Turning ILI cases into H1N1 cases – How the Lipsitch paper redefines reality

In Science, it is important to use words in very precise ways. This leads to specialised vocabularies which may be difficult for laymen to follow but a godsend to pseudoscientists who can either use real terms incorrectly or simply make up their own. However, nothing is more dangerous than for scientists to misuse words in their publications and then spread this misinformation to the press.

And that is exactly what has happened in a paper recently published in PLoS Medicine by a group of scientists working with Dr. Marc Lipsitch.

There are several ways to determine if someone is infected with the new H1N1 virus. Real time PCR is the most commonly used, but virus can also be cultured and sequenced. [Flu A tests fail about 50% of the time and are not considered reliable]. Seroprevalence studies could be done to determine whether people have been infected with H1N1 in the past by looking for antibodies to it, but no results from such studies have been released.

Influenza-like illness (ILI) is a dumping ground for a wide variety symptoms and diseases. Although some people with ILI actually have influenza, most do not. This is well known and has been experimentally verified, as I discussed previously – Estimates of case fatality rates based on influenza-like illness are wrong. I had written that blog in response to statements to the press by Dr. Lipsitch back in September. Apparently, those comments in September relate to the paper which has just now appeared in PLoS Medicine. As I pointed out back in September, an empirical study done in Sweden found that only 5% of ILI cases actually had pandemic H1N1. Most were infected with rhinoviruses or had other diseases. So, it should be clear to anyone that ILI does not equal an H1N1 case.

Should be, but apparently isn’t.

From the Lipsitch study:

We estimate the severity of pH1N1 infection from data from spring–summer 2009 wave of infections in the United States. The New York City and Milwaukee health departments pursued differing surveillance strategies that provided high-quality data on complementary aspects of pH1N1 infection severity, with Milwaukee documenting medically attended cases and hospitalizations, and New York documenting hospitalizations, ICU/ventilation use, and fatalities. These are the numerators of the ratios of interest. The denominator for these ratios is the number of symptomatic pH1N1 cases in a population, which cannot be assessed directly. We use two different approaches to estimate this quantity. In the first (Approach 1), we use self-reported rates of patients seeking medical attention for ILI from several CDC investigations to estimate the number of symptomatic cases from the number of medically attended cases, which are estimated from data from Milwaukee. In the second (Approach 2), we use self-reported incidence of ILI in New York City, and making the assumption that these ILI cases represent the true denominator of symptomatic cases, we directly estimate the ratio between hospitalizations, ICU admissions/mechanical ventilation, and deaths (adjusting for ascertainment) in New York City.
[emphasis, mine]

Given that we already know that in one empirical study 95% of ILI cases were not due to influenza, it is obvious that the Lipsitch study has made a fatally incorrect assumption. The fact that the rest of the study involves complicated statistics is irrelevant (although quite impressive to gullible reporters). If your assumptions are wrong, your model cannot produce valid results.

Some may forgive Dr. Lipsitch his extensive (and quite wrong) assumptions on the grounds that that is the best that we can do. However, this is simply not true. We could test vast numbers of people with high-throughput Real Time PCR, if the CDC wished to (or perhaps, knew how to).  We could have done seroprevalence tests months ago, if the CDC wished to (or perhaps knew how to). The decision not to gather actual, real data was a choice on the part of the CDC, not a limitation in technology, as Dr. Lipsitch implies.

We should not blame Dr. Lipsitch for the CDC’s failings. However, he can be blamed for grossly overinterpreting his own data:

From, The Age, December 9, 2009

‘I think it is very likely to be the mildest pandemic on record,” said Marc Lipsitch, a professor of epidemiology at the Harvard School of Public Health, who led a federally funded analysis with researchers at the Centres for Disease Control and Prevention (CDC).

Such a broad statement about a new virus based on such a shaky foundation (actually, no foundation at all) would be inappropriate at any time. Making this statement in the middle of a pandemic is irresponsible.

So far over 500 children are estimated to have died in the US from this virus. Doomed pregnant women on ventilators have had emergency C-sections to save their babies. Teachers and health care workers are dying.

And flu season is just beginning.

I don’t claim to know the future. I do know that those who cannot tell the difference between ILI and an actual case of H1N1 have no business talking to reporters about what is going to happen.

No business at all.


Presanis et al. (2009) The Severity of Pandemic H1N1 Influenza in the United States, from April to July 2009: A Bayesian Analysis. PLoS Medicine.


19 thoughts on “Turning ILI cases into H1N1 cases – How the Lipsitch paper redefines reality

  1. so, you claim the CFR is much higher,
    but due to fewer infections while the deaths
    are still lacking.
    Even if true, would it matter for calling
    it “a mild pandemic” ?

  2. I have been saying this for a while, gsgs.

    No, I would not call this a “mild” pandemic. I would say that we are in the early stages of a pandemic with a nasty virus. The final death toll cannot be known at this time. Too many variables. But I would say the potential for large numbers of deaths is high.

  3. gsgs, you miss the point of the blog.

    ILI is NOT equal to an H1N1 case. Unless someone actually tests people, they can’t calculate CFR.

  4. They are bad estimates based on a false assumption, ie, that ILI is a good marker for panflu cases.

    I cannot explain why they choose to do this rather than real-time PCR or seroprevalence studies. It seems to be a basic issue of competence. They know how to count ILI cases, they apparently do not know how to high throughput PCR or seroprevalence studies.

  5. Monotreme, this actually is not a pandemic at all… it is an atypical A(H1N1) influenza that has infected hundreds of millions of people around the globe with a fraction of the deaths normally [claimed to be] associated with influenza. This is only a pandemic because the WHO changed the criteria for one AFTER the A(H1N1) virus emerged in the USA.

    You fall into the very trap you accuse Lipsitch et all to have fallen into… You say, “So far over 500 children are estimated to have died in the US from this virus. Doomed pregnant women on ventilators have had emergency C-sections to save their babies.”

    The 500 children dead statistic is a guestimate based on unconfirmed deaths due to pneumonia… with no H1N1 testing. So you are criticising beat up statistics on the one hand yet using beat up stats on the other.

    You can’t argue it both ways.

  6. Ron Law, you apparently do not know what a pandemic is. Here is one definition:

    “Pandemic influenza is a global outbreak of disease that occurs when a new influenza A virus appears in humans, causes serious illness and then spreads easily from person to person worldwide.”

    The new H1N1 clearly meets all those criteria.

    There are over 300 child deaths from documented H1N1. The current estimate of child deaths due to H1N1 is now over 1,000.

    You can see a partial list here:

    1. It is easy to make up a definition of what a pandemic is/isn’t. The simple fact is that if there were three pandemics in the 20th century, then the definition on the WHO website in April 2009 was at least partially correct… there had to be a new subtype… H1N1 is not a new subtype, therefore, according to the WHO definition in in place when this atypical H1N1 emerged it is not a pandemic. The second criteria for a pandemic according to the WHO is that enormous numbers of death and disease occur… clearly that has not happened either.

      If this is a pandemic, then there have been several this decade already… and we can expect many more… the fact is that it is not a new sub-type… The Virginian State pandemic plan uses the original WHO definitions… so your choice of ‘google a reference’ supports my argument. Also note the same website simply defines a pandemic as,”An epidemic occurring over a very wide area, crossing international boundaries and usually affecting a large number of people.”

      That means that seasonal influenza causes a pandemic every year… I repeat, by the WHO definition in place as at April 2009, when the atypical H1N1 virus emerged in the USA, this is not a pandemic. It is not a new influenza virus and it is not causing enormous numbers of death and disease… it is causing enormous numbers of infections but most of then are asymptomatic.

  7. Ron Law, I think you are confused about virology. Let me explain. The new H1N1 virus is not related in any significant way to the seasonal H1N1 virus. The source of the 8 genomic segments is completely different. This is not a matter of the small changes that occur to flu viruses each year, the pandemic virus is a completely new virus of unknown origin. I have written blogs describing the closest matches for PB1 and PB2 which you might read for further edification, if you wish.

    As regards Wesley Husband, the actual cause of death was not revealed. The news stories I read said that the local tests were negative but that the other tests might be positive. At this point, we don’t know what the results were. Further, I don’t think you picked this case at random. I think you deliberately ignored the many cases above and below it that had confirmed H1N1. Not very honest of you, is it?

    For anyone interested in the truth, I suggest going to PFI_Main and watching the videos.


    1. I’m not confused at all… it is not a new sub-type… it is H1N1… The simple fact is that if measured by the WHO definition in place at the time it emerged it could not have been classified as a pandemic on both counts… it’s not a new sub-type and it’s never caused enormous numbers of deaths and disease.

      As for Wesley Husband I can assure you it was totally random… there is no evidence I can find to even suggest it was positive…

  8. You are confused if you think that because the pandemic strain and and one of the seasonal strains are both labeled H1N1 that they are similar. They are completely different viruses. Anyone who spends five seconds looking at the sequences can see this. You obviously have not done so and are instead arguing semantics rather than biology.

    When was the last year we had so many young people die of the flu?


    Again, anyone interested in the truth can watch the videos here:


  9. Reading the Bayesian paper right now, but really – the sample size in the Swedish ILI study you cited was only 79. 1 ILI does NOT equal 1 flu case, much less 1 H1N1 case, but you ought to dig up some better ILI: case ratios than that one.

  10. CarolynM, if you can find a better paper than the Swedish study, please let me know. In any case, it appears that we agree that conflating ILI with H1N1 is a bad idea.

  11. monotreme, not only is the Swedish study a small study size, but it was also highly selected so should have been biased toward detecting the panic flu.

    It’s now seven months on since your original blog and the earth is still rotating… we were not at the flu season wasn’t just beginning, it had ended!

    As an aside, ILI rates in NZ are at there lowest for this time of winter in 20 years… in all but one year (2004) we would be into the peak phase of ILI by now… This panic flu seems to have been far more effective in reducing both disease and deaths than the regular vaccine… so why are we still trying to ‘stamp it out’ knowing that any void will be filled by something much worse…?

  12. Mr. Law,

    I hope that more comprehensive studies are done. However, the existing data suggests that the 2009 H1N1 virus was far deadlier than regular flu. Two separate studies established that it was 5 to 10 times more lethal to children than seasonal flu. It was also more deadly to pregnant women and people with a variety of health problems. Finally, a study that I noted yesterday at PFI_Forum, reported that 50% of the people who died of 2009 H1N1 were previously healthy. Seasonal flu usually kills very old people who already have many diseases that are going to kill them anyways. 2009 H1N1 kills children, pregnant women and middle-aged people in the prime of their lives. There is no serious disagreement among scientists about this.

    The earth is indeed still rotating, but it is doing so without people who would otherwise be alive without this flu.

    I, and millions of others, have been vaccinated against 2009 H1N1 with no ill effects. Although a few people did experience serious adverse effects, this risk paled against the risk of the virus itself.

    As far as worrying about “filling the void” if 2009 H1N1 is stamped out, you are little confused about pandemics. We will all develop antibodies against this virus, one way or another. We can get them by vaccination or we can get them by being infected. Personally, I prefer the former, but in terms of the future evolution of the virus, it doesn’t matter.

    Influenza often peaks in the Southern Hemisphere in August, so it is premature to declare victory over this particular virus. As you note, something worse may be next. If so, then the risk-reward ratio for getting vaccinated will be even better than it is now. Right?

  13. Hi monotreme… you can cut the formalities… Ron will do… never have been one for titles…

    Here’s more drivel from WHO…

    “In New Zealand, rates of ILI have steadily increased over the month of June; however, only small numbers of predominantly pandemic influenza virus have been detected so far. In both Australia and New Zealand, current levels of ILI are similar to those observed during the same period in 2008, when the influenza season was noted to have arrived and peaked late in winter.”


    Firstly, the flu in 2008 peaked bang on normal… secondly, despite the heightened surveillance, testing and public ‘awareness’ ILI consultation rates at week 26 are the second lowest (just) in nearly two decades!

    Flu in NZ peaked mid winter the same as in 2007, and 2006, and 2005

    See fig 1

    The only late winter flu over the past decade was in 2004
    See fig 1

    See fig 1

    See fig 1 going back to 1996

    Another interesting facet of surveillance in NZ this year highlights that the sentinel surveillance is much much less sensitive than simply monitoring the testing of hospital patients tested for influenza.

    There had been 20 positive 2009 H1N1 swabs spread over 24 weeks detected via routine laboratory testing before any were detected by the official sentinel surveillance system. In other words the virus had been circulating for 6 months before the sentinel surveillance system identified its first case…

    It has been stated that after the first 50 cases of influenza the virus will have been spread far enough that containment is impossible… this suggests that sentinel surveillance is of no use other than for academic purposes.

    Makes you wonder how ‘expert’ WHO is, doesn’t it?

    You say the influenza often peaks in August in the Southern Hemisphere… I’m talking about NZ… I’m sure it’s possible cherry pick countries and data to support any claim, such as the one made that the US was about to enter its flu season when in fact it had just finished. Apparently according to our Minister of Health, about 400 New Zealanders die each year from influenza… last year 35 died… three of those were under 20 years old… the same number that died in 2005. You discount old frail people’s worth. Sad. You say that getting influenza and being vaccinated both confer immunity… maybe… but isn’t it strange that apparently people who got the flu before about 1950 still have protection but we have to have the vaccine every year because it only lasts about 6 months. You say pregnant women were at more risk… my calcs are that the chances of a pregnant woman dying WITH the flu are about 1:15,000… the chance of dying in a car crash are about 1:8,000… makes one think, doesn’t it?

  14. Ron, flu seasons are variable as your own links suggest. Pandemics are so rare that we have even less basis for determining when they will peak. In the US, the number of 2009 H1N1 cases peaked at different times in different States. We don’t know why. I won’t claim to predict when the flu season will peak in New Zealand. All I’m saying is that it is way too soon to declare all clear.

    I have a low opinion of the WHO, but for opposite reasons to you. I think they have deliberately understated the harm this virus has caused. Further, I would suggest that you are playing right into their hands by calling them alarmists. They want people to do this. That way they won’t be blamed for not issuing a stronger warning should the virus become more virulent. I don’t think that the WHO claims about what is and what is not possible in terms of containment are credible. Indeed, I think these claims are politically and economically motivated because they do not want countries to institute movement restrictions.

    I don’t discount the worth of the elderly. However, I do think most people would agree that a virus that kills 30 year olds is different than a virus that primarily kills 90 year olds. I’m not sure where you get your calculations for the risk of a pregnant woman with the pandemic flu dying, but I don’t think they are remotely in the range of reality. And in any case, your comparison with auto deaths is irrelevant. I think people should have access to vaccines and seatbelts. They aren’t mutually exclusive.

    Here are some references you might want to look at:

    Libster et al (2010) Pediatric Hospitalizations Associated with 2009 Pandemic Influenza A (H1N1) in Argentina. New England Journal of Medicine. Volume 362:45-55.

    Siston et al (2010) Pandemic 2009 influenza A(H1N1) virus illness among pregnant women in the United States. JAMA Volume 303:1517-1525.

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 )

Google+ photo

You are commenting using your Google+ 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 )


Connecting to %s