Logistic regression analysis of independent predictors for post-stroke infections

Shown are the nominal p-values and estimates for the parameters remaining in the model after backward selection. Factors and covariates included for selection were: NIHSS at admission, age, tube feeding, mechanical ventilation, HLA-DR expression (Day 1), treatment, and interaction between treatment and HLA-DR (Day 1) expression. The final logistic model was: log( p/1-p) = μ+treatment+ßtreatment*log(HLA-DR); p: number of infected patients within 11 days/total number of patients (in the respective treatment group).