Epidemiology of influenza-associated hospitalization in adults, Toronto, 2007/8

In prospective surveillance for laboratory-confirmed influenza infection in patients admitted to six participating hospitals in Toronto during the 2007/8 influenza season, a triage temperature ≥38.0°C, the presence of respiratory symptoms, admission diagnosis of respiratory infection, or exacerbation of COPD or asthma, and admission during peak influenza weeks were independent predictors for influenza. The likelihood of influenza exceeded 15% in patients with suspected respiratory infection or exacerbation of COPD/asthma if they were admitted in peak weeks during the influenza season, and exceeded 20% for patients who were febrile at triage and had a respiratory diagnosis or respiratory symptoms in the emergency department. These findings may assist clinicians in decisions regarding influenza testing of patients being hospitalized, and may better guide decisions regarding empiric antiviral therapy in adult acute care.

Our data suggest that these findings hold true for both influenza A and B, for both younger and older adults, and in patients being admitted to both medical services and ICUs. While they are limited to a single geographic area during a single influenza season, they are similar to those from previous studies in other years in both North America and Europe. Walsh et al., in a single hospital over three influenza seasons from 2001–2004, identified that 154 of 1471 (10.5%) older adults admitted with a cardiopulmonary diagnosis during the influenza season had influenza, and that fever was more than twice as common in patients with influenza as patients without [15]. van den Dool et al., during the peaks of two influenza seasons, identified that 8.5% of patients (23 of 271) admitted to internal medicine, geriatrics, respirology, or infectious diseases wards with any diagnosis had influenza [16].

These studies, and that of Babcock et al., also found that clinical signs and symptoms are of limited value in diagnosing influenza [17]. Walsh et al. found that, during a one-year epidemic period, the complex of cough, temperature of 38°C or higher, and illness duration of 7 days or less in older adults with a cardiopulmonary admitting diagnosis yielded a sensitivity of 78% and a specificity of 73%, a positive predictive value of 40%, and a negative predictive value of 91% [15]. However, other authors have, in general, not been able to identify symptoms complexes with positive or negative predictive values much above 50%. As Call et al. noted in their 2005 meta-analysis: “Unfortunately, no specific symptom or combination of symptoms is diagnostic of this common infection” [18].

The goal of this study, however, was not to identify symptom complexes that could diagnose influenza, but, rather, to identify populations of patients in whom the probability of influenza was high enough that either diagnostic testing or empiric antiviral therapy were warranted. An increasing number of studies suggest that specific antiviral therapy is effective in reducing the morbidity and mortality associated with influenza [7, 8], and current guidelines for both seasonal and pandemic influenza recommend the treatment of patients requiring hospitalization for influenza [19, 20]. In outpatients, early therapy with specific antivirals is more effective than later therapy [21], and there is good evidence for bacterial infection requiring hospitalization that early, effective anti-bacterial therapy improves outcomes [22‐24]. Our data may be particularly important for clinicians considering empiric antiviral therapy, where understanding the pre-test probability of acute infection is needed for rational use.

As in other studies of influenza in adults, our data suggest that reverse transcriptase polymerase chain reaction (RT-PCR) is significantly more sensitive than viral culture or DFA for influenza diagnosis, and these tests are, in turn, more sensitive than commercial rapid influenza tests. In adult patients requiring hospital admission, RT-PCR is the only diagnostic test with adequate performance characteristics for the diagnosis of influenza [19].

There are a number of limitations to our study. Only 63% of eligible patients were tested for influenza, suggesting that the proportion of persons actually infected might be significantly higher than our estimate. The fact that patients in whom clinicians accepted recommended orders were more likely to be febrile, have respiratory symptoms, or have a high-risk diagnosis suggests that clinicians were selecting out patients more likely to have influenza for testing. As in Babcock et al.’s study [17], it is possible that patients whose clinicians did not order specimens did not have influenza. However, other unknown biases may have existed. Data collection by chart review limited the number of risk factors that might have been considered, and meant that we could not consider any differences that might have been found between vaccinated and unvaccinated persons. In addition, we assessed only respiratory symptoms overall rather than individual specific respiratory symptoms (e.g., cough, shortness of breath). Although our results are similar to those of studies in other seasons and geographic areas, there may still be significant year-to-year and area-to-area variation. In support of this, risk factors for infection with influenza A(H3N2) and influenza B appear to be somewhat different in our analysis. In addition, as expected, patients admitted with influenza during weeks with predominant A(H1N1) activity were younger than others [25]. Although there were too few A(H1N1) infections diagnosed to compare risk factors for A(H1N1) compared to A(H3N2) and B infections, such predictors may be different, in part, because patients infected with influenza A(H1N1) are expected to be younger than patients infected with influenza A(H3N2) or B [25]. Finally, patients who are admitted to hospital with influenza infection likely do not constitute a homogenous group. The proportion of patients who need specific antiviral therapy to control influenza infection versus the proportion who are suffering from a complication of influenza but are able to control viral replication themselves remains unknown. Further study to define the causes of hospitalization due to influenza and the potential of antiviral therapy are clearly needed.

The identification of patients shedding influenza virus is also important for infection prevention. If additional precautions for patients admitted with influenza are an important element of the control of transmission of influenza in acute care hospitals, then our data suggest that additional precautions should be implemented for all patients with respiratory infections and COPD or asthma exacerbations during the influenza season, until the results of influenza testing rule out influenza. The cost and impact on patient care of such an approach highlights the need for a better understanding of when patients with influenza are infectious and of the epidemiology of healthcare-associated influenza transmission.

In conclusion, we believe that, in febrile patients with suspected respiratory infection or exacerbation of COPD or asthma, influenza testing and empiric antiviral therapy may be warranted during influenza seasons. During weeks in which >15% of surveillance specimens are expected to be positive for influenza, testing and empirical treatment might be indicated in any patient with respiratory symptoms. This approach may not contribute to the development of resistance, as the treatment of patients who do not have influenza does not create selective pressure for drug resistance, because current antivirals are influenza-specific.