Sunday, May 21, 2006

Is the virus becoming more deadly?

The recent cases of bird flu in Indonesia have had an unusually high fatality. In the Sumatran cluster six of seven family members succumbed. Overall the case fatality in Indonesia is 78%. This has prompted a number of news sources to conclude the virus is becoming deadlier:
The bird flu is becoming more deadly, now boasting a 64 percent fatality rate.

According to the World Health Organization (WHO) 47 victims have died out of 73 cases in 2006. In 2004, however, 41 patients died out of 95 cases, which is only 43 percent.

The WHO believes that since 2003, 123 people have died from the virus, from the 217 documented cases. (All Headline News)
Maybe this is a good time for a quick review of Case Fatality Rate (CFR).

The first thing is that technically CFR isn't what epidemiologists today call a rate. Rates are the number of disease events per person per time unit or sometimes number of disease onsets in a time span. Thus a mortality rate for cancer might be 1 per 10,000 population per year. A CFR, by contrast, is a proportion. In technical terms it is an incidence proportion, a measure of average risk, in this case the average risk of dying of the disease given that you contracted it. It therefore varies between zero and 1.0 (rates vary from zero to an undetermined upper number that can be much greater than 1.0).

The CFR has a numerator (deaths from bird flu) and a denominator (all diagnosed cases of bird flu). What the number expresses is the risk of dying of bird flu once you get bird flu. From this viewpoint, that risk seems to be increasing (43% in 2004 to 64% today, overall, and 78% in Indonesia).

However there are complications. The numerator (deaths in bird flu diagnosed cases) may be accurate (death is a definite endpoint), but the CFR measures the risk of dying once counted in the denominator and that risk is subject to other factors than the deadliness of the virus--in particular care seeking behavior, timely access to medical care, the skill and resources of the care itself, and the general state of health of the patient. Thus the numerator may differ from place to place even though the virus remains exactly the same.

The denominator (diagnosed bird flu cases) is less firm than the numerator. WHO has a very strict standard for defining a case (requiring laboratory confirmation by one of its reference laboratories), so it is possible many truly existing cases are not being counted because they are not being diagnosed by anyone or are mild and do not seek medical care. This would artificially inflate the CFR by having a denominator smaller than it should be and many people thought this was a likely reason for the abnormally high CFR in this disease. Unfortunately there is no confirming evidence that many cases are being missed. While data from seroprevalence surveys is sparse (these are blood studies of the general population or contacts of confirmed cases to see if they have evidence of mild infection), so far they show stunningly little evidence of undiagnosed infection. Maybe the CFR slightly overestimates the true value, but there is as yet no convincing evidence it overestimates it by much.

Because the accuracy of both the numerator and denominator of the CFR might vary from country to country and within a country, comparing CFRs over a time period in cases from ten very different countries does not clearly measure a change in the deadliness of the virus. While it is quite possible the virus is becoming more virulent, we don't have a good way to determine that at the moment.

Treat the news headlines with the appropriate caution.