Influence of time to diagnosis of severe influenza on antibiotic use, length of stay, isolation precautions, and mortality: a retrospective study

Abstract Background Timely diagnosis of influenza infection in patients might help reduce antibiotic use during influenza seasons and, consequently, antibiotic selection pressure. In this retrospective cohort study, we aimed to evaluate whether time to influenza diagnosis in patients with severe influenza is associated with the duration of antibiotic therapy. Methods We retrospectively included all hospitalized patients >16 years who tested positive for influenza A or B by polymerase chain reaction during influenza seasons 2013/2014 or 2014/2015 at the University Hospital Zurich. The primary aim was to assess the association between timing of laboratory‐confirmed influenza diagnosis and duration of antibiotic therapy. Secondary outcomes were length of hospital stay, duration of isolation precautions, and mortality. Early diagnosis was defined as laboratory confirmation on the day of or the day after hospital admission or symptom onset. Results A total of 126 patients were included (median age 57 years). Timing of influenza diagnosis was not associated with the duration of antibiotic treatment, the duration of isolation precautions, or mortality. Early influenza was associated with reduced length of hospital stay (median 7 vs 9 days [P=.014]) in patients with community‐acquired influenza. Conclusions Although the duration of antibiotic therapy and mortality were found unaffected by early influenza diagnosis, our data indicate that it is linked with a reduction in the length of hospitalization in patients with community‐acquired influenza. This highlights a need to also fully understand the effect of time to diagnosis of bacterial pathogens on antibiotic prescribing patterns in order to exploit the potential of early influenza diagnosis in patient care.


| INTRODUCTION
Unspecific symptoms hamper influenza diagnosis based on clinical case presentation alone and differential diagnoses can be quite diverse, misinterpreting influenza infections as common cold, bacterial pneumonia or even non-infectious causes of fever. 1,2 Although the likelihood of influenza is increased during winter season in patients with cough and fever, 1,3,4 these predictors do not enable physicians to differentiate adequately between the various differential diagnoses.
Due to these unspecific and often misleading symptoms and potentially severe complications, hospitalized patients with suspected or confirmed influenza diagnosis often receive antibiotic treatment. 5,6 Recommendations are to base influenza treatment on neuraminidase inhibitors together with symptomatic therapy and to use antibiotics only in severe cases or patients with high risks. [7][8][9] In severe cases, an antibiotic agent active against Staphylococcus aureus is recommended, as staphylococcal pneumonia is known to be frequently associated with influenza. 10 Interestingly, even though early influenza treatment with neuraminidase inhibitors was previously shown to be associated with shorter duration and reduced severity of illness, a faster resolution of fever, and a faster return to normal health and activity, 11 13 This management strategy, however, might also be influenced by study results that show only moderate benefit of neuraminidase inhibitors in reducing influenza symptoms in healthy outpatients and the lack of controlled clinical trials supporting a beneficial effect on severe outcomes in hospitalized patients despite a growing body of evidence from observational studies. [14][15][16][17] A recent survey highlights that up to date, antiviral substances are underutilized and antibiotics overused in influenza patients. 5 Despite the fact that in hospitalized adults with acute respiratory illness during winter season, viral infection seems to be more likely than bacterial, 6 patients often receive antibiotic treatment despite a diagnosis of respiratory virus infection. 18 More rapid influenza diagnosis may thus help streamline antiinfective treatment and reduce antibiotic consumption. As antibiotic overuse is associated with emergence of antibiotic resistance, 19 rapid influenza diagnosis may therefore reduce antibiotic selection pressure, resistance development, and support appropriate treatment.
Consequently, rapid diagnosis of influenza has previously been shown to help reduce the use of antibiotics in adult outpatients 20 and was shown to reduce the duration of antibiotic therapy, the amount of prescription, and the length of hospital stay in children. 21,22 Recent studies suggested a general benefit of hospitalized patients from early influenza diagnosis by reducing the need for extended care after hospital discharge in older adults, 23 mortality, 24 and the risk of lower respiratory tract complications. 25 Oosterheert et al. and Shiley et al. reported that viral respiratory tract diagnosis per se does not impact the use and duration of antibiotic treatment in hospitalized adults. 18,26 In contrast, Rogers et al. showed that rapid influenza diagnosis did decrease duration of antibiotic use, length of hospitalization, and duration of isolation in children admitted with respiratory symptoms. 21 To clarify and validate the benefits of early diagnosis, we aimed to determine whether rapid laboratory confirmation of influenza infection in hospitalized adults reduced the duration of antibiotic therapy in a retrospective cohort study of adults with influenza diagnosis hospitalized at the University Hospital Zurich, Switzerland, in seasons 2013/2014 and 2014/2015. Secondary aims of our study were the analysis of the association between timeliness of influenza diagnosis, length of hospital stay, duration of isolation precautions, and in-hospital mortality.

| Study setting, design, and procedures
We performed a single-center retrospective cohort study of hospitalized patients with laboratory-confirmed influenza diagnosis at the University  [27][28][29] Our study retrospectively included all hospitalized and consenting patients greater than 16 years of age with influenza A or B infection confirmed by polymerase chain reaction (PCR) 30 32 and the CURB-65 Score 31 were used to classify the severity of respiratory disease. "Generalized weakness" was defined as feeling weaker than before onset of symptoms.

| Predictor variable
The main predictor variable was the duration between symptom onset or hospital admission (whichever occurred first) and laboratory confirmation of influenza. Early influenza diagnosis was defined as laboratory confirmation on the day of or the day after hospital admission in patients with community-acquired influenza infection and the day of or the day after symptom onset in patients with nosocomial influenza infection, respectively. Nosocomial influenza was defined as occurrence of the first symptoms after admission to the hospital.

| Outcome variables
Duration of antibiotic therapy was the primary outcome variable of interest. To estimate the duration of antibiotic therapy, all antimicrobials with antibacterial activity were included if treatment was initiated after the first symptom and stopped after laboratory confirmation of influenza infection. The variable beta-lactam antibiotics was divided into a broad-spectrum group with (including piperacillin, cefepime, ceftazidime, meropenem, and imipenem) and a narrowspectrum group without (including penicillin, amoxicillin, amoxicillin/ clavulanic acid, cefuroxime, cefotaxime, ceftriaxone, ertapenem) antipseudomonal activity.
Secondary outcomes were length of hospital stay, duration of isolation precautions and in-hospital mortality. Length of stay was defined as time between admission and discharge. Associations with length of hospital stay were analyzed for patients with communityacquired influenza infection only.

| Statistical analysis
Categorical data were tested for differences using Fisher exact tests, whereas continuous variables were tested using Wilcoxon rank sum tests or the Student's t test, as appropriate. Cox regression analysis was used to detect differences in length of antibiotic therapy and length of hospital stay in patients with early as compared to late influenza diagnosis. Multivariable logistic regression analysis was used to determine differences in mortality. Potential confounders among patient characteristics with P-values <.05 in univariable analyses were considered for inclusion in multivariable models based on clinical judgment, with final models representing those that best balanced parsimony and fit. The limited number of outcomes was factored in when building the models to prevent overfitting. 33 Data were analyzed using Stata ® version 13.1 (Stata Corporation, College Station, TX, USA).
Two-tailed P-values <.05 were considered statistically significant.

| Ethics
The research was conducted in accordance with the Declaration of Helsinki and national and institutional standards. The study was reviewed and approved by the ethics committee of the Canton of Zurich (BASEC-Nr.: PB_2016-00182). A general consent was given by the ethics committee for all patients hospitalized before January 1, 2014, and for deceased patients, tourists, or patients who could not be contacted by mail or phone. Informed consent was obtained from all patients that were hospitalized with influenza at or after January 1, 2014.

| RESULTS
In total, 126 consenting patients with positive influenza PCR could be included in the study cohort. Fifty-three (42%) were diagnosed with influenza on the day of or the day after admission or symptom onset (early diagnosis group), respectively. Patient characteristics and outcomes are shown in Tables 1 and 2 and S1. According to univariable analysis, patients with delayed influenza diagnosis had lower total neutrophil counts (P=.006), lower total leukocyte counts (P=.008), lower diastolic blood pressure on admission/symptom onset (P=.045), and were more likely to have acquired influenza infection nosocomially (P=.032) than patients with early influenza diagnosis.

Late influenza diagnosis (n=73) P-value
Pneumonia severity index on admission or first symptom, median (range) 89 (24-222) 88  Our study, however, in which antiviral use was low, is not able to confirm or rebut any of these findings.
Our study failed to demonstrate an effect of early diagnosis on the duration of isolation precautions. Even though our study population is not suitable for determining the effect of early diagnosis on days of isolation in patients with suspected-rather than confirmed-influenza diagnosis, such an effect may be small in our setting as preemptive isolation precautions were only taken in a small fraction of patients with influenza diagnosis.
Our study has several limitations. First, the findings of our study may not be generalizable to other settings. The University Hospital Zurich is a tertiary care referral center where, because of its many highly specialized departments, many patients with severe comorbidities, such as lung and heart transplant recipients, or oncological patients receive treatment. The patients included in this study may thus not represent a cohort of hospitalized influenza patients as seen in other hospitals. Nevertheless, our patient population represents those at highest risk for severe influenza complications and therefore is of particular interest. Our findings may also not be translated to other influenza seasons or other regions in the same influenza seasons due to antigenic drift and changes in predominant influenza strains over time and even between geographically distinct settings. Second, our observations are only applicable to hospitalized adults with severe influenza; they cannot be extrapolated to adult outpatients with mild influenza or children. Early diagnosis might help optimize treatment and influence antibiotic selection pressure in these populations, as shown by others. [20][21][22][23][24][25]38 Last, the interpretation of our findings is limited by the retrospective design of the study. Our results should be confirmed in the setting of a prospective study to eliminate bias.
In conclusion, we were able to demonstrate that diagnosis on the day of or the day after admission in hospitalized patients with community-acquired influenza infection is associated with shorter hospitalization. In contrast, timely diagnosis did not influence the du-