Study patients and data collection
The Taiwan Severe Influenza Research Consortium (TSIRC) which includes eight tertiary referral centers in Taiwan , conducted this retrospective study. From 1 January to 31 March in 2016, all of the patients who were admitted to ICUs with virology-proven influenza were screened, and all of these patients who had the diagnosis of ARDS according to the Berlin definition  and also receiving prone positioning were enrolled. The demographic data, laboratory data, treatment record, ventilator setting data and outcomes were obtained from electronic medical records, while a standard case report form was used to collect these data. The local Institutional Review Board for Human Research of the involved hospitals approved this study (Linkou and Kaohsiung Chang-Gung Memorial Hospital IRB No. 201600632B0, National Taiwan University Hospital 201605036RIND, Taipei Veterans General Hospital 2016-05-020CC, Tri-Service General Hospital 1-105-05-086, Taichung Veterans General Hospital CE16093A, China Medical University Hospital 105-REC2-053(FR), Kaohsiung Medical University Hospital KUMHIRB-E(I)-20170097) and the need for informed consent was waived due to the retrospective nature of the study.
Diagnosis of influenza
Influenza was proven by more than 1 of the following tests: influenza rapid antigen test, nucleic acid reverse-transcriptase polymerase chain reaction (RT-PCR), or viral culture, or serum antibody serologic test (antibody titers changed more than 4 times from acute to convalescent stage) according to the different patient or different hospital. The rapid antigen test was sampled from nasopharynx swab or throat swab, and the RT-PCR or viral culture might sampled from nasopharynx swab, throat swab, sputum or bronchoalveolar lavage
We collected the laboratory data when the patient was admitted to the ICU including complete blood count, white blood cell differential count, and biochemistry data. Arterial blood gas data before and 1 day after the prone positioning were collected. We also collected the severity scores on the ICU admission day including Pneumonia Severity Index (PSI) , Acute Physiology and Chronic Health Evaluation II (APACHE II) , CURB-65 pneumonia severity score , and Sequential Organ Failure Assessment (SOFA) score .
Mechanical ventilator setting and prone positioning
The ventilator settings were adjusted according to the ARDSnet with lung protective strategy protocol , and the setting data such as peak inspiratory pressure, positive end expiratory pressure (PEEP), fraction of inspired oxygen (FiO2), and tidal volume before and 1 day after the prone positioning were all recorded. The pressure control mode was used in all of the patients. The dynamic driving pressure was defined as peak inspiratory pressure minus PEEP, and the compliance was defined as tidal volume divided by dynamic driving pressure. The methods of prone positioning were according to the PROSEVA study  that more than 16 hours/day was used in all of the patients. According to the previous studies, the PaO2 responders were defined as the post-prone positioning PaO2/FiO2 ratio increasing by ≥20% or by ≥20 mm Hg compared to pre-prone positioning, while PaCO2 responders were defined as post-prone positioning PaCO2 decreasing by ≥1 mm Hg compared to pre-prone positioning [12, 13, 24-26].
Number (percentages) for nominal variables, and means ± standard deviation for continuous variables were presented in this study. Pearson’s Chi Square test was used to compare nominal variables, and the independent Student t-test was used to compare continuous variables, with the binary logistic regression test used to analyze univariate and multivariate factors. In this study, the two-tailed test was used, and the definition of significance was p value <0.05. Statistical analyses and database management were performed using SPSS version 22.0.0 (SPSS Inc., Chicago, IL).