Influenza-like illness criteria were poorly related to laboratory-confirmed influenza in a sentinel surveillance study

OBJECTIVE: To analyze the most related clinical data for influenza and the utility of influenza-like illness criteria as the clinical threshold for sampling in an influenza sentinel surveillance over a 3-year period. METHODS: Sentinel physicians collected throat specimens and data from outpatients with acute respiratory infection (< or = 72 hours duration). Laboratory-confirmed influenza infection was compared with independent symptoms and the influenza-like illness criteria, as defined by the Classification Committee of the World Organization of Family Doctors. RESULTS: From 1934 patients, 359 (18.56%) yielded positive results for influenza viruses. Only 199 (55.4%) of laboratory-confirmed cases fulfilled clinical criteria of influenza-like illness: positive and negative predictive value (PPV and NPV) of 0.36 and 0.88, respectively. Fever, cough, and rhinorrhea individually correlated with influenza infections (PPV: 0.30, 0.20, and 0.20, respectively; NPV: 0.92, 0.87, and 0.85, respectively). Multivariate analysis demonstrated that the correlation of influenza infection with the presence of fever and cough was similar to the correlation between influenza infection and influenza-like illness criteria (odds ratio 2.24 vs. 2.71, respectively). CONCLUSION: Influenza-like illness criteria are poorly related to laboratory-confirmed influenza. For early detection of influenza viruses in surveillance systems, a less restrictive clinical criterion (specifically, acute respiratory infection) perhaps should be followed.