Rethinking thresholds for serological evidence of influenza virus infection