“There are known knowns. These are things we know that we know. There are known unknowns. That is to say, there are things that we know we don’t know. But there are also unknown unknowns. There are things we don’t know we don’t know.”- Donald Rumsfeld
How do we know what we know? It is said by some anti-vaccine proponents that vaccines are not needed because the diseases they prevent are either gone or no longer as severe as they were in pre vaccine times. People may have suffered and died in the distant past, but no longer. The risk now is from the vaccines not the diseases they no longer prevent.
36,000 people, more or less, die every year from influenza. That is the number of deaths according to the CDC web site; the NEJM review uses the higher number of 56,000 (7). Which number is correct? Isn’t that why the flu vaccine is recommended: to prevent all those people from dying.
36,000 is a lot of people. That’s about 120 deaths per million people in the US. In Oregon, population about 3 million, that would be about 360 people a year, which is two deaths a day for the six month flu season.
“Death is caused by swallowing small amounts of saliva over a long period of time.”–George Carlin.
2,400,000 people die every year in the US, about 6600 a day. In Oregon, that is about 65 deaths a day. No one outside a epidemiologist is going to notice 2 extra deaths a day during flu season. I have seen a lot of people die of influenza, but I have a biased experience: I am an infectious disease doc, so I am likely to see people with influenza, especially patients with disease severe enough to kill them.
About the same number of people die from car accidents and die from handguns in the US each year as die from influenza. I have never known a person in my real, as opposed to my professional, life to die from influenza or handguns or a car accident. My personal experience suggests no one dies from these causes, but since I take care of patients at one of the Portland trauma hospitals, I know what cars and guns do to people. My professional life confirms that people do indeed die from being shot or car accidents. I would wager that most people reading this blog have not known anyone who has died from influenza, guns or car accidents. The fact that people do die of influenza seems contradicted by experience. Why get the vaccine? I don’t get the flu and and no one I know has ever died from it.
As an illustrative example, a relative of mine, a retired physician, mentioned that he thought the shingles vaccine was a waste of time and money as he had never known anyone to get shingles. Using personal experience to judge disease prevalence is unreliable. If I applied the same rationale to driving, I would not wear a seat belt as I have never been in a high speed crash.
36,000 people die of influenza each year. What is the source of that statistic? From “Mortality associated with influenza and respiratory syncytial virus in the United States”. JAMA 2003. Is that really how many deaths are there from influenza? It depends on what you mean by ‘death’ and what you mean by ‘influenza’ and what you mean by ‘from’.
“It depends on what the meaning of the words ‘is’ is.” – Bill Clinton
Influenza deaths are an estimate. A sample of a population is evaluated for a disease, like influenza, and the the results are extrapolated to the whole population. The bigger the sample, the more reliable the extrapolation. It is not a simple task to estimate disease prevalence. How do you define a case of influenza? Most people with influenza do not get cultures to confirm the diagnosis. What combination of cultures, rapid testing, serology and clinical syndrome do you use to define a case of influenza?
There is a lack of precision in determining if influenza was the cause of death since many people with influenza do not have the tests done to prove the diagnosis. Someone is admitted to the hospital with what starts as a febrile coughing illness and dies of heart failure and bacterial pneumonia. The initial illness could very well have been influenza that set them down the road to death. If you are interested in influenza deaths, you need to define what constitutes death from influenza. Did they die directly from influenza? Did they die from secondary infections? Did they die from underlying diseases such as heart or lung disease that are worsened by the physiologic stress of influenza?
To complicate matters, the US does not have a unified health care system and as a result deaths are not tracked uniformly. You might be surprised at how difficult it is to accurately fill out a death certificate since people often die from multiple diseases.
Lets peruse the JAMA article to get a sense how influenza deaths are estimated. How do they find that grain of sand that is influenza deaths from the mountain of routine mortality? Lets look at the reference in detail, because it gives insight as to how to determine influenza deaths in the US.
“The doctors killed Garfield; I just shot him.” – Charles Guiteau.
“Numbers of deaths attributable to influenza are difficult to estimate directly because influenza infections typically are not confirmed virologically or specified on hospital discharge forms or death certificates. In addition, many influenza-associated deaths occur from secondary complications when influenza viruses are no longer detectable. Nonetheless, wintertime influenza epidemics have been shown to be associated with increased hospitalizations and mortality for many diagnoses, including congestive heart failure, chronic obstructive pulmonary disease, pneumonia, and bacterial superinfections.”
Here is how they arrive at an annual average of 36,000 influenza deaths.
First, they defined a time frame to evaluate: the respiratory season is July 1 to June 30.
Second, find influenza cultures. Positive cultures define a case of influenza. They looked at all the influenza isolates reported to the CDC. This is an underestimate of the amount of influenza as many cases of influenza are defined clinically and a a small fraction of suspected cases of influenza are confirmed with cultures.
Third, look at mortality from the National Center for Health Statistics. They included “the underlying cause-of-death because it represents the disease or injury that initiated the chain of morbid events that led directly to the death.”
Why did they die? The US lacks a national data base for cause of death, so they used diagnosis codes rather than death certificate reports, which also has issues. Every disease has an ICD-9 code associated with it. Fever is 780.6. HIV is 042. You know a doctor by the codes she has committed to memory from repeated use. Diagnosis codes exist for billing purposes and the rigorousness with which a diagnosis is made is often less than perfect, limited by time and potential morbidity. I could diagnosis every case of influenza if I could do an open lung biopsy on each patient, but that is not worth either the harm to the patient or the cost of the procedure. Often the diagnosis code represents a best, but honest, guess as to what diseases a patient had.
“The 3 death categories modeled were underlying pneumonia and influenza deaths (ICD-9 codes 480-487), underlying respiratory and circulatory deaths (ICD-9 codes 390-519 and ICD-10 codes (I00-I99, J00-J99), and all-cause deaths (all ICD codes).”
Diagnosis codes are not an optimal way to know what the ‘real’ diagnosis was, but the best we have. So we using a 2 megapixel camera to take a snapshot of influenza deaths, but the picture is going to be a reasonably accurate representation of the impact of the disease.
“While the individual man is an insoluble puzzle, in the aggregate he becomes a mathematical certainty. You can, for example, never foretell what any one man will be up to, but you can say with precision what an average number will be up to. Individuals vary, but percentages remain constant.” – Arthur Conan Doyle
Fourth, and final, part of the study is the application of magic. Or statistical analysis. Same thing to me. They used an “age-specific Poisson regression model that used weekly influenza circulation data.” Here is where I have to defer to people who have a deeper understanding of statistics as I thought a Poisson model would be fish counting in the Seine.
“Poisson regression has been mainly applied to compare exposed and unexposed cohorts and to evaluate the clinical course of ill subjects (2)”. Sounds perfect. That is what we want to do, so the statisticians appear to be using the right statistical analysis. Here is another factor that could lead to lack of trust in the death rates from influenza: the processes by which these numbers are generated are really understood by a tiny fraction of the population. It appears from reference (2) that the Poisson is a reasonable statistical analysis to apply to the question of the number of deaths from influenza, but I almost have to accept that proposition on the basis of faith given my bleak understanding of statistics.
So what did the analysis show?
For the period in question there was an yearly average (some years more, some years less) of
69,140 deaths from pneumonia and influenza,
1,135,724 deaths from underlying pulmonary or cardiac disease, and
2,126,740 deaths from all causes.
Applying the Poisson model to the data they looked at the deaths when influenza was circulating in the community and when influenza was not circulating in the community and compared the deaths. They determined that on average there were
5977 deaths from pneumonia and influenza,
25,420 deaths from underlying pulmonary or cardiac disease, and
34,470 deaths from all causes
could be attributed to influenza. These were the extra deaths, by type, that occurred when influenza was circulating in the US.
See copywrited material here: . I feel the same way.
It is important to note that the flu does not kill most people directly, but indirectly through secondary infections or exacerbation of underlying chronic medical problems. There is year to year variation in the number of deaths depending on the virulence of influenza, the effectiveness of the flu vaccine and the number of people protected by past exposure to the strain.
The dogma, as an aside, has been that during the 1918 pandemic most people died from acute influenza infection, as occurs with the current strain of avian influenza. That may have not been the case, as an archeological epidemiologic study in Clinical Infections Diseases (8) suggests that deaths from secondary bacterial infections may have accounted for the majority of the deaths. I keep drumming my fingers nervously on the table, waiting for the perfect infectious disease storm to hit. If we have a bad influenza season combined with the new widespread MRSA strain, we could see a huge spike in deaths from secondary bacterial pneumonias from MRSA. The current USA 300 strain is impressively virulent when it gains access to virus damaged lungs, killing patients in less than 48 hours, and I can’t reliably kill MRSA.
The JAMA study also demonstrated that the older you are, the more likely you are to die of influenza, that if the patient is >85 years old, there is a 16 times higher chance of dying of flu than if they are less than 65.
Short of a universal health care, where everyone is in a database with precise diagnosis, this kind of epidemiologic study is as good as we can do to determine how many people die from influenza. The best way to determine death rates from influenza would be to take a defined population of people and routinely evaluate them by culture and serology through the flu season, see who developed the flu and determine who died as a result. Not an ethical prospective study, but accomplished retrospectively in Tennessee nursing homes, where they had patients for a total of 81,885 person years and evaluated them for influenza. In this population, most of which had multiple co-morbid conditions, when influenza was circulating there was associated increase in mortality of an extra 15 deaths/1000 (3). If the US population were all similarly old and debilitated, then we would have over 4 million extra deaths a year from influenza. Similarly, when influenza is circulating in the community, there is an associated extra 9% of deaths in those with lung disease (4).
There are other ways to determine deaths from influenza. Deaths have been reported and influenza tracked since 1900. The onset of the influenza season has always been associated with an increase in death of all kinds. Of course, until diagnostic tests were available, the diagnosis of influenza has to be taken with a grain of salt substitute. Using historical reporting data you can get a rough idea of how often influenza kills. This kind of data is an approximation (5), but probably close enough to the truth to understand the impact of influenza. The problem is knowing with reasonable assurance that the deaths were indeed due to influenza.
Given that limitation, over time deaths from influenza have fallen since 1900 from 10.2 deaths per 100,000 population in the 1940s to 0.56 per 100,000 by the 1990s. That would be about 16,800 deaths directly from influenza each year, not including the secondary deaths from heart and lung diseases, more than the JAMA article, but they used different case definitions.
Using a similar methodology, it is estimated that influenza results in 365,000 or more hospitalizations each year (6).
”An approximate answer to the right question is worth a good deal more than the exact answer to an approximate problem.”—John W. Tukey
All the studies demonstrate the same basic principal: when influenza arrives, so does the grim reaper. So how many people does influenza kill each year? 16,000? 36,000? 56,000? Even more? Or less? As you can see, it depends on how you do the counting. All the studies put the number of deaths in the same ballpark. One says that influenza directly kills ‘only’ hundreds of people a year. I think hundreds is a lot of people, especially as the deaths were preventable. It is true that deaths directly from influenza are unusual. It was a mild flu season last year, but 76 children died directly of influenza, all unvaccinated, and the number of deaths is probably an underestimate given the current surveillance system. I could not find a number for adult deaths last year at the CDC site.
“A single death is a tragedy; a million deaths is a statistic.” —Joseph Stalin
By our best estimates, influenza, one way or another, kills 36,000 people in the US every year. 36,000 stacked corpses would make column 6 miles high. Many of those deaths were preventable with the flu vaccine. Every death was a tragedy.
For many people influenza is like being pushed off a cliff. Neither the push (the flu) nor the fall (underlying condition) kills you.
It is the sudden stop.
Next time: how effective is the influenza vaccine?
(1) Thompson WW, Shay DK, Weintraub E, et al. Mortality associated with influenza and respiratory syncytial virus in the United States. JAMA 2003;289:179–86.
(3) J Am Geriatr Soc. 2003 Jun;51(6):761-7. Influenza- and respiratory syncytial virus-associated morbidity and mortality in the nursing home population.
(4) Arch Intern Med. 2002 Jun 10;162(11):1229-36. Winter viruses: influenza- and respiratory syncytial virus-related morbidity in chronic lung disease.
(5) Am J Public Health. 2008 May;98(5):939-45. Epub 2008 Apr 1. Trends in recorded influenza mortality: United States, 1900-2004.
(6) Influenza-Associated Hospitalizations in the United States. JAMA. 2004;292:1333-1340.
(7) Prevention and Treatment of Seasonal Influenza NEJM. 359;24 www.nejm.org december 11, 2008
(8) Predominant role of bacterial pneumonia as a cause of death in pandemic influenza: implications for pandemic influenza preparedness. J Infect Dis. 2008 Oct 1;198(7):962-70.