Hemagglutination-inhibitory antibodies are usually highly strain specific with little effect on infection with drifted or shifted strains. The significance of broadly cross-reactive non-HAI anti-influenza antibodies against conserved domains of virus glycoproteins, such as the hemagglutinin (HA) stalk, is of great interest. We characterize a cohort of 40 H1N1pmd09 influenza-infected patients and identify lower respiratory symptoms (LRSs) as a predictor for development of pneumonia. A binomial logistic regression of log10 pre-existing antibody values shows that the probability of LRS occurrence decreased with increased anti-HA full-length and stalk antibody ELISA titers. However, a multilevel logistic regression model adjusted by other potential serocorrelates demonstrates that only antibodies directed against the stalk of HA correlate with protection from lower respiratory infection, limiting disease progression. Our predictive model indicates that a threshold of protective immunity based on broadly cross-reactive HA stalk antibodies could be feasible.