Predictors for fatal human infections with avian H7N9 influenza, evidence from four epidemic waves in Jiangsu Province, Eastern China, 2013‐2016

Background Four epidemic waves of human infection with H7N9 have been recorded in China up to 1 June 2016, including in Jiangsu Province. However, few studies have investigated the differences in patients' characteristics among the four epidemic waves, and the analyses of factors associated with fatal infection lacked statistical power in previous studies due to limited sample size. Methods All laboratory‐confirmed A(H7N9) patients in Jiangsu province were analysed. Patients' characteristics were compared across four waves and between survivors and those who died. Multivariate analyses were used to identify independent predictors of death. Results Significant differences were found in the lengths of several time intervals (from onset of disease to laboratory confirmation, to onset of ARDS and respiratory failure, and to death) and in the development of heart failure. The proportions of overweight patients and rural patients increased significantly across the four waves. Administration of glucocorticoids and double‐dose neuraminidase inhibitors became the norm. Predictors of death included complications such as ARDS, heart failure and septic shock, administration of glucocorticoids, and disease duration. Conclusion Characteristics of H7N9 patients and clinical treatment options changed over time. Particular complications and the use of particular treatment, along with disease duration, could help clinicians predict the outcome of H7N9 infections.

Provinces/Municipalities. The virus has raised severe concerns due to its ability of binding to receptors in the upper respiratory tract 4 and the high death rate in human infections. Several published studies have presented preliminary analyses of the risk factors or predictors for fatal human infections with H7N9. [5][6][7] However, statistical analyses, especially multivariate analyses, were underpowered in most studies due to small sample size (≤40 cases) or patients' limited epidemiological/clinical information. As of 1 June 2016, the H7N9 virus has caused four epidemic waves in Jiangsu Province, resulting in a total of 103 laboratory-confirmed H7N9 cases with 47 deaths. This study aims to comprehensively identify the epidemiological or clinical predictors of death in human infections with H7N9 using multivariate analyses, based on a relatively large sample size obtained from Jiangsu province from 2013 to 2016.

| Subjects
All laboratory-confirmed human infections with H7N9 are reported through a national system for reporting of notifiable infectious diseases. 8 Demographic, epidemiological and clinical information on patients infected with H7N9 was collected using standardized questionnaires by local CDC staff or trained clinical doctors in Jiangsu Province and was reported through this system. All patients infected with H7N9 as of 1 June 2016 in Jiangsu Province were included in this analysis.

| Ethic statement
The National Health and Family Planning Commission ruled that the collection of data from cases of H7N9 was part of the public health investigation of an emerging outbreak, and thus, the investigation was exempt from institutional review board assessment. 8 Patients' information was collected and reported to Jiangsu Provincial CDC and China CDC through a national system for reporting of notifiable infectious diseases. Jiangsu Provincial CDC is responsible for checking and monitoring the reported information and will take part in patients' investigations if necessary. The data set was not anonymized in the reporting system but was anonymized before data analysis.  Table S1. Cases with missing data on variables being analysed were excluded. .736

| RESULTS
Chronic liver disease, n (%) .028 Renal dysfunction, n (%) .012 Antibiotics, n (%)  patients. A similar increasing trend was also observed in Zhejiang province, which was another hot spot of H7N9 epidemic in China. 12 The time interval from onset of disease to laboratory confirmation of H7N9 patients has been shortened from a median of 12 days to 8 days across four epidemic waves. This might be attributed to improved laboratory capacity of municipal Centers for Disease Control and Prevention (CDC), 13 avoiding repeated tests for confirmation by Provincial CDC. 3 In addition, increased awareness of clinical doctors and the public might also be involved. 13  We found in this study that clinical complications such as ARDS, heart failure and septic shock could pose an elevated risk of death from H7N9 infection. This finding is consistent with previous studies and case reports. 5-7 A longer disease duration was found to be associated with a reduced risk of death in our study. Previous studies also reported that the disease duration of fatal cases was shorter than that of patients who recovered. 22,23 Infections with several avian influenza subtypes, such as H9N2, H7N7, H7N2, H7N3, H10N7 and H6N1, are commonly regarded as self-limiting diseases, 24 and essential supportive care has been proved to be effective for recovery. 25,26 For H7N9, studies showed that viral load peaked within the first 10 days after symptom onset during the acute exudative phase, 27 and individuals'

| DISCUSSION
anti-H7N9 antibodies could be detected from day 21, along with a significant decline of viral load. 28 This indicated that infection with H7N9 can also be considered self-limiting, which highlighted the critical role of supportive health care in treatment. We suggest that a disease duration of 25.5 days might serve as the time indicator for patients' survival.
Interestingly, administration of glucocorticoids was associated with a significantly elevated risk of death, while its clinical implementation jumped from a proportion of 61.5% to 100% across the observed four epidemic waves in this study. Glucocorticoid is mainly used as an anti-inflammatory agent and is also an alternative for the treatment of septic shock. [29][30][31] The administration of glucocorticoids thus reflects disease severity to some extent. However, as observed in this study, glucocorticoids have been used more and more comprehensively for treating H7N9 patient, not only restricted to these with special need. Treating influenza with glucocorticoids remains controversial. A number of studies have reported that use of glucocorticoids increased the death risk from H1N1 32-35 and from H5N1. 36 Slower viral clearance of influenza A (H3N2) virus was observed in patients treated with systemic glucocorticoid. 37 Furthermore, a systematic review and meta-analysis using data from 19 studies of glucocorticoid treatment and human infection of influenza virus have concluded that glucocorticoids were related with mortality, nosocomial infection, longer mechanical ventilation and longer ICU stay. 38 As a result, a much more rational use of glucocorticoids based on sturdy scientific evidences is urgently needed, 39 for instance, assessing patients' disease progression with objective medical indicators along with the types, doses and administrative routes of glucocorticoids.
In conclusion, the differences in epidemiological and clinical char- were also provided.