OBJECTIVE: We aimed to determine the complex of symptoms which has the highest predictive value for the diagnosis of influenza. METHOD: A questionnaire study with questions regarding the symptomatology of influenza among patients aged 60 and older (n = 1838). Thirty-four participating GPs recorded the symptomatology of patients who came to their general practice with influenza-like complaints. The validity of the diagnostic conclusion of the GP, as well as the diagnostic validity of the criteria of the International Classification of Health Problems in Primary Care (ICHPPC-2) and the Sentinel Stations in The Netherlands, was determined with the help of the predictive value and odds ratio, using serologically confirmed influenza as the gold standard. The same method was used to determine which complex of symptoms has the highest predictive value for influenza. The results were verified using logistic regression analysis. RESULTS: The predictive value of the diagnostics of the GP amounted to 35%. The predictive values of the diagnostics according to the criteria of the two classification methods were 24% (Sentinel Stations) and 18% (ICHPPC-2). Of the individual symptoms, the combination of fever, coughing and acute onset had the highest predictive value (30.3%) for the diagnosis of influenza. CONCLUSION: It is recommended that the criteria of the Sentinel Stations in The Netherlands and the ICHPPC-2 be adapted in the following way: influenza is likely if, out of the entire complex of symptoms, at least fever, coughing and an acute onset occur.