Colorized transmission electron micrograph of Avian influenza A H5N1 viruses (seen in gold). Image provided by CDC/C. Goldsmith, J. Katz, and S. Zaki.
Illness and Death During the Pandemic
By Grattan Woodson, MD, FACP
When studying how many people may become ill or die during a pandemic, one runs the risk of looking heartless. Statistics do not capture the uniqueness of each person, or the value of their lives. Rather than being an exercise in insensitivity, by understanding how serious the pandemic may be is necessary for making appropriate plans. No one is “only” a number; but numbers matter. For instance, knowing how many patients will be seriously ill during a pandemic affects the number of staffed hospital beds that will be needed.
When trying to gauge the severity of a pandemic, scientists look for two key variables. First is the clinical attack rate, which is the percentage of people within a population that get sick with the disease. So, if there were 100 people in the population, and 35 got sick, the clinical attack rate for that disease would be 35%. The second is the case fatality rate. This is the percentage of people with the illness who died. So, if there were 35 people sick with the disease and 3 die, the case fatality rate would be about 8.5%. If you know these numbers, predicting the number of people that will die during a pandemic is straightforward. Simply multiply these values by the number of people in the population (CAR x CFR x Population = Number of Deaths). The fatalities rise with an increase in any of the other factors.
Pandemics have higher clinical attack and fatality rates than during seasonal flu explaining why there are more deaths. It is as simple as that. The progressive rise in the human population is also an important factor in the expected death number. In 1918 the US population was 100 million while today it is 296 million.
During seasonal flu the typical clinical attack rate is in the range of 5% to 15% with the case fatality rate between 0.2% and 0.35%. It is common for the clinical attack rate to increase by about 5 times and the clinical fatality rate to zoom up as much as 30 times during severe pandemic outbreaks.
In the US, the Department of Health and Human Services released its updated Pandemic Influenza Plan in November 2005.¹In that plan, the US Government projects a clinical attack rate of 30% for the next flu pandemic, whether moderate or severe.
Case fatality rates are more difficult statistics to come by. They are available for recent pandemic and seasonal flu in the developed nations, but unavailable for past pandemics and present seasonal flu in the less developed nations where 87% of people live. The most reliable pandemic statistic is the number of deaths attributed to influenza in the developed nations, but even this statistic is subject to variation and depends upon the assumptions used in calculating it. For instance, with the exception of primary viral pneumonia, death is not usually caused by influenza directly. Instead, the flu places great stress on one or more of the body’s critical organ systems, whose failure is the direct cause of death manifesting as a heart attack, stroke or liver, kidney, lung, or heart failure. In other patients, flu sets the stage for a fatal secondary bacterial infection in the lungs or blood stream that is the actual cause of death. This indirect way influenza has of contributing to the death of patients can make it hard for doctors, coroners, or statisticians to sort out what role it had in the patient’s death.
US pandemic mortality projections
The US Government projects flu illness and deaths they expect in the next pandemic based upon two scenarios, moderate and severe respectively.
| US DHHS mortality and morbidity estimates for the US during the next pandemic | ||
| Case Fatality Rate Estimate | Clinical Attack Rate Estimate | US Deaths Estimate |
|---|---|---|
| 0.23% (Moderate) | 30% | 207,000 |
| 2.1% (Severe) | 30% | 1,903,000 |
|
US Population 2005 = 296,000,000 Adapted from the DHHS Pandemic Influenza Plan, November 2005 |
||
In a 2005 New England Journal of Medicine article on influenza, Dr. Michael Osterholm, Director of the Center for Infectious Disease Research and Policy, provided a death number projection for the US during a severe influenza pandemic of 1.7 million². He did not give any estimates for clinical attack rate or case fatality rate but we can extrapolate them based upon the number of deaths expected.
| Osterholm’s pandemic case fatality rate prediction for the US | ||
| Case Fatality Rate Prediction | Clinical Attack Rate Estimate | Number of Deaths in USA Estimate |
|---|---|---|
| 2.30% | 25% | 1,700,000 |
| 1.64% | 35% | 1,700,000 |
| 1.15% | 50% | 1,700,000 |
| US Population 2005 = 296,000,000 | ||
Inspection of the projections by the US Government and Dr. Osterholm reveals that the death numbers for both are similar. It is of interest that the calculations used to derive them were different. Dr. Osterholm used the simple expedient of extrapolating the same death rates observed during the 1918 pandemic to the present adjusted for the increase in population. The US Government states in their Planning Assumptions that they based the clinical attack rate of 30% upon the “average of the three 20th-century pandemics”. While the clinical attack rate in the developed nations for the 1957 and 1968 pandemics are accurate estimates, the rates for the 1918 pandemic are not known, and any figure for the 1918 Spanish Influenza used by the government or anyone is sheer guesswork only.
The derivation of the case fatality rate for a severe pandemic is less secure. In the US Government’s Planning Assumptions they admit that the severity of the pandemic cannot be predicted before hand. They chose a case fatality rate of 2.1% for the next major pandemic. This compares with a case fatality rate estimate of 2.5% for the 1918 pandemic used by the Institute of Medicine’s 2004 conference on influenza pandemic preparedness, and Dr. Osterholm’s estimate of 1.6%³.
Adjusting the official mortality projections for a severe bird flu pandemic
In creating a model for a severe influenza pandemic, I began with the assumptions used by the US Government’s projections and adjusted them for a higher clinical attack rate and the loss of effective hospital care during the pandemic.
Why the clinical attack rate may be higher than 30%
There are several important considerations that were not included in the US Government’s model that support the concern that our population is at higher risk from a pandemic than ever before. First is the concentrated nature of people living in cities today, as crowding is a well-understood epidemic accelerant. Second is the average age of the US population is higher now than in any prior pandemic. Older adults are more susceptible to influenza infection. Third is the large number of frail people alive today who suffer from chronic medical conditions that would not have survived in the past. These include people with diabetes, heart and lung disease, and cancer. There are also a fairly large number of people living with chronic diseases that affect immunity. They include those with HIV-AIDS, chronic hepatitis, renal insufficiency, and inflammatory bowel and joint disorders. All people with chronic medical disorders are at higher risk from influenza than those without them. These factors all support a higher clinical attack rate than projected in the US Government plan. A truly severe pandemic deserves to have a higher clinical attack rate than moderate or mild one.
Why the case fatality rate may be higher than 2%
How well or poorly a patient does during a serious illness is referred to as their outcome. Outcomes can be expressed in many ways, with death being one of the most conclusive. For this reason, death is called a hard outcome meaning that it is one that all observers can agree on. Case fatality rates during seasonal influenza are highest among the oldest and youngest members of our population. Those with chronic diseases like diabetes, heart and lung disorders and those with chronic conditions that impair immunity also have a much higher case fatality rate with seasonal influenza than people without these conditions. Never in history has humanity faced an influenza pandemic enriched with so large a number of people possessing these high-risk characteristics. This factor was not considered in the US Government’s assumptions for case fatality during pandemic influenza. As a consequence, the case fatality rate stemming from a truly severe pandemic is likely to be higher than projected.
The relationship of severity of illness, treatment setting, and patient prognosis
The severity of the patient’s illness has an important influence on case fatality rate. Severity of illness simply refers to how sick the patient was during their illness. The treatment setting is defined as the place where the patient’s treatment is provided and is strongly related to outcomes. If the severity of illness and treatment setting for the patient is known then this can be used to predict the patient’s prognosis. A patient’s prognosis is simply their medical outlook or expected course during an illness. Patients with a poor prognosis are not expected to do well while those with a good prognosis are expected to make a complete recovery. A patient’s prognosis is related to their severity of illness combined with their state of health prior to admission. Treatment setting must also be considered when calculating a prognosis. For example, a moderately ill patient with a chronic but stable medical disorder may have a good prognosis if treated in a staffed hospital bed and a poor one it treated at home.
Deducing the severity of illness expected during the next pandemic
A breakdown of the severity of illness expected by the US Government during a severe influenza pandemic can be deduced from the data presented in Table One of the DHHS Pandemic Influenza Plan. This table projects where they expect the patient will be treated for influenza. For instance, the patients requiring respiratory ventilators or Intensive Care Unit (ICU) treatment carry the highest severity rating, with those being admitted to a regular staffed hospital bed being assigned an intermediate severity, and those projected to be treated at home having the lowest severity of illness.
| Assumptions used for assignment of prognostic types during pandemic influenza |
| Type 1 |
|---|
|
Patients designated in the Department of Health and Human Services Pandemic Influenza Plan
A 50% CFR is assumed for these patients treated in the hospital and a 95% CFR for patients in Type 1 when treated at home both under good conditions. |
| Type 2 |
|
Patients designated in the DHHS PIP:
It is assumed this group will have a hospital CFR of 15% with a home rate of about 50%. |
| Type 3 |
|
Patients designated in the DHHS PIP:
A CFR of 1% is assumed for this group if the hospital is functioning compared with 5% CFR for those treated exclusively home under good conditions. |
| ICU = Intensive Care Unit, CFR = case fatality rate, DHHS PIP = Department of Health and Human Services Pandemic Influenza Plan |
Since severity of illness is directly related to the patient’s prognosis, this analysis can be used to assign patients into 3 simple prognostic types. Prognostic type refers to the patient’s risk of dying from the flu and is related to the severity of their illness and their state of health before becoming ill. For this model I assumed Type 1 patients are critically ill, Type 2 patients are moderately ill, and Type 3 patients are mildly ill. Simply stated, prognostic Type 1 and Type 2 patients would be those the US Government expects to require hospitalization with the Type 3 patients being those they expect to be treated at home.
The table below is the US Government projections for a severe pandemic modified by a higher clinical attack rate of 40% and by prognostic type. The DHHS plan assumes that the hospitals will remain open and functional during the 18-month pandemic period. In addition, every severely ill patient will have access to a staffed hospital bed, ICU bed, respiratory ventilator, antiviral drugs, antibiotics, and all the other critical care supplies during the duration of the pandemic. While these assumptions seem optimistic to me, for the table below I have maintained the same treatment setting assumptions used by US Government. Adjusting the clinical attack rate and application of prognostic type to their projections result in a doubling of the expected number of deaths from 1.9 to 4.1 million in the US.
| The US DHHS Pandemic Influenza Plan mortality projections adjusted for clinical attack rate and prognostic type | ||||
| Treatment: Hospital Setting | ||||
|---|---|---|---|---|
| Prognosis Type | Patient Number | Prognosis by Clinical Attack Rate | Case Fatality Rates | Deaths Expected |
| Type 1 | 4,144,000 | 3.50% | 50% | 2,072,000 |
| Type 2 | 6,512,000 | 5.50% | 15% | 976,800 |
| Type 3 | 107,744,000 | 91.00% | 1% | 1,077,440 |
| Total | 118,400,000 | 100% | 3.49% | 4,126,240 |
The patient’s treatment setting has a powerful impact on outcome. The two treatment settings considered here is a staffed hospital bed verses home treatment. For my model, I have assumed that a severe pandemic will exhibit 3 discrete pandemic waves lasting 2 or 3 months each during the 18-months pandemic period and that the US hospital system will not be able to accommodate more than 1 in 3 of the patients projected to require admission by the US Government. My assessment is the hospital system lacks the capacity to accommodate this many additional patients. In the table below, I have assigned expected case fatally rates for all 6 possible combinations of prognostic type and treatment setting.
| Effect of prognostic type and treatment setting on death rates during influenza pandemic | ||
| Prognosis Type | Expected Outcome for Hospital Treatment Setting | Expected Outcome Home Treatment Setting |
|---|---|---|
| Type 1 | High mortality 50% | High mortality 95% |
| Type 2 | Moderate mortality 15% | High mortality 50% |
The next table combines all the adjustments made I think are reasonable to make to the US Government’s model.4
| Adjusted US HHS Pandemic Influenza Plan for prognosis, clinical attack rate and treatment setting | ||||
| Treatment: 1/3 Hospital Setting & 2/3 Home Setting | ||||
|---|---|---|---|---|
| Prognosis Type | Patient Number | Prognosis by Clinical Attack Rate | Case Fatality Rates | Deaths Expected |
| Type 1 | 4,144,000 | 3.50% | 83.33% | 3,453,333 |
| Type 2 | 6,512,000 | 5.50% | 38.33% | 2,496,267 |
| Type 3 | 107,744,000 | 91.00% | 3.67% | 3,950,613 |
| Total | 118,400,000 | 100% | 8.39% | 9,900,213 |
There are three primary reasons explaining the increased mortality rate expected in this adjusted model compared with the one presented in the US Government. These are the higher clinical attack rate, adjustment for prognostic type, and a change in treatment setting for 2 in 3 severely ill patients. These changes in the official assumptions result in an increase in the number of deaths during a severe pandemic to 9.9 million from 1.9 Americans. This corresponds to an increase in the overall case fatality rate from approximately 2% to 8% during the expected 18-month pandemic period.
Global pandemic mortality projections
For the worldwide death number, McKibbin and Sidorenko in their Lowy Global Economic Pandemic Study projected 142 million deaths for a severe pandemic5. Dr. David Nabarro, the UN coordinator for avian influenza, told the press in October 2005 that the death toll from a severe influenza pandemic could reach 150 million. Dr. Osterholm’s range of 180 million to 360 million is based on the current best estimate of world deaths during the 1918 event of 60 to 100 million deaths adjusted for an increase in the world’s population from 1.6 to 6.6 billion people.
The 8% case fatality rate seen in the adjusted US model proposed here is only possible by assuming that 1 in 3 of the critically ill would be treated in a modern hospital setting. This is an assumption that does not apply to 87% of the world’s population who live in the developing and third world. While the average age of the people that live in these regions is lower than in the developed nations, this may not be an advantage in the first instance because of cytokine storm and in any event is offset by a high prevalence of people at risk from influenza. These conditions include malnutrition, political instability, HIV-AIDS, tuberculosis, malaria, and chronic hepatitis C and B; conditions that will predictably inflame the pandemic. Health information for providing good home care to influenza victims using simple resources like the oral rehydration solution is unlikely to be available to the majority of these folks. The result is an increase in expected case fatality rate to 12.5% for the world as a whole.
| Estimates of the worldwide impact of a server pandemic | ||
| Case Fatality Rate Prediction | Case Fatality Rate Prediction | Deaths Worldwide Estimate |
|---|---|---|
| 7.36 | 30% | 142,000,000◊1 |
| 5.7% | 40% | 150,000,000◊2 |
| 6.8% | 40% | 180,000,000◊3 |
| 12.5% | 40% | 333,000,000◊4 |
| 13.6% | 40% | 360,000,000◊4 |
|
World Population 2005 = 6,600,000,000 |
||
Models are simple estimates of complex events
As you can see there is a lot of guessing that goes into making projections like this and the results can easily be manipulated by simply changing the clinical attack or case fatality rates used in the model. The adjusted US model proposed here is made more complex by introducing the notions of case severity, prognostic type, and treatment setting. While valid adjustments to make, the value assigned to each is pure guesswork on my part. There is a lot to recommend simple models but not when they fail to adequately fulfill their function as a reasonably facsimile of the event. While no model is perfect, the more realistically it prospectively patterns the actual event, the better it performs for those relying on it for planning. In my view it is better for a model to include factors with the potential to significantly affect patient outcomes when they are likely to occur than to ignore them. Including them provides a more accurate estimate of the morbidity expected with a model of pandemic influenza than one that fails to consider them.
¹ The US Department of Health and Human Services Pandemic Influenza Plan. November 2, 2005
² Osterholm M, Preparing for the next pandemic., N Engl J Med 2005;352:1839-1842
³ The Institute of Medicine’s Pandemic Influenza: Assessing Capabilities for Prevention and Response June 16 — 17, 2004.
4 In all honesty, the US Government’s estimates and certainly my adjustments and model are pure guesswork. The crux of the matter relies on how many severely ill people will make it into a functioning hospital and how many won’t. This is a very important adjustment because the value assigned to it has such a dramatic effect on the number of deaths expected during the pandemic. 5 McKibbin WJ, Sidorenko AA,. Global macroeconomic consequences of pandemic influenza, Lowy Institute, February 2006, www.lowyinstitute.org.