Calculating Unmeasurables

+ See all authors and affiliations

Science  22 Jan 2010:
Vol. 327, Issue 5964, pp. 395
DOI: 10.1126/science.327.5964.395-c

In assessing an epidemic, it would be helpful to know how many people were being infected each day. The objective data, however, are more likely to reflect how many people either became ill or died on a given day, time-lagged tallies that represent subgroups of the total infected population. For latent infections that simmer for years before producing symptoms, such as HIV, or for acute infections where time to death is variable, derivations of the desired incidence curve can be uncertain.

Goldstein et al. apply a mathematical method that was originally used for the purpose of extracting images from blur. Information about the time to death and the deaths per day was combined to calculate the incidence distribution, and the authors applied their approach to the influenza epidemic that struck Philadelphia in 1918. Their analysis suggests that in the few days between when the size of the epidemic became clear and when the city enacted closure of public gathering places, the spread of influenza was already being slowed significantly, probably by changes in individual behavior.

Proc. Natl. Acad. Sci. U.S.A. 106, 21825 (2009).

Navigate This Article