used on large numbers of patients at the point-of-care, would provide rapid results (in 30 minutes or less), and would require minimal skill to administer and interpret (Griffin, 2007; Grijalva et al., 2007).

Current diagnostic tools include the following:

• Culture (of the virus) is traditionally considered the gold standard, with 100 percent specificity, but a culture also requires 48 hours or longer, so it cannot be used to determine treatment with antivirals that require prompt use for maximum efficacy.

• Real-time polymerase chain reaction (RT-PCR) is at least as, and perhaps more, sensitive than culture. It currently is expensive, of limited availability, and requires somewhat sophisticated laboratory capabilities.

• Direct immunofluorescence (or direct fluorescent antibody assay, DFA) provides results in approximately 1–6 hours and has 80–95 percent sensitivity.

• Enzyme immunoassay (EIA) rapid test gives results in 30 minutes, and has 50–90 percent sensitivity (median 70 percent) and 90–95 percent specificity.

• Serology is mostly useful for epidemiologic studies and takes more than 10 days.

Most tests available do not provide information about influenza A subtypes (CDC, 2006a; FDA, 2007b). Only PCR and culture currently can detect the H5N1 strain. Even fewer tools may be available if a different strain with pandemic potential emerges.

The performance of laboratory tests varies by the type of sample (swab versus aspirate or wash; nasal versus nasopharyngeal versus throat), the age of the patient (generally more sensitive in children because of higher viral loads), and the time since onset of symptoms (CDC, 2006b; FDA, 2007b). In addition, the operating characteristics (e.g., the sensitivity, specificity, likelihood ratio) of any laboratory or clinical diagnostic tool depend on the prevalence of influenza in the population, which influences the predictive value. For example, the positive predictive value of rapid tests is fairly good (i.e., most positives are true positives) when disease is highly prevalent, helping rule in true influenza. However, negative predictive value at the peak of a pandemic will be poor, so rapid tests would not be a good tool for denying treatment.

The clinical diagnosis of influenza is difficult because the symptoms overlap with those of other respiratory viral infections. A number of studies have examined the ability of clinical symptoms to predict which patients with respiratory infections have influenza (Call et al., 2005). The results are influenced by study design, the age of the patients, the