Predicting influenza infections during epidemics with use of a clinical case definition

Combined pharyngeal and nasal swab specimens were collected from 100 subjects who presented with a flu-like illness (fever >37.8 degreesC plus 2 of 4 symptoms: cough, myalgia, sore throat, and headache) of <72 hours' duration at 3 different clinics in the province of Quebec, Canada, during the 1998-1999 flu season. The rate of laboratory-confirmed influenza infection was 72% according to cell culture findings and 79% according to the results of multiplex reverse-transcription polymerase chain reaction (RT-PCR) analysis (85%, influenza AH3; 15%, influenza B). All subjects for whom these results were discordant (negative culture and positive PCR) presented with a temperature <greater than or equal to>38.2 degreesC as well as 3 or 4 of the symptoms in the clinical case definition. Stepwise logistic regression showed that cough (odds ratio [OR], 6.7; 95% confidence interval [CI] 1.4-34.1; P =.02) and fever(OR, 3.1; 95% CI, 1.4-8.0; P =.01) were the only factors significantly associated with a positive PCR test for influenza. The positive predictive value, negative predictive value, sensitivity, and the specificity of a case definition including fever (temperature of >38 degreesC) and cough for the diagnosis of influenza infection during this flu season were 86.8%, 39.3%, 77.6%, and 55.0%, respectively.