Viral co‐infections are associated with increased rates of hospitalization in those with influenza

Abstract Background Influenza causes significant morbidity and mortality in the United States. Among patients infected with influenza, the presence of bacterial co‐infection is associated with worse clinical outcomes; less is known regarding the clinical importance of viral co‐infections. The objective of this study was to determine rates of viral co‐infections in emergency department (ED) patients with confirmed influenza and association of co‐infection with disease severity. Methods Secondary analysis of a biorepository and clinical database from a parent study where rapid influenza testing was implemented in four U.S. academic EDs, during the 2014–2015 influenza season. Patients were systematically tested for influenza virus using a validated clinical decision guideline. Demographic and clinical data were extracted from medical records; nasopharyngeal specimens from influenza‐positive patients were tested for viral co‐infections (ePlex, Genmark Diagnostics). Patterns of viral co‐infections were evaluated using chi‐square analysis. The association of viral co‐infection with hospital admission was assessed using univariate and multivariate regression. Results The overall influenza A/B positivity rate was 18.1% (1071/5919). Of the 999 samples with ePlex results, the prevalence of viral co‐infections was 7.9% (79/999). The most common viral co‐infection was rhinovirus/enterovirus (RhV/EV), at 3.9% (39/999). The odds of hospital admission (OR 2.33, 95% CI: 1.01–5.34) increased significantly for those with viral co‐infections (other than RhV/EV) versus those with influenza A infection only. Conclusion Presence of viral co‐infection (other than RhV/EV) in ED influenza A/B positive patients was independently associated with increased risk of hospital admission. Further research is needed to determine clinical utility of ED multiplex testing.


| INTRODUCTION
Influenza infections are caused by influenza viruses with multiple circulating types, subtypes, and antigenic-distinct strains and are responsible for up to 650,000 annual deaths worldwide, and 61,000 deaths in the United States annually since 2010. [1][2][3] These infections pose a substantial burden on the healthcare system, given the significant annual morbidity and mortality. Certain groups of individuals are at greater risk for influenza-related complications, including the elderly, the immunocompromised, and individuals with chronic co-morbid conditions such as cardiovascular disease, cancer, diabetes, HIV/AIDS, and kidney disease. 1 Previous studies have shown that in individuals infected with influenza viruses, rates of bacterial co-infections are substantial, ranging between 11% and 35%. 4,5 The presence of bacterial co-infection in those with influenza has also been widely reported to be associated with more complicated disease course, higher risk of intensive care unit (ICU) admission, and increased mortality. 4,6,7 There has been relatively less research describing the overall burden and risks associated with viral co-infection among those with influenza. Reported rates of influenza viral co-infection from prior studies in varied settings range between 4% and 6%. [8][9][10] There have been a few studies that report increased risk of severe illness in influenza-infected patients found to have viral co-infections, but these have principally been restricted to inpatients. 8,11 The recent developments and now widespread availability of multiplex viral testing platforms provide an opportunity to identify viral co-infections in patients with influenza. We examined a large biorepository of nasopharyngeal (NP) specimens from a parent multi-center emergency department (ED) study of patients who tested positive for either influenza A or B (influenza A/B). Our objective was to define rates, and demographic and clinical correlates of viral co-infections in those with influenza A/B, as well as determine whether the presence of viral co-infection was associated with more severe illness, using inpatient admission as a proxy for disease severity.

| Study design and data collection
We conducted a secondary analysis of a biorepository and clinical data set from a parent study, which was designed to assess the impact of systematic influenza testing on treatment for influenza in the ED. 12 For the parent study, four participating sites (

| Patient consent statement
Patient written consent was obtained. The parent influenza cohort (JHU IRB00041135, JHU IRB00141101) was approved by the JHU Institutional Review Boards and those of each participating institution for collection and analysis of specimens and clinical data; the biorepository analytic plan was approved by the JHU Institutional Review Board (JHU IRB00135664).

| Specimen collection
From the parent study, a total of 5916 patients across the four sites met the CDG criteria and were tested for influenza by the GeneXpert assay, of which 1070 tested positive for influenza. A total of 999samples had adequate NP specimen available (>500 μl) allowing for further molecular detection using the ePlex RP RUO cartridges (Genmark Diagnostics, Carlsbad, CA) and were included for detailed analysis ( Figure 1).

| Outcomes and statistical analysis
Samples from patients who received rapid influenza testing were ana-  14,15 Notably, IAV/IBV co-infections with RhV/EV made up almost half of coinfections observed in our population. Of note, there were two patients that tested positive for both IAV and IBV. As these did not fit in either IAV only or IBV only groups and given the small number, they were included in the IAV/IBV + non RhV/EV co-infection group. The primary outcome of interest was hospital admission, which was used as a marker of disease severity. Chi-squared analysis was used to determine the rates and patterns of viral-co-infections. Univariate analysis was conducted to determine the odds of admission based on presence of viral co-infection; a multivariate logistic model was used to adjust for sex, age (categorical), and the presence or absence of certain underlying medical conditions (see Table 4), which would likely influence the clinical decision to admit (primary outcome).

| RESULTS
A total of 5916 patients were tested for influenza across the four EDs. The majority of patients had only one ED encounter (5649/5916, 95.5%); 267 (4.5%) had two or more encounters ( whereas IBV plus other non RhV/EV co-infection rates were much lower (4/999, 0.4%), and these included AdV, PIV and RSV. Of note, 6 specimens initially identified as IAV by rapid Xpert testing were reported as IBV by ePlex testing, and 4 specimens initially identified as IBV by Xpert testing were reported to be influenza A/H3 virus by ePlex. Two specimens noted to be IAV on Xpert testing were found to have IAV plus IBV co-infections by ePlex testing. The ePlex cartridges used in the study include bacteria; there was only 1 specimen positive for influenza and a bacterial co-infection. Of note, there were no significant differences in age and gender for those excluded    Associations were further quantified using univariate and multivariable analysis ( there was an increased risk of hospital admission, consistent with our findings here, although that was not specific to an ED population. 16 As noted, although we did not see worsening infection in our cohort study with IAV/IBV + RhV/EV co-infection, we did not appreciate any protective impact in this study either. Multiplex upper respiratory viral testing has been described in the literature as a potential approach to aid in antibiotic stewardship. [17][18][19] In a pediatric randomized control trial led by Rao  This resulted in a projected cost savings of >$578,000 due to deferred admissions and reduction in antiviral prescribing at the study site. 21 While the impact on admission and prescribing is well documented, there were 12 cases where the influenza type results from GenXpert and ePlex platform did not match. While that number was relatively small (<2% of cases overall) and might at least in part be explained by co-infections, it is possible that one or both assays had false positive or false negative result. Finally, we used hospital admission as a proxy for severity of illness. However, a clinician's decision to admit may not be strictly influenced by disease severity and there are other factors not captured here (e.g. psychosocial) that may affect the admission decision, independent of the influenza infection itself. In our regression analysis, however, we control for a variety of potential confounders including demographic, multiple comorbidities, and hospital site. Future studies could prospectively assess clinician's reason for admission and/or a numerical scoring system for severity of illness (such as MEWS score) at the time of admission, to provide more unbiased measure of severity of illness. 25

| CONCLUSION
In conclusion, we found that the presence of viral co-infection (other than RhV/EV) in ED influenza A/B positive patients was independently associated with increased risk of hospital admission. Our data suggest that multiplex testing may be valuable in practice when used to test higher risk populations or incorporated into a CDG for those patients found to have influenza A/B viral infection and could aid clinicians in predicting patient trajectory and helping with triage decisions for discharge or admission. The routine use of multiplex testing in the ED as a predictor for patient outcomes still requires more research.