Predicting pneumonia and influenza mortality from morbidity data

PLoS One. 2007 May 23;2(5):e464. doi: 10.1371/journal.pone.0000464.


Background: Few European countries conduct reactive surveillance of influenza mortality, whereas most monitor morbidity.

Methodology/principal findings: We developed a simple model based on Poisson seasonal regression to predict excess cases of pneumonia and influenza mortality during influenza epidemics, based on influenza morbidity data and the dominant types/subtypes of circulating viruses. Epidemics were classified in three levels of mortality burden ("high", "moderate" and "low"). The model was fitted on 14 influenza seasons and was validated on six subsequent influenza seasons. Five out of the six seasons in the validation set were correctly classified. The average absolute difference between observed and predicted mortality was 2.8 per 100,000 (18% of the average excess mortality) and Spearman's rank correlation coefficient was 0.89 (P = 0.05).

Conclusions/significance: The method described here can be used to estimate the influenza mortality burden in countries where specific pneumonia and influenza mortality surveillance data are not available.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Humans
  • Influenza, Human / epidemiology
  • Influenza, Human / mortality*
  • Pneumonia / epidemiology
  • Pneumonia / mortality*
  • Poisson Distribution
  • Population Surveillance
  • Seasons