You can search for articles in back issues of Contingencies from July/August 2000 to March/April 2009 using the search box to the right. Simply type in subject words, author's name, or article title and click search. To search for articles from May/June 2009 to the present, go to the current digital issue of the magazine and use the search function on the left of the top navigation bar.
The Cost of Avian Influenza
by Howell Pugh
SOUTHEAST ASIA is experiencing a severe outbreak of avian influenza (H5N1). If the potential for this outbreak to turn into a pandemic in the human population is real, as many experts believe, then the U.S. insurance industry is not adequately prepared or reserved. Health insurers need to begin planning, and regulators need to start stress-testing the life-insurance block.
Research has identified three essential prerequisites for the start of a pandemic: Transmission of a novel viral subtype to humans, the ability of the virus to replicate in humans and cause disease, and efficient human-to-human transmission.
Since 1997, the first two prerequisites have been met on four occasions. The most recent has occurred in Vietnam and Thailand in 2004 and continues well into 2005. Currently, the H5N1 influenza strain is endemic in poultry in many parts of Asia. This means the world is vulnerable to pandemic and will likely remain that way for years.
The Spanish flu pandemic in 1918 was the most deadly infectious outbreak in recent history. Recent research has placed the worldwide death toll at between 50 million and 100 million. What began in the spring as a relatively mild “herald wave” returned in the early fall as a fast-moving and fast-killing pandemic wave.
The second wave erupted simultaneously in the United States, France, and Sierra Leone and then spread worldwide through the sea lanes. This version of the flu was marked by unusually severe symptoms such as internal and external bleeding, difficulty breathing to the point of turning blue (cyanosis), and severe pain.
Most striking was the pattern of mortality. Normally, influenza causes the most deaths in the very old and the very young. This infection was different. In the United States, 99 percent of the deaths occurred below age 65. The highest number of deaths was between ages 25 and 29. This pattern gave the famous W curve (Fig. 1, page 24).
Avian influenza threatens to become one of the worst pandemics in recent history. Yet North America contains only 4 percent of the world’s capability for manufacturing flu vaccine. Is the U.S. life and health insurance industry ready?
The virus affected between 25 percent and 30 percent of the population. The overall mortality rate was about 2.5 percent of those infected; hardest hit were pregnant women. There were reports that among some populations, such as troops stationed in close quarters, and immunological naïves such as Alaskan Eskimos, the rate of mortality could have reached 20 percent of those infected. For the entire United States, the death toll is estimated at more than 600,000.
Medical treatments common today, such as vaccines and antibiotics, had not yet been discovered. Communities resorted to control efforts such as isolation, quarantine, disinfection, and banning public gatherings. The extreme speed of the spread and the high morbidity overwhelmed the response capabilities. Hospitals were forced to turn away patients, who had to resort to makeshift facilities. Cities ran out of grave space and coffins. Philadelphia used steam shovels to dig mass graves and brought in supply trains of coffins under guard.
The pandemic of 1957 arrived in a world much better equipped. In 1933, influenza was found to be caused by a virus. There were vaccines that had proved effective in reducing the incidence of morbidity in seasonal epidemics by two-thirds. Antibiotics were now available to treat the complications of influenza, especially bacterial pneumonia. The World Health Organization (WHO) had established the global surveillance network as an early-warning system. Thus, cases in Hong Kong alerted the world community in early May. By mid-month, the virus had been identified and samples provided to vaccine manufacturers.
This time, the virus was much milder than in 1918. In the United States, it hit the population in two waves, starting in September and then again in January of 1958. There was still high morbidity, but deaths conformed more to typical seasonal flu epidemics. The second wave mainly struck the elderly and caused the increased mortality. In the United States, the excess deaths were more than 60,000.
Limited quantities of the vaccines became available in the United States in August. There was manufacturing capacity to produce only enough vaccine to protect priority groups. No country, including the United States, had the capacity to cover its entire population, much less export vaccines elsewhere.
The Hong Kong flu pandemic beginning in 1968 was even milder than that of 1957. The geographical spread was rapid, as usual, but symptoms were mild and morbidity was low. The United States was an exception. There was a significantly higher number of influenza-related pneumonia deaths and about 34,000 excess deaths. By contrast, Canada had only a slight increase in morbidity and no excess deaths.
Once again, vaccine delivery was too little and too late. Even with a six-month lead, there were only 20 million doses in the United States when the epidemic peaked.
Though current science can’t predict how deadly a virus strain will become, we do know a few things about flu pandemics in general.
A pandemic will create a surge in cases over a short number of weeks. There will be a strain on health services even before considering the loss of health workers to the illness.
The deadly impact seems to be even more prominent in the second wave. This should mean that there is some window of time, whether one month or six, that could be used to create a vaccine and distribute it. This strategy has never been successfully tested in past epidemics.
Pandemics appear to have a different mortality pattern by age than the usual seasonal epidemic. While this was especially true in 1918, both of the two later episodes show a tilt toward more excess deaths at the younger ages (Fig. 2, page 25). This pattern is meaningful for life insurers as the spike occurs in the prime insurance ages. This effect dissipates as the years pass and the seasonal epidemic pattern reasserts itself.
Even if vaccines are available and used, countries with domestic manufacturing capacity will be the first to receive the vaccine. This is a particular concern of the United States. In the list of 21 influenza vaccine manufacturers maintained by the WHO, only one is based in the United States. Over 70 percent of the world’s manufacturing capacity is headquartered in Europe, while only 4 percent is based in North America.
Indeed, in the United States, recent events have shown that the country is failing to provide properly for regular annual and seasonal vaccinations. In 2003–2004, an early flurry of flu deaths in children caused strong demand for flu vaccine that exceeded the supply. In the fall of 2004, the British government shut down the factory for Chiron, one of the two European suppliers of that year’s U.S. supply, due to contamination problems. Chiron had been expected to provide one-half of the U.S. supply for 2004–2005.
As a result, the Centers for Disease Control and Prevention (CDC) of the U.S. Department of Health and Human Services issued an advisory urging healthy adults to forgo their shot for that year in order to supply doses for the elderly, the youngest, and the sickest. There were scenes in the media of people in long lines and even stories of Americans traveling to Canada seeking shots. By January of 2005, the CDC reversed itself and allowed shots to be given to anyone.
If the United States can’t provide routine influenza vaccine in a routine influenza season, how equipped will it be to respond to a real influenza crisis or a pandemic?
Today’s Threat: H5N1
First appearing in Hong Kong in 1997, virus H5N1 originated in Chinese geese and found its way into the poultry markets. It was able to infect humans directly. Eighteen cases were identified, and there were six deaths. This outbreak was halted by the quick elimination of 1.5 million live birds in only three days, which removed the opportunities for further human exposure.
A striking feature of the Hong Kong outbreak was the presence of primary viral pneumonia in severe cases. When pneumonia occurs in influenza patients, it’s usually a complication caused by a secondary bacterial infection. In the H5N1 cases, pneumonia was directly caused by the virus, didn’t respond to antibiotics, and was frequently and rapidly fatal. With one exception, none of these patients had underlying disorders that could explain the severe course of the disease.
The next sighting was in February 2003, when a Hong Kong family became sick after a trip to southern China. The man, age 33, died, and his son recovered. First wave—January to March 2004. Early in January 2004, Vietnam reported a cluster of 11 children in a hospital with severe respiratory disease that did not respond to antibiotics. Seven had already died, but subsequent lab work was able to confirm H5N1 in three of those cases.
At the same time, outbreaks in poultry farms were beginning. In three weeks, more than 400 sites were reported involving 3 million birds. In Japan, a farm near Kyoto reported an outbreak of highly pathenogenic avian influenza. In late January, Thailand reported its first human cases of H5N1 in two small boys. A flock of 70,000 birds was infected. By the end of the month, Thailand reported 156 outbreaks affecting 11 million birds. And a continuing trickle of human cases was reported.
In early February, the avian flu had also been discovered in China, Indonesia, Cambodia, Laos, and Korea. More than 120 million birds were culled to try to stem the outbreak. In the scale of the countries involved, the number of birds, and the number of human cases, this was the largest outbreak ever experienced. With rigorous culling, the outbreak subsided by March and seemed to have disappeared.
Second wave—July 2004 to March 2005. In July, fresh outbreaks began in many of the same countries. Malaysia joined the list in August. By November, Thailand was reporting the first probable case of human-to-human transmission. Fewer poultry were affected—only 1 million in the second wave. The number of human cases in Vietnam and Thailand totaled nine, with eight fatalities.
There was clear evidence that the influenza virus was mutating. Comparisons of samples over time showed the virus becoming increasingly pathogenic and lasting longer in the environment. It was found in its highly pathogenic form in dead migratory birds. This was highly unusual since wild waterfowl are the normal reservoir of Influenza A, but usually carry only low-pathogenic forms. Finally, ducks were found to excrete H5N1 in its highly pathogenic form.
By mid-March 2005, a total of 69 confirmed cases of avian influenza had been counted. There were 46 deaths, overwhelmingly children and young adults (Fig. 3, page 26). It’s not known how many cases had milder forms of the disease and thus would not have sought out medical attention. This means we can’t determine what the case mortality is for this disease.
Normally, human infection with avian influenza is extremely rare. At most, it causes viral conjunctivitis. H5N1 was able to pass directly from bird to human and caused severe disease with high mortality. This creates the potential for pandemic if the virus is able to change into a form that can be transmitted from human to human.
There are two ways this could occur. The influenza virus is capable of a reassortment of genes if both avian and human forms are in the same individual. This would involve a human host serving as a mixing vessel that would allow the avian form to become more easily transmissible. Both the pandemics of 1957 and 1968 are known to have been caused by the exchange of genes between avian and human influenza viruses.
The second method is for the influenza virus to undergo a series of step-like changes, which occur as the virus mutates during the infection of humans and gradually allows the virus to improve its transmissibility among humans. The 1918 pandemic is believed to have begun from an avian virus that acquired the adaptations to sustain human-to-human transmission. This would mean that surveillance workers should look for independent chains of limited human-to-human transmission to see if the virus is changing.
Controlling the Pandemic
In the best of all possible worlds, the pandemic would be controlled by worldwide surveillance, effective vaccines, sufficient antiviral supply, and adequate medical facilities to handle flu surges. In the real world, all of these present problems as well as opportunities.
Surveillance. The WHO has created a Global Influenza Surveillance Network consisting of 133 national centers in 84 countries. The centers collect influenza viruses circulating in different parts of the world and forward them to four research centers. This is the basis for deciding which virus strain should be in the annual vaccines that are prepared for the northern and southern hemispheres.
An example of the effectiveness of this system was the containment of the severe acute respiratory syndrome (SARS), which is related to the cold virus, outbreak in 2003. The WHO has identified and isolated the H5N1 virus and supplied samples to labs around the world.
Vaccines. The traditional method of manufacturing influenza vaccines involves the use of fertilized egg embryos. This is the only proven and rigorously tested means of large-scale vaccine manufacture. But it creates a technical problem for the H5N1 vaccine since the avian flu virus often kills the egg embryo.
One way around this is to use reverse genetics to deactivate the virus. This is considered intellectual property, however, so extra licensing and fees are anticipated. In addition, Europe regulates reverse genetic material as a genetically modified organism. This means that safety concerns require higher biosafety standards for manufacturing facilities. Currently, 70 percent of the world’s vaccine manufacturing capability is located in Europe. Upgrading plants can be done but at extra cost and additional time.
Clinical trials of H5N1 vaccines have already been planned. The National Institute of Allergy and Infectious Diseases (NIAID) initiated vaccine production and testing in May 2004. The current plans are to proceed with 450 subjects in clinical trials in 2005.
The main problem posed by vaccines is the inadequate manufacturing capacity to provide sufficient vaccine doses to handle the population in the event of a pandemic. Vaccine in limited doses means that there would have to be decisions about which groups should be targeted for delivery.
Antivirals. Antiviral drugs are useful in reducing the severity or length of the flu disease. They operate by reducing the ability of the virus to enter the cells of the body or by reducing its ability to multiply.
There are two classes of antiviral drugs used for influenza. The oldest type is prone to drug resistance by the flu virus. The newer drugs—oseltamivir and zanamivir—have a better safety profile and are less prone to the development of drug resistance. For these two drugs, the main constraints are price and supply. Drug supplies are very limited. Surge capacity for production is negligible, but the drugs can be stored for long periods.
Antivirals do have a role to play in the current situation and in the event that human-to-human transmission becomes more efficient. They are currently used as a treatment for diagnosed cases in Vietnam and Thailand and as a prophylactic (preventive) drug for clearly defined risk groups such as health workers, family members, or others closely exposed to existing cases.
In the event of increased transmissibility, antivirals could be used in a mass treatment to try to hold the pandemic in partial check and thus buy time to develop a vaccine. Once a pandemic has been declared, the drug would be invaluable before the development of a vaccine by giving victims some ease and hope. In order for antivirals to be effective, however, there should be enough in the national stockpiles to serve the population. Currently, the United States has plans in place to order doses for 7 million people.
Consequences for Insurers
The main risks to the insurance industry are morbidity, asset deterioration, and mortality. In order to assess the effects of an influenza pandemic on the insurance industry, it’s necessary to employ models. With only three data points in the past century, there’s little in the way of experience to draw from. The three pandemics are statistical outliers from the normal range of mortality. Medical science hasn’t gleaned enough to come to the rescue; it’s unable to tell either when the pandemic will emerge or how severe it will be.
The CDC commissioned a study of a pandemic influenza model. This study generated best and worst cases:
This model based the estimated deaths on rates from 1957 at the low end and on a multiple at the high end. As such, it doesn’t include either the 1918 or 1968 pandemics within its low-to-high range of deaths.
As Max Rudolph points out in his article, “Influenza Pandemics: Are We Ready for the Next One?” in the July 2004 issue of Risk Management, the equivalent of 1918 in today’s time would cause a $100 billion loss event. David Ingram, in the March 2005 issue of Risk Management, has argued that a relative estimate would be a $15 billion-to-$20 billion loss event.
Science doesn’t give us the answer. We can’t know the amount of the loss in advance; we can only know that there is a high probability of some loss occurring.
Morbidity. In the CDC model, hospitals would sustain an influx of patients equal to 0.1 percent to 0.3 percent of the population over a six- to 10-week period. In the United States, this would severely affect the hospitals. According to the Organization for Economic Cooperation and Development (OECD), the country has 2.9 acute-care hospital beds per 1,000 people. Thus, anywhere from one-third to 100 percent of the acute-care hospital beds would be needed for influenza patients.
This would occur during a time when hospital and health workers would be affected equally by the pandemic, suffering losses of workers by about 15 percent to 25 percent. In 1918, the medical services staff had been drafted in large numbers into the armed services. This prevented adequate help in the hospitals and clinics during the course of the epidemic.
Assets. The asset base for the insurance industry would also be at risk. In 2003, both Hong Kong and Toronto suffered through an episode of the SARS virus. The cost in tourism and trade was tremendous, and there was a long lag period before tourism levels returned to normal.
A full-scale epidemic would have enormous impact on business. During 1918, in Geneva, Switzerland, all public gatherings were banned. Concerts and theaters closed. Even open-air church services were forbidden. Similarly, pictures of the main shopping streets in Philadelphia show an empty landscape. There were no people who would brave being out in public to expose themselves to possible infection. Factories closed down due to the absence of workers, who were sick.
In today’s economy, what would be the impact of closing movie theaters, shopping malls, and sporting events? What if restaurants were closed or limited to take-out? If a large number of long-distance truck drivers were unavailable to drive for two weeks, for example, there might be difficulties in distributing items such as perishable food or components for just-in-time manufacturing. Equity and bond investments would suffer right at the moment when they would be called upon to pay claims. Even a relatively small asset impairment, say between 2 percent and 5 percent, would increase the adversity of heavy mortality.
Mortality. The risk of mortality would depend on the severity of the influenza. At present, we can only guess at the range of possibilities, from mild to severe.
“Fat tails” are statistical outliers on the far edges of bell curves. They’re the things that are never supposed to happen but actually happen rather frequently. Because they’re not supposed to happen, they’re often priced incorrectly. For example, you can buy insurance against an asteroid hitting your house fairly cheaply because it never happens; insurance companies figure they’ll never have to pay off. But all the models treat events as if they were purely random, known events, such as you might get at Las Vegas.
When you flip a coin, you don’t know what you’ll get. You can build a model with a 50/50 probability that you’ll get either heads or tails. But in real life, the possible outcomes are unknown and completely unpredictable. They also happen to be the most dangerous as well as the most profitable.
A reoccurrence of 1918 severity would create a $100 billion event. This equals half the surplus of the entire life insurance industry at year-end 2002.
Clearly, the impact wouldn’t be shared equally by all companies. Some would certainly fail and go into state receivership. Reinsurers seem to be the most vulnerable due to the term focus of their business model. As one actuary quipped: “Reinsurers would drop faster than people.” State guaranty funds would have to kick in if insurers failed, and no one knows how readily they could respond to the demand for payment of claims due.
The final backstop for the states may be the federal government. Several states may band together and ask for consideration of a relief bill in the nature of the Terrorism Risk Insurance Act (TRIA) of 2002. However, a leading Republican spokesman, Rep. Tom Delay, has opposed efforts up to now to renew TRIA in this Congress. He stated, “It is imperative that the industry... work with Congress to develop a long-term solution that does not involve the federal government serving as a reinsurer or a permanent backstop. Nor can the government become a funding mechanism for the insurance industry.”
We can draw some analogies from the property/casualty industry. The events of 9/11 caused a $35 billion loss. The Florida hurricane season of 2004 was a $28 billion loss. In the most recent annual statement of the Berkshire-Hathaway Co., Warren Buffett stated that the property/casualty industry should prepare for a $100 billion loss in either hurricanes or earthquakes.
The insurance industry is subject to solving problems by using reserves, rating agencies, or regulators. It’s foolish to imagine that for pandemic influenza there will be valuation actuaries who will take the lead by increasing reserves. There are simply too many unknowns in advance of the actual pandemic event. This means there is a clear role for industry groups to play.
The Academy could begin to bring its expertise in low-frequency/high-severity events from the property/casualty side over to the life insurance side. It could also assist health insurers in creating pandemic preparedness plans.
The Society of Actuaries could create a task force to look at the possible range of mortality in the event of an influenza pandemic and recommend appropriate stress tests for a company’s insurance block.
The alternative is that the industry heads into a possible severe condition with no planning and no preparation. It would remain for the other R’s—regulators and rating agencies—to assert their roles in the planning for the next pandemic.
HOWELL PUGH is a consulting actuary in Indianapolis.
Avian Influenza, Assessing the Threat, World Health Organization, Geneva, Switzerland, January 2005.
The Threat of Pandemic Influenza, Are We Ready? Stacey L. Knobler, Alison Mack, Adel Mahmoud, Stanley M. Lemon, Editors, Institute of Medicine, Washington, D.C., Nov. 2004.
Barry, John, The Great Influenza, Viking Press, New York, N.Y., 2004.
Simenson, L.E., Clarke. M.J., Schonberger, L.B., Arden, N.H., Cox, N.J., Fukada, K., “Pandemic Versus Epidemic Influenza Mortality: A Pattern of Changing Age Distribution,” Journal of Infectious Diseases, 1988;178:53-60.
Contingencies (ISSN 1048-9851) is published by the American Academy of Actuaries, 1100 17th St. NW, 7th floor, Washington, DC 20036. The basic annual subscription rate is included in Academy dues. The nonmember rate is $24. Periodicals postage paid at Washington, DC, and at additional mailing offices. BPA circulation audited.
This article may not be reproduced in whole or in part without written permission of the publisher. Opinions expressed in signed articles are those of the author and do not necessarily reflect official policy of the American Academy of Actuaries.