Introduction
Influenza is a known risk factor for invasive pulmonary aspergillosis1, as are other viral respiratory infections including SARS-Cov-2 (COVID-19)2, parainfluenza and respiratory syncytial virus3. Since its first case description in 19524influenza-associated pulmonary aspergillosis (IAPA) is increasingly recognized as a severe complication in critically ill influenza patients5-8. IAPA incidence ranges from 10-32% of influenza patients admitted to ICU 1, 9,10. Differences in prevalence might be related to different awareness and screening practices11. Risk factors for developing IAPA include male sex, smoking, chronic lung disease, influenza A, solid organ transplant, haematologic malignancy and treatment with corticosteroids within 28 days prior to influenza infection but are otherwise poorly defined1,26. IAPA carries a high mortality of 30-60% 1, 5, 6,12, 13 and commonly results in need of organ supportive therapies14. Optimal diagnostic and preventive strategies are unclear15.
During the 2017/18 season we identified IAPA in 11% of critically ill influenza patients in two Swiss centers which was associated with high risk of complications and mortality14. In the current study we aimed to analyze the epidemiology and clinical outcome of IAPA in a multicenter study of Swiss ICUs combining our 2017/18 data with new data from 2019/20.

Study design and methods

We performed a multicenter cohort study in ICUs of seven tertiary care hospitals in Switzerland during the 2019/20 influenza season. Participating ICUs received an educational training on influenza and IAPA in fall 2019 and a screening algorithm for IAPA was implemented (figure 1) in order to increase awareness of this entity and homogenize diagnostics and treatment strategies. We proposed to screen all patients with influenza infection systematically for IAPA on admission to the ICU using fungal cultures from bronchoalveolar lavage (BAL) or tracheal secretion (TS), if BAL was not possible, and galactomannan (GM) from BAL or serum. A positive GM was defined as >0.5 in serum and >1.0 in BAL. Patients were assessed daily for new clinical signs of IAPA, which triggered microbiological sampling for aspergillosis from BAL (or TS) and serum. IAPA was diagnosed according to diagnostic criteria by Schauwvlieghe et al. and Blot et al1, 16. Anti-mold treatment was initiated as soon as criteria of IAPA were fulfilled or pre-emptively if patients were unstable and had suspicion of IAPA (figure 1). All data were retrospectively collected from electronic medical records. Our inclusion criteria and definitions were reported previously 14.
We combined data on influenza patients from seven ICUs during the 2019/20 influenza season with those from the 2017/18 influenza season of two tertiary care hospitals (University Hospital Geneva, Cantonal Hospital St. Gallen). The primary outcome was to find risk factors for IAPA and secondary outcomes were to define predictors for poor outcome, which was a composite of in-hospital mortality, ICU length of stay (LOS) ≥ 7d, mechanical ventilation ≥ 7d or extracorporeal membrane oxygenation (ECMO).
The study was approved by the local ethics committees (EKOS 2018-01994 and 2019-02173). Funding was provided by the research funding of the Cantonal Hospital of St. Gallen and the Swiss network on fungal diseases (FUNGINOS).
Statistical analysis
No missing data were observed in 158 data sets. Continuous variables were not categorized. Continuous variables were assessed by Wilcoxon ranksum test, categorical values by Fisher’s exact test. Multivariable logistic regression was performed to assess risk factors for IAPA and the composite poor outcome. To prevent multicollinearity, we first removed variables which were obviously (per definition, or clinically) related to the relevant outcomes (antifungal therapy for IAA, and intubation, renal replacement therapy, ARDS, bacterial superinfection and delirium for bad outcome). The last two variables e.g. are known to be associated with ICU-LOS.
Variables with very low numbers were also removed from the model (i.e. neutropenia, solid organ transplantation). Restricted cubic splines were used for modelling continuous variables. We performed a stepwise backward elimination procedure. Collinearity was tested with variance inflation factor and variables that were highly collinear were eliminated. Bootstrapping procedures were used to check validation and calibration of the model. Kaplan Meier curves were analyzed for assessment of duration of ICU stay. Univariate analysis of LOS-ICU was performed by the logrank test. All statistical analyses were performed by R 4.0.2 (2017, R foundation for statistical computing, Vienna, Austria).

Results

We included 158 influenza patients (81 [51%] from the 2017/18 and 77 [49%] from the 2019/20 influenza season). 17 patients (10.8%) were diagnosed with IAPA. We did not observe a different proportion of IAPA over the two influenza seasons (2017/18: 9 cases [11.1%], 2019/20: 8 cases [10.3%], p=1.0). Baseline characteristics were similar in patients with and without IAPA, except for a significantly higher prevalence of asthma among patients with IAPA (p=0.05, table 1). Patients with and without asthma received corticosteroids in 3/8 (37.5%) vs. 27/150 (18%, p=0.2) before influenza diagnosis and in 7/8 (87.5%) vs. 76/150 (50.7%, p=0.05) during hospitalization.