infectious agents, estimates of influenza-like illness would over-state the effects of the influenza virus. (The degree of exaggeration depends on the relative prevalence of other viruses at the time the estimate is made.) Second, influenza viruses not only cause death directly and in association with bacterial pneumonia, but very often may contribute to death in patients with other, serious primary illnesses, such as heart, lung and renal disease. In fact, during a typical year, the “influenza-related” deaths due primarily to other diseases are believed to outnumber those directly due to influenza and pneumonia.41
In order to detect the occurrence of influenza epidemics and to capture their full impact on mortality, the CDC has for years relied on a derived index, called “excess mortality.” Mortality rates normally show a regular year-round fluctuation, highest in winter and lowest in summer. The CDC currently receives weekly mortality counts (total deaths and those attributed to pneumonia and influenza) from 121 urban centers around the country, comprising about 30 percent of the U.S. population. The CDC compares the observed mortality with the “normal” curve, which is based on a composite of several years’ experience. If the reported mortality exceeds a certain threshold for two consecutive weeks, this is considered indicative of an epidemic. CDC sums the number of excess deaths reported by the 121 cities during the flu season (usually 2-3 months), computes an “excess mortality” rate per 100,000 population covered, and then extrapolates to the entire population to derive a total number of excess deaths in the country.42
Computation of excess mortality is a sensitive way to identify the occurrence of an epidemic, but it may be an inaccurate indicator of influenza’s importance as a national health problem. First, urban centers, having relatively dense concentrations of people, would be more likely to experience epidemic outbreaks; extrapolating from 70 million city dwellers to the entire country may therefore exaggerate national experience. Second, restricting the excess death counts to the influenza season fails to correct for those patients who would have died shortly (within the year) without any influenza. One old study concluded this effect was present, but small, and that most excess deaths occurred in people who were not just about to die anyway.43 A recent comparison of CDC’s calculated excess mortality with annual mortality data compiled by the National Center for Health Statistics (NCHS) suggests that CDC’s excess mortality estimates have tended to be too high in recent years.44
In addition to possible inaccuracies of this sort, the number of deaths is an incomplete measure of the importance of influenza virus as a cause of death For purposes of setting priorities among health programs, a vital, supplementary measure is the “years of life lost” due to disease.45 This is a function both of mortality rate and age at death. Everyone is going to die, and what is important is not the fact of death, but its prematurity, the number of years of life expectancy, foreclosed. This could be calculated from age-specific death rates compiled on an annual basis. But so far as we know, the calculation has not been done by CDC. Therefore, its attributions of mortality cannot be adjusted in this way. Elderly persons make up such a high proportion of influenza deaths that the adjustment could reduce flu’s relative importance as a cause of death in this country.
A further limitation to the CDC’s “excess mortality” measure is its inability to reflect the extent of non-fatal influenza. No mortality measure, even if otherwise perfect,