Figure 3.

Vaccine effectiveness as a function of patient age. A LASSO logistic regression model was fitted for influenza regressed on vaccination, with adjustment for age at enrollment, sex, race, current smoking, home oxygen use, underlying medical condition, timing of admission relative to influenza season, specific influenza season, insurance, and site of enrollment. Restricted cubic spline was applied to age and week of influenza season variables, with 3 knots for each variable. The interaction between vaccination and age was also included in the model. Vaccination, age (linear and nonlinear), and interactions between vaccination and age were not penalized. The gray zone is a 95% point-wise confidence band.