Swine flu updateDeath rate of swine flu difficult to measure

Published 15 July 2009

To formulate an effective policy to cope with the swine flu there is a need for an accurate set of numbers about the disease’s spread and morbidity; these number are hard to come by

Estimates of the proportion of people who will die if infected with swine flu are flawed, say U.K. researchers. At present, the estimate of the death rate in the United Kingdom and the United States is 0.5 percent, meaning that about five people die for every 1000 people infected. Accurate estimates are needed so that health authorities can best target treatment and vaccination strategies.

A new analysis suggests three main reasons why current estimates may be wide of the mark.

Andy Coghlan writes that the first and main source of uncertainty is the unknown number of infected people, who recover at home without notifying their doctors that they are ill, or receiving a diagnosis. Thus, although doctors know how many patients are dying of swine flu in hospitals, they do not know what proportion of all cases are life threatening. They need both figures, however, to work out the “case-fatality ratio” — calculated by dividing the number of fatal cases by the total number of cases.

We don’t know the denominator,” says Azra Ghani, head of a team at Imperial College London tracking development of the epidemic in the United Kingdom. “For that reason, dividing the number of deaths by the number of cases may be flawed,” says Ghani’s colleague Tini Garske, the lead author of the study exposing gaps in the data.

Coghlan writes that a second source of uncertainty is the possibility that deaths from swine flu are being attributed falsely to other causes of death, such as heart attacks or pneumonia from other causes. This would lead to underestimates of the death rate.

Finally, statistics are distorted by a time-lag between the point at which someone is infected and the time they die. This could lead to an apparent surge in deaths which may falsely be interpreted as the virus becoming more deadly through mutation.

Taken together, these factors make it difficult to rely on existing data sources to accurately calculate the death rate or to predict the course of the epidemic.

Ghani says, however, that studies are already planned to rectify these shortcomings. The aim is to monitor how the virus is spreading among people in the community, giving a representative “snapshot” of how the virus behaves and spreads between people in the real world. Once a reliable pattern has been established, it can be extrapolated to give a better national estimate of the denominator figure.

To do this, researchers hope randomly to select and monitor individual households initially free of the virus. By monitoring these, they can track the spread of the virus when it does arrive in the household, even among those who are infected but show no symptoms. “It’ll give a much better idea of those whose symptoms are so mild they don’t make it into the family doctor’s surgery,” says Ghani.

Coghlan writes that another possibility for future study is randomly to test people in the community to see if they have antibodies to the virus. This will have to wait, says Ghani, until a test to detect these antibodies is available.

Equally vital is research to detect any changes in the genetic make-up of the virus that are likely to make it more infectious or dangerous. “We have past examples of genetic markers of mutation that would make it more virulent, so we can monitor for those,” said Ghani. “Also, we can see how virulent these strains are in animal studies, and obtain data sets in the human population to see whether it changes virulence, and that would take away the uncertainty of statistics alone,” says Ghani.