To predict influenza related pneumonia - a continuous challenge during pandemics

Abstract

Background: The last influenza pandemic showed the importance of the early beginning of antiviral therapy for a successful outcome of influenza related to pneumonia (IRP). Thus the ability to differentiate the influenza pneumonia from the bacterial one, during the first hours after the patients admission, is crucial for further management of the case. A study was performed to investigate whether adults with severe pneumonia could be distinguished clinically form patients with n o n - H ^ community acquired pneumonia (CAP). Methods: Clinical and epidemiological data of 75 adults admitted for severe H Nj IRP were compared with a prospective study cohort of 127 adults with severe n o n - H ^ CAP admitted during interpandemic period. A multivariate logistic regression model was generated for prediction of H N influenza related pneumonia. Results: In-hospital mortality in both cohorts was pretty similar, to 20% in H N | IRP cohort compared with 19,7% in n o n - H ^ CAP cohort. A diagnostic prediction model was derived by assessing one point for each of the seven criteria: demographic (age< 65 years), clinical (presence of myalgia/arthralgia, absence of hypotension, absence of pathological bronchial breathing), laboratory (leucocyte count < 12*109/1) and radiological (presence of bilateral involvement, extension to superior pulmonary areas). The threshold yield of the model was obtained for 4 points value of that, with a negative predictive value of 92,4% and 88% of sensibility.Accuracy of the obtained model was appreciated using the value of area under receiver operatingcurve, which corresponds to a very good one - 0,93 (95%CI 0,89 - 0,96). Conclusion: There are substantial clinical differences between HjNpnfluenza related to pneumonia and inter-pandemic CAP. A model based on seven accessible criteria allows the early identification of adults with severe influenza related pneumonia.