Influenza in the US, week 23

The Centers for Disease Control and Prevention has published their weekly analysis of influenza activity for the week of 7-13 June 2009. They conclude that

…influenza activity decreased in the United States, however, there were still higher levels of influenza-like illness than is normal for this time of year.

During week 23, 38.7% of specimens (2,765 out of 7,149) tested positive for influenza virus. Of these, 82% (2,263) were identified as the pandemic H1N1 strain. The H3N2 strain, and last season’s H1N1 strain, accounted for only 21 and 22 of the positive specimens, respectively. These results are shown in the following graph:


The numbers raise at least two questions. First, the seasonal H3N2 and H1N1 strains are clearly disappearing. Is this because it is the end of the influenza season in the northern hemisphere, a time when influenza typically drops to very low levels? Or is it because the older viruses are being ‘forced out’ by the new pandemic strain? The fact that the pandemic H1N1 strain is also supplanting last year’s strains in Australia suggests that these viruses will not return in the fall.

The other question is why the pandemic H1N1 strain continues to cause a significant number of infections in the northern hemisphere in June. We’ve previously discussed how temperature and humidity are believed to be important factors in determining the seasonal patterns of influenza in temperate climates. Clearly the 2009 pandemic strain isn’t following these ‘rules’. Neither did the 1918 pandemic strain. Could it be that seasonality of  influenza is regulated not only by temperature and humidity, but also by levels of population immunity? Perhaps, after a season of influenza, the high levels of immunity coupled with high temperatures and humidity together lead to reduced disease in the summer. Because population immunity to a pandemic strain is extremely low or nonexistent, the virus can circulate even in conditions of high temperature and humidity. This hypothesis is undoubtedly an oversimplification, and other factors, such as transmissibility, are also likely to play a role. For example, the pandemics of 1889, 1957, and 1968 did display seasonality in the northern hemisphere – see Figure 1 in the paper cited below.

Miller, M., Viboud, C., Balinska, M., & Simonsen, L. (2009). The Signature Features of Influenza Pandemics — Implications for Policy New England Journal of Medicine, 360 (25), 2595-2598 DOI: 10.1056/NEJMp0903906

12 thoughts on “Influenza in the US, week 23”

  1. Vince,

    Look at the chart you posted. Notice how both the total number of cases and the percentage of positives are as high as the highest levels (or nearly so) from this past flu season. I don't understand how the CDC can say that influenza is “declining”. From where I'm sitting, it looks like its going like crazy. Can you explain this? Am I missing something?


    — Lenn

  2. I don't know exactly how CDC prepares its conclusions, but they do
    seem to make statements on a weekly basis. Therefore, the total
    influenza positives did decline from week 22 (40.2%) to week 23
    (38.7%). The graph shows total number of samples and the positive
    samples for each strain. Here are the percent positive samples for
    each week: 15 (6.2%), 17 (13.2%), 18 (11.9%), 19 (15.1%), 20 (22.4%),
    21 (31%), 22 (40.2%), 23 (38.7%). The percent positive samples don't
    jive with their conclusion; perhaps they are also looking at other
    data – number of states with extensive spread?

  3. I think it is very important to note that the CDC graph is the percentage of samples taken that are testing positive, not the total percentage of the population with the flu. As such, this value will be influenced by changes in testing practices. In particular I can imagine that in an average flu season the elderly are most likely to go to the doctor with symptoms (because they have been told that they are most at risk for complications) and some given percentage of everyone who shows up with ILI are tested for the flu. In addition, armed with Medicare and spare time, many elderly are frequent visitors to the doctors' office. When flu circulation is low during the summer, only a very small percentage of them will actually have the flu. But now we have swine flu, which is known to be a higher risk for younger people. This lowers the threshold for when a younger person will decide to visit the doctor, but the young people who are visiting will still be a sicker crew than the regular elderly population. In addition, because of the onslaught of panicked ILI patients many doctors have stopped testing all but the sickest — apparently Kaiser will only test you now if you are actually hospitalized. This can decouple the percentage of positive samples from the real H1N1 infection rate.

    I think a more telling CDC statistic is the percentage of all deaths due to pneumonia/potential flu related causes. This is following seasonal trends, with the exception of the last week of reporting, when it just nudged above the epidemic threshold.

  4. Nice analysis. But what do you conclude from the pneumonia/influenza
    mortality data? Clearly the pandemic H1N1 strain continues to
    circulate in a non-seasonal manner, and this is not reflected in the
    P/I data.

  5. I am not sure that the epidemiological data support the idea that temperature and humidity are the factors in seasonality. In Nicaragua there is sometimes a bimodal distribution, with the second peak in July, the hottest and rainest month of the year. We wrote about this here:

    The lack of data on seasonality in the tropics (and flu data from the tropics more generally) is a major gap in our knowledge.

    As far as your recent Tweet (Greg Poland's Editorial), his piece contains the false assertion that 56K people die of flu every year (we don't know how many people die, but the measure, as you point out in a recent post and which I have explained is excess mortality and represents and average over many flu seasons. It also depends on how it is modeled (Serfling's Method or other regression techniques) and what the part of the death certificate is being used. Moreover, as Lone Simonsen, Cecile Viboud and Mark Miller (whom you cite) have pointed out on a number of occasions, the evidence that mortality among the elderly is affected much by flu vaccine is weak. It doesn't show up in cross-sectional data or the scant RCT data, only in the observational studies. I am not knocking observational design (I am an epidemiologist and that's what I do for a living) but in this case the discrepancy is problematic.

    Love your blog, though. Just thought I'd add some additional info.

  6. Clearly H1N1 is continuing to circulate, and to replace other strains but the P/I data (as well as google flu trends) suggests that the rate of spread is definitely being tempered by seasonality, with total flu rates across the country probably not higher than 5% – 10% higher than average for this time of year (my guesstimate by eye from google flu) with clusters going on in Utah, New Jersey, New York, and probably other smaller places. Also note that the CDC percentage of outpatient ILI visits is reported as normal in every region except Region II, which probably represents the cluster going on in New Jersey, which is also a strong indicator that throughout most of the country the high percentage of positive tests is a consequence of changes in sampling practices. By the way, what is the reason for the high P/I mortality in the winter of 2007/2008?

  7. Hi Vincent,
    do you think that the extensive use of air conditioning may play a role in the flue spreading thorugh the northern emisphere in spite of the summer?

  8. “Summer Flu”- we are starting to hear this word quite often now. Although flu is primarily winter season infection and with evidence that flu virus spreads efficiently at low temp/ low humidity, words like “summer flu” were not used often. But with recent pandemic strain H1N1, we are changing our thoughts. Not only in US, but also in Europe, incidence of summer flu is getting reported. In fact Wales/England reported summer flu incidence rate above 10 years average. With all these observation, it will be convincing to learn how this strain of H1N1 is able to spread at hot temperatures and humid conditions.

  9. Hi Vince et al,

    I did some more research on this and it seems that they have been saying about the same thing since week 19. In week 17 (the first week they reported pandemic numbers), they said activity increased. In week 18, they reported that activity stayed about the same. Starting with week 19, they say it is decreasing. I don't know where they are getting this from, as the percentages have increased every week until this past one, while the actual numbers have increase ever week.

    I'm hoping that the table below works correctly. I pulled the number and percentages of Influenza positives from the CDC Weekle Reports. The approximate sample size I calculated from the figures provided (they should be correct to within one or two people). The missing week is when the novel strain hit and they took a week of to regroup.

    I'm not sure what we can really say about this, other than the fact that the numbers are increasing and that the percentages of positives in the last few weeks are larger than at the height of the normal flu season. To me, this latter point is the most important. The number of people infected could be an artifect of increased surveillance, but the percentages (if the sampling is done correctly) should be about the same regardless of the number sampled (within certain minimum sample size requirements, of course).

    Anyway, I hope you guys find this interesting.


    — Lenn

    Week Number % Positive Approx Samp
    40 8 0.7% 1143
    41 7 0.5% 1400
    42 10 1.0% 1000
    43 17 0.4% 4250
    44 11 0.6% 1833
    45 40 1.7% 2353
    46 16 0.9% 1778
    47 52 2.5% 2080
    48 43 2.0% 2150
    49 72 2.8% 2571
    50 103 3.5% 2943
    51 103 3.4% 3029
    52 130 5.0% 2600
    53 79 3.2% 2469
    1 242 7.1% 3408
    2 409 11.5% 3557
    3 588 15.8% 3722
    4 792 16.2% 4889
    5 1154 20.6% 5602
    6 1313 24.4% 5381
    7 1405 24.6% 5711
    8 1418 21.2% 6689
    9 1252 23.0% 5443
    10 1102 21.7% 5078
    11 1104 21.4% 5159
    12 722 16.8% 4298
    13 348 12.3% 2829
    14 259 9.0% 2878
    15 151 6.2% 2435
    17 1892 13.2% 14333
    18 1454 11.9% 12218
    19 1074 15.1% 7113
    20 1450 22.4% 6473
    21 2074 31.1% 6669
    22 2681 40.2% 6669
    23 2765 38.7% 7145

Comments are closed.

Scroll to Top