductive rate than influenza -- it's estimated to be around 2-4. People with SARS are not infectious before their symptoms appear, and their peak levels of contagiousness seem to occur several days after the appearance of symptoms. This means that if you get the symptomatic people, you've got the infectious people, and also that you have more time to find their contacts, and more reason to hope that those contacts will independently present themselves at a hospital before they get infectious. All of this worked in favor of attempts to control SARS. By contrast, just about everything about influenza works against them.

Here's an illustration from the PNAS article that makes the point clearly. It summarizes the results of applying a rather complicated mathematical model to SARS, smallpox, influenza, and HIV. (The lines are different containment measures and/or assumptions; what's above a line is not contained, and what's below it is.) The top picture shows what happens if you isolate symptomatic individuals, trace back their contacts, and quarantine them, and if your efforts are 100% effective (as they would obviously not be.) Given this assumption, influenza (the red blur that comes down from the top in a sort of bar shape) can be contained. But if the effectiveness of your isolation and contact tracing program falls to 90%, which is pretty high, it is probably not. And bear in mind that the sorts of isolation and contact tracing that the illustration envisages are a lot more focused than quarantining a city.

Description (from article): "Black lines correspond to isolating symptomatic individuals only. Colored lines correspond to the addition of immediate tracing and quarantining of all contacts of isolated symptomatic individuals. The black (isolation only) line is independent of distributional assumptions made (low or high variance), whereas the colored (isolation + contact tracing) lines match the variance assumptions made in Fig. 1 (red = high variance; blue = low variance). The efficacy of isolation of symptomatic individuals is 100% in A, 90% in B, and 75% in C."

***

There are a lot of preparations for avian flu that we desperately need to undertake. (Rebuilding our public health infrastructure is an obvious one, as is doing some really serious planning.) Some responses to avian flu, for instance the maintenance of order, might for all I know involve the military. But using the military, or anyone else for that matter, to quarantine a neighborhood or a city will not work against the avian flu. It might even make things worse: if people react to the imposition of a quarantine and the deployment of the military by fleeing in panic, they could end up dispersing the disease more widely than it would have been dispersed without a quarantine.

As I said earlier: adopting a policy with very serious costs is only justified if you can expect it to yield very serious benefits. Quarantines have very serious costs. These might be justified if a quarantine prevented a serious epidemic. But if they are not going to work, then there is no reason at all to adopt them, and all the reason in the world not to.

After all, if we want to take completely ineffective measures against an epidemic, we have a whole universe of possibilities to choose from. Stopping the avian flu by making everyone eat a daily serving of ice cream, for instance; or by urging people to take their loved ones out for special protective romantic dinners; or by issuing everyone a special protective pony. Given these alternatives, each of which would work just as well against the avian flu as a quarantine, why on earth would we choose quarantine?

Comments

[hilzoy]

Yes: it's the official Penitent Wonky Thread ;)

[Frank]

Hmm so martial law and quarantine woundn't help contain an Avian flu pandemic. I wonder if Bush might have another reason for wanting to declare martial law...

[Jackmormon]

My reading on quarantines doesn't go much past 1850, but in what I've read, the well-connected and moneyed rarely had much trouble getting out of the quarantined areas, with or without symptoms. For a truly harrowing account of slipping in and out of a quarantined area, I recommend Charles Brocken Brown's Arthur Mervyn, based loosely on the author's experiences in the 1790 yellow fever epidemic in Philadelphia.

And if we're going to advocate for cosmestic epidemic preventatives, why not suggest healthy diet? The 19th-c hygenists loved that one.

[hilzoy]

Out of curiosity, was this comprehensible?

[Gary Farber]

Wait, no one said there was math involved!

[Jackmormon]

I could follow it, although it took some concentration. Then again, I've done a bit of history-of-science type reading on epidemiology. The reproductive-rate calculus was new to me (beating on self for being a lazy researcher), but I think I came to understand its importance over the course of the post. The only suggestion I could make would be to add a sentence after

Thus, if R(t) for a given disease at a given time is 2, then each person who gets the disease at that time, on average, will infect two new people with it

to suggest whether that given time as 2 would justify full alarm bells or not. Later in the post, it becomes clear that 2 is not an acceptable standard of transmission, but not one justifying quarantine. Still, that was the point at which I realized I was in over my head.

[Jackmormon]

Oh, I meant to add: if you could give more examples of r(t) in familiar epidemics (TB, yellow fever, black plague), people without specialized knowledge could relate better.

[hilzoy]

I updated, to account for your first point. The second is harder, since R(t) is variable over time, within a given epidemic.

[Francis / BRGORD]

so, quarantining will work for influenza if (a) the efficacy of isolation of symptomatic indivduals is 100%, there is tracing and quaranting of contacted individuals and the flu has low variance or (b) the efficacy of isolation is 90%, there is tracing and quarantining of contacted individuals, the flu has low variance and, in addition, the R(0) is around 8.

you could explain high vs low variance in greater detail. otherwise, good post.

instead of a pony, i'd like some mission style furniture.

[Jackmormon]

I think the update is pretty good, although it doesn't confer the point that if r(t)=25, more people will be dead than immune. What the average layeperson is looking for is an understanding of what the scale means, from bad to awful.

Should I shut up now? (My day job has been freshman writing for years now; don't encourage me.)

[hilzoy]

Jackmormon: r(t) just tells you how many new cases the average infected person produces. Whether they die or become immune depends on the severity of the disease.

Thus: imagine my hypothetical infectious dandruff. Even if it had r(t) of a million, so more or less the whole world came down with it in no time, all they'd get would be dandruff. Then, when the dandruff disease had run its course, they'd be immune.

By contrast, almost no one ends up immune to ebola, because almost everyone who gets it dies. This would be true even if it was really, really hard to get, so that you had to e.g. roll around in the blood of ebola victims after undergoing scarification rituals, or something. (In that case, there would be the question why the disease is still around, so let's assume it normally hangs out in some animal or other. As in fact ebola does, though last time I checked, no one knew which.)

[hilzoy]

Oh -- and I did one more update, for clarity.

[rilkefan]

Umm, if a population center is full of R=10**5 people, wouldn't the right response be not to quarantine it but to nuke it? Or is that already a plot point in 24?

[Anarch]

Umm, if a population center is full of R=10**5 people, wouldn't the right response be not to quarantine it but to nuke it?

Only if the nuke destroys and doesn't disseminate the virus (e.g. trapped in particulate fallout). Which, now that I think about it, is probably next season's plot point on 24.

[liberal japonicus]

Can I just say, I love the exclamation point in the title. Reminds me of that John Wayne film Hatari!

[otmar]

Btw, this is yet another example where I get the impression that Bush has read too many Clancy novels.

[Anarch]

One other thing that might help is a brief discussion of the nature of exponential growth (which is what constant R implies). For example, to use the simple doubling scenario: assuming that every person infects two other healthy people (R = 2) during one "iteration", it will take approximately 28 iterations to infect the entire population of the United States. At lower R, say R = 1.5, it'll take approximately 48 iterations; at higher R, say influenza's worst-case R = 25, it'll take only 6 iterations. Assuming an "iteration" (infection->new infection cycle) of about 3 days, this means that R = 1.5 gives us around 6 months to annihilation, R = 2 gives us about 3 months and R = 25 gives us around 2 1/2 weeks.

This is obviously a first-order estimate -- to be more precise you'd want to use, say, logistic growth [y' = ky(1-y)] instead of exponential growth [y' = ky] to simulate dampening, and that's still horrifically crude -- but it should help give people a sense of the time-scales involved.

[Tim]

But for evaluating the efficacy of a quarantine probably first order is fine. By the time you need to factor in dampening you're almost certainly SOL.

Understandable post to me. You could add that diseases that are quickly fatal usually give lower R values.

[john miller]

This is the kind of post that makes me wish I had paid more attention in my science and math classes 40 years ago.

But I think I understand what you are saying.

Two points. You talk about the need for planning beginning now. As has been pointed out on other threads lately, one of the apparent failings of the current administration is failure to paln for anyhting other than roses being tossed at our feet.

Secondly, I think it is important to look at the psychological effects of a quarantine, both good and bad. Could be bad for those areas placed under quarantine, as you mention the possibility of panic, etc. However, at the same time, it may be positive for other areas who sense that their government is actually doing something.

Then again, they may feel that since they are not quarantined they have nothing to worry about, which would be bad.

Anyway, my point, for which I don't have a definitive answer (way too early) is what would the psychological inmpacts of a quarantine be, and would they offset either the positive or negative practical consequences of one?

[tonydismukes]

Excellent post, hilzoy. Very informative.

(I do have a decent general math & science background, so the presentation was at exactly the right level for me to follow quickly.)

[Jake - but not that one]

I thought the post very clear. The addition I would make is not to add math (which is not a problem, I just don't think it useful here) but to discuss, at least briefly, the alternatives to quarantine. As it is, the article comes across as "don't quarantine, just pray" or die, or whatever.

This thought stems from your opening - a few good and thoughtful people early in the debate can make a difference. Well, they can make a difference if they have a plan.

It's that old plan thing, again. :)

"Hey, if you're so smart, tell me what we SHOULD do, not just what we shouldn't do!!!"

Jake

[Slartibartfast]

Admittedly just skimming, but R(t) seems to be a sort of average rate of retransmission given unfettered movement. Wouldn't it be more meaningful to examine what happens to R(t) under quarantine? What does effectiveness of a quarantine mean, other than the degree to which it affects R(t)? Is it reasonable to suppose that a quarantine that reduces R(t) from 100k to 0.5 is awfully difficult to execute? Consider this more of a questioning of how linear R(t) is with effectiveness of quarantine than a failure to pay attention, although the latter is certainly everpresent.

Also, it seems I recall reading that most of the deaths in various flu pandemics were due to secondary infections coming on top of the flu sledgehammer. But I'm too fargan busy to Google.

[Tim]

The alternates as I understand it are to use vaccine and antivirals to firebreak any outbreak (we don't have the vaccine or antivirals yet); or to isolate any outbreaks in hospitals (we ditched the spare hospital capacity to make this possible to save money long ago). So proposing quarantines is the only solution that's not immediately going to be seen as bogus.

[hilzoy]

Slarti: R(0) is retransmission given unfettered movement and a virgin population. R(t) is retransmission in whatever circumstances actually prevail at t. Quarantines are an attempt to affect R(t), and specifically to lower it below 1, so that the epidemic extinguishes itself.

The relevance of the baseline R(0) is: what a quarantine (and whatever other measures) need to do is to move R from R(0) (which is roughly R at the beginning of an outbreak of a new disease) to less than 1. So noting R(0) gives you an idea of how effective a quarantine would have to be to contain the disease. If R(0) is 1.5, then the answer is: not all that effective; even a pretty leaky quarantine would work. If the answer is 10, on the other hand, then the answer is: only a really, really, REALLY effective quarantine would do the trick.

When you then add in the fact that flu is very infectious before symptoms appear, and thus before people who have it can be identified, it looks way unlikely that any quarantine could be nearly effective enough to contain the flu.

And then add in the speed. -- Suppose that the flu worked very, very slowly: e.g., that it took a year for someone who had gotten the flu to become infectious. Then you'd have a whole year to locate and isolate that person before he or she became infectious. You could, for instance, try to find everyone who was in an airport that an infectious person was in, or who was in the same subway car, or whatever, and have some hope of finding most of them. But if you only have a couple of days, and people who are very infectious don't necessarily give any outward sign of having the flu, this looks a lot less feasible.

[liberal japonicus]

This article was interesting and had this

The Public Health Service Act gave the U.S. Public Health Service responsibility for preventing the introduction, transmission, and spread of communicable diseases from foreign countries into the United States. Under its delegated authority, the Division of Global Migration and Quarantine is empowered to detain, medically examine, or conditionally release individuals and wildlife suspected of carrying a communicable disease. The list of quarantinable diseases is contained in an Executive Order of the President and includes cholera, diphtheria, infectious tuberculosis, plague, smallpox, yellow fever, and viral hemorrhagic fevers, such as Marburg, Ebola and Congo-Crimean.

Here's the list and the FAQ for the update for bird flu.

[Tim]

H5N1 shares the characteristic with the 1918 flu that it can kill by itself. There's a some evidence that it is more dangerous to young adults than children or seniors.

[Jake - but not that one]

Ok, what happened? Did the day before thanksgiving exert some kind of mind control over everyone, or what?

Or did y'all decide posting was over for the holiday and nobody told me?

Jake

[Gary Farber]

"Or did y'all decide posting was over for the holiday and nobody told me?"

Yes. We're all meeting over at the other place, to do the thing.

You'll qualify next year.

But you can come over to my place around 8, and watch some DVDs with me, if you'd like.

[Slartibartfast]

Actually, I composed a rather lengthy reply to hilzoy whose contents currently escape me (probably something along the lines of, if quarantine can knock R(t) down to 5 or so, and if things like masks and improved cleanliness habits knock it down to 2 or so, and increased paranoia knocks it down to 0.8 or so, is quarantine still worth it?) but then I realize what I know about this issue would probably fill a thimble up to the lowest row of dimples, at most, and so I should probably shut up. Plus, I got one of those you-may-be-a-spammer pages, copied the text because I couldn't get it to post, and then promptly copied something else in another application, which wiped the whole deal out. Or something resembling that.

[Jackmormon]

Slartibartfast, I'm sorry to hear that the software suspects you of spamming, seeing as though I'd like to see more of you around here. Post about your health, post about your garden--I'm interested, if that can serve as a baseline for anything.

Thanks, Hilzoy, for the updates, which do help to clarify how epidemiologists understand risk. I can only hope that if a flu pandemic does occur, sensible and articulate epidemiologists find their way on air.

In other, more self-centered news: should I get a flu shot this year? I've never gotten one before. I'm a generally healthy 28 year-old, who had until recently always thought that flu-vaccinations were for the elderly and immune-impaired. I know full well that the current vaccination dose won't protect against the feared Avian Flu strain. Is there any point besides making my mother happy?

[liberal japonicus]

Is there any point besides making my mother happy?

Does there need to be any other point? ;^)

[hilzoy]

Jackmormon: yes, there are two points to getting it, besides the usual. (1) The flu is serious in any case; if avian flu hits, you do not want to have it after already being weakened by the regular flu, or conversely. (2) Flu viruses have this delightful capacity to swap genes with one another when two separate strains find themselves in the same host. Sometimes this results in a virus that just doesn't work, but sometimes it results in a virus with properties of avian and human flu. And one possible combination is: the deadliness of avian flu; the easy transmission among humans of the normal flu. By getting the flu shot, you prevent this from happening in you.

[hilzoy]

To be clear: point (2) above is less about benefitting you than about doing your part to prevent an unlikely but by no means impossible event that would greatly harm others from happening in your body.

[Jackmormon]

Okay, internet moms (half the neuroses, twice the statistics!), I'll get my flu shot.

Good thing I live in NYC, where the mayor has donated obscene amounts of money from his obscene personal wealth to finance public health. Between his philanthropy and my university's outreach, I should be able to find a vaccination that doesn't make me feel like I'm stealing health from the poor.

[hilzoy]

And Slarti: I don't think that will work with the flu. One reason is that when handwashing etc. cut transmission over a population, the extent to which they do so is a function of the average effectiveness of the hygiene measures and the percentage of people who are convinced to adopt them. And you'd have to have very, very, very effective hygiene measures to make up for the fact that they would in all likelihood not be adopted by everyone. This is especially true for flu, which is spread not just by hands but by airborne droplets. You'd need everyone wearing masks and gloves or something.

Also, think of it this way. Anarch described exponential growth above. In fact, R(t) isn't constant, and so growth isn't really exponential. But it comes close at the outset of an epidemic, when everyone is susceptible; it tails off later, as the number of people who have had the disease becomes significant, and so the number of susceptible people an infected person comes in contact with shrinks even if the number of people she comes in contact with is constant.

Now: what you try to do with a quarantine is: confine the disease to a particular area, leaving the rest of the country untouched. Anyone who escapes from the infected, quarantined area into the surrounding, not yet infected area will set off basically exponential growth. (Which is another way of saying why R(0) is important: it tells you what number is being exponentially increased.) If R is low-ish, you have time to work; but if it's high, the effects of even a relatively small number of escapees will pretty quickly come to approximate the number who would have been produced w/o a quarantine (since the small number of initial cases at large will be compensated for by the fact that with more escapees or no quarantine, the number of susceptible people begins to drop more quickly.)

Did that make any sense?

[hilzoy]

Oh: the "also, think of it this way" para. is on another point from the one that preceded it: it's not on handwashing, but on the question of the effects of people who elude quarantines when R(0) is high.

Jackmormon: not being particularly neurotic, I prefer stats any day. (My parents were delightfully casual about things like my making explosives as a child -- my mom had read Emile -- so I never really got into obsessive worrying.)

[CaseyL]

About whether to get a flu shot: I've noticed, over the years, that I get fewer winter colds when I've had a flu shot. I have no idea why, unless the shot somehow boosts anti-viral immune response in general as well as specifically against flu bugs. So, yeah, I'm a big advocate of flu shots.

My feelings about the possibility of a deadly avian flu epidemic or pandemic, though, are that we might be getting a wee bit more worked up than is necessary. The factors that made the 1918 pandemic so catastrophic were particular to that era. The global population just isn't as isolated into discreet sectors as it was then; the international epidemiology response network we have now plain didn't exist in 1918; and we know immensely more now about how to isolate a virus and make a vaccine for it quickly.

I'm not counseling complacency. I'm just saying that scenarios involving quarantining entire cities, while public health infrastructures collapse utterly, are a long shot.

[trilobite]

Has anyone else noticed that Bush really really likes using the military? He seems to think of it as a sort of Swiss Army Knife, good for invading, occupying, nation-building, interrogating, jailing, disaster relief, and now for enforcing quarantine. Perhaps Condi Rice or someone could tell him that there are other tools available to the government.

Oh, but wait, that would be Big Government. Can't have that. Much better to do all the same things Big Government does, for more money, less effectively, and concentrating all the power in one institution--with guns. Yes, I see now how much more sensible that is.

[John Thullen]

Trilobite, I suspect that if you rooted around in the conservative think tank literature of the last 30 years, you would find recommendations that the military be used for all manner of exigencies, the better to avoid the truly awful, world-ending occurrence of unsunseted bureaucracies, funded by that thing we apparently must protect ourselves against -- worse than millions dead from avian flu or nuclear war -- taxes!

Speaking of rooting around and birds, this paragraph appears today in this obituary for one Ruth M. Siems, the inventor of Stove Top stuffing:

"Stove Top's premise is threefold. First, it offers speed. Second, it divorces the stuffing from the bird, sparing cooks the nasty business of having to root around in the clammy interior of an animal. Third, it frees stuffing from the yoke of Thanksgiving; it can be cooked and eaten on a moment's notice any day of the year."

Clearly, This invention, Patent #3,870,803, along with the birth control pill and political correctness, was the beginning of the end for Western Civilization.

Happy day, all. I must now engage in the nasty business of the day, yoked that I am.

[Gary Farber]

"...this paragraph appears today...."

Assuming today is November 22nd.

[Gary Farber]

"Trilobite, I suspect that if you rooted around in the conservative think tank literature of the last 30 years, you would find recommendations that the military be used for all manner of exigencies...."

As someone who has as one of many hobbies doing such rooting around, in point of fact, one will find in such lit endless papers on the dangers, horrors, and ill outcomes of such use, and they all tend to have the point that one should never, ever, ever, use the military save to win battles against other armies, and then get the hell out.

There are some exceptions, but, as it happens, Meester Bush simply goes against the vast accumulated wisdom of said papers in Iraq. He was going with the flow back in the 2000 campaign, when he decried nation-building and using the military for non-military purposes.

Not that I want to get in the way of freeform fantasizing. Especially not that I want to discourage thullenosity.

[psh]

'was this comprehensible?'

Yeah, it was great. It would help to see the contagion model used, at least in some closed form. One confusion: the 3 graphs show different containment approaches; why do they also vary R-sub-zero? Holding it constant would help compare the 3 approaches, no? (or is that a very tiny R-sub-theta on the Y axis? Do different approaches cause that to vary?)

[psh]

Oh. same 3 approaches per chart, but efficacy varies and efficacy affects R-sub-0 (?)

[hilzoy]

psh: it doesn't, actually. The lines tell you: for a given set of control measures, and for a disease with a given theta (proportion of infection occurring before symptoms), what is the maximum R(0)that that disease can have and still be controlled? Thus, the lines differ for the three graphs, since they chart control measures with different rates of efficacy.

The four diseases charted are the blue and red blobs, which are all at the same place on all three graphs. SARS (in blue) is in the bottom left; smallpox (in blue) is just above it; flu is the red bar-shaped thing coming down from the top (bar-shaped since its R(0) is unclear); and HIV is the red blur over to the right.

I think the situation of HIV is not exactly right, here. They are emphasizing the proportion of infections that occur before a person becomes symptomatic as a variable in this model, I think, because normally becoming symptomatic is what allows an infected person to be identified as infected, and thus for measures like contact tracing and isolation to be taken. HIV testing, which of course is possible long before a person develops symptoms of AIDS, complicates the use of this variable -- lots of people are identified as HIV+ long before they develop symptoms.

[hilzoy]

"It doesn't, actually" means: R(0) doesn't vary from one chart to another.

[psh]

Ah.

[xanax]

For the edification, safety and continued good health of the ObWi "community:"

Bird Flu Symptoms:

The Center for Disease Control has released a list of symptoms of bird flu.

If you experience any of the following, please seek medical treatment immediately:

1. High fever
2. Congestion
3. Nausea
4. Fatigue
5. Aching in the joints
6. An irresistible urge to crap on someone's windshield.

Happy holiday weekend...