UCSF

Background: Through years of observational research it has become accepted that the influenza (flu) vaccination reduces mortality in an older population. However, despite 50 years of targeted administration, the national statistics for influenza-related mortality among the elderly are unchanged.

Objectives: We attempt to distinguish vaccine effective from selection bias using data from a California Veterans Medical Centers (n=38,736) age 65 and older.

Methods: Two methods were used to differentiate effect from bias. The first method was the development of a 24 month mortality prediction logistic model. The second method was stratification by risk of death into deciles from low risk (0.03) to high risk (0.59) and evaluating the vaccine's effect on mortality within each decile.

Results: There was very little change in model concordance between the model with influenza vaccine (77.8%) as a variable and the model without influenza vaccine (77.7%). For the division by risk deciles, we had precision to the third decimal point and could not detect a vaccine effect.

Discussion: We evaluated all-cause mortality across 7 years, starting in October of 2002 and ending September 2009, in an elderly California Veteran population. During this time there were 7,944 deaths. Vaccine coverage averaged 42%. We could detect no overt effect of the current vaccine dose given to veterans.