Etiologies of severe acute respiratory infection (SARI) and misdiagnosis of influenza in Indonesia, 2013‐2016

Abstract Background Severe acute respiratory infection (SARI) accounts for a large burden of illness in Indonesia. However, epidemiology of SARI in tertiary hospitals in Indonesia is unknown. This study sought to assess the burden, clinical characteristics, and etiologies of SARI and concordance of clinical diagnosis with confirmed etiology. Methods Data and samples were collected from subjects presenting with SARI as part of the acute febrile Illness requiring hospitalization study (AFIRE). In tertiary hospitals, clinical diagnosis was ascertained from chart review. Samples were analyzed to determine the “true” etiology of SARI at hospitals and Indonesia Research Partnership on Infectious Diseases (INA‐RESPOND) laboratory. Distribution and characteristics of SARI by true etiology and accuracy of clinical diagnosis were assessed. Results Four hundred and twenty of 1464 AFIRE subjects presented with SARI; etiology was identified in 242 (57.6%), including 121 (28.8%) viruses and bacteria associated with systemic infections, 70 (16.7%) respiratory bacteria and viruses other than influenza virus, and 51 (12.1%) influenza virus cases. None of these influenza patients were accurately diagnosed as having influenza during hospitalization. Conclusions Influenza was misdiagnosed among all patients presenting with SARI to Indonesian tertiary hospitals in the AFIRE study. Diagnostic approaches and empiric management should be guided by known epidemiology. Public health strategies to address the high burden of influenza should include broad implementation of SARI screening, vaccination programs, clinician education and awareness campaigns, improved diagnostic capacity, and support for effective point‐of‐care tests.


| INTRODUC TI ON
Severe acute respiratory infection (SARI) was defined in 2011 for purposes of global surveillance. The 2011 definition harmonized heterogeneous definitions used by three WHO regions, thus facilitating comparisons. The revisions include one definition for all age groups to simplify implementation, dropping "shortness of breath" and "breathing difficulty," adding "history of fever" and increasing the onset of symptoms to 10 days. SARI is now defined as an acute respiratory illness with a history of fever or measured fever of ≥38°C and cough, with onset within the past 10 days and requiring hospitalization. 1 This case definition enables monitoring of severe influenza-related diseases and assessment of burden.
Severe acute respiratory infection criteria are not specific for influenza. A number of respiratory viruses other than influenza and bacterial etiologies in patients that met SARI criteria have been reported from previous studies. [2][3][4] Several viruses and bacteria that cause systemic diseases such as dengue virus, 5,6 chikungunya virus, 7,8 Salmonella spp., 9 Leptospira spp., 10 and Rickettsia typhi 11 may present with a respiratory illness and fulfill the SARI criteria. A few studies have evaluated the sensitivity and specificity of SARI criteria for influenza detection. Sensitivity and specificity range from 37% to 84% and 23% to 78%, respectively. [12][13][14] SARI criteria may also be used for pneumonia surveillance. 13 Severe acute respiratory infection and influenza-like illness (ILI) surveillance have been conducted in Indonesia since 1999 at several hospitals and primary health centers. These studies reported that the proportion of influenza cases varied from 14% to 20% of all enrolled subjects. [15][16][17][18] As the aims of these studies were to confirm influenza infections, other causes of respiratory infections or "systemic" viral or bacterial infections were not analyzed.
Since bacterial, influenza, and non-influenza viral respiratory infections are prevalent in Indonesia, there is a need to evaluate strategies for respiratory pathogen surveillance in the region.
Our study aimed to identify pathogens associated with SARI in Indonesia, to compare these with hospital assessed etiology, and to describe their demographic and clinical characteristics. In addition, we determined performance characteristics of SARI criteria for identification of influenza virus.

Funding Information
This project has been funded in whole or in part with Federal funds from the NIAID, NIH, under contract Nos. HHSN261200800001E and HHSN261201500003I. NIAID collaborators contributed to design of the study; collection, analysis, and interpretation of data; and writing of the manuscript. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.

K E Y W O R D S
diagnostic accuracy, etiology, Indonesia, influenza, severe acute respiratory infection 2 | ME THODS

| Participants
Patients fulfilling the criteria for SARI 1 were identified from an observational cohort study of patients hospitalized with acute febrile illness requiring hospitalization (AFIRE) conducted in Indonesia from 2013 to 2016. 19 The AFIRE study recruited patients presenting to 8 tertiary hospitals for evaluation of acute fever, at least 1 year old, hospitalized within the past 24 hours, and not hospitalized within the past 3 months. These hospitals were top referral provincial hospitals representing 7 major cities (Bandung, Denpasar, Jakarta, Makassar, Semarang, Surabaya, and Yogyakarta) in three populous islands. Patients were eligible for participation if they or a legal guardian/representative provided written consent following an explanation of the study objectives and procedures in Indonesian. Assent was obtained from children ≥13 years old or who were old enough to understand the proposed research.
Informed consent or assent was obtained prior to collection of clinical data, laboratory data, and specimens. Blood was collected from all subjects at enrollment, 14-28 days, and 3 months after enrollment. Other biological specimens such as nasopharyngeal swabs, sputum, urine, or feces were collected per the attending physician. Records were reviewed for demographic and clinical information, including clinical diagnoses made during hospitalization. The "true" diagnosis was determined based on hospital and/ or Indonesia Research Partnership on Infectious Disease (INA-RESPOND) laboratory testing in conjunction with available clinical information.
Severe acute respiratory infection was assessed per the WHO definition requiring (a) an acute respiratory illness (ARI), (b) history of fever or measured fever of ≥38°C, (c) cough, (d) onset within the past 10 days, and (e) requiring hospitalization. 20 As no standard definition of ARI was available, we defined ARI as any sign or symptom related to the respiratory tract, including coryza, nasal congestion, sore throat, hemoptysis, and dyspnea. Presence of cough was determined by chart review. For patients who had pneumonia, it was assumed that cough was present even if not recorded. The requirements of fever, onset <10 days, and requiring hospitalization were already met by participants through AFIRE inclusion criteria. Ethical clearance was obtained from F I G U R E 1 Study flow, specimen tested, and pathogen by diagnostic method. Urine and CSF are not included because no pathogens were identified from urine samples (n = 23) and CSF specimens were unavailable. The number of cases is shown in parenthesis when more than 1 case was identified the institutional review board of the National Institute of Health

| Diagnostic testing
Diagnostic microbiology tests were conducted at hospitals immediately after specimens were collected and at the INA-RESPOND laboratory retrospectively. The diagnostic laboratory evaluations and identified pathogens are shown in Figure 1. Details of the laboratory assays listed below and interpretation for each pathogen are listed in Table S1.

| Hospital laboratory
At each hospital, all blood specimens, and other biological specimens, when available were cultured. Bacterial culture and identification were performed via VITEK® 2 (bioMérieux) and BD Phoenix™ (Becton-Dickinson) automated systems, according to manufacturer instructions. Based on the standard of care, the two most common "systemic" pathogens, dengue virus and Salmonella typhi, were tested using rapid NS1 antigen and IgM/IgG antibody test; IgM Salmonella typhi was evaluated by Tubex TF. In a subset of cases, microscopic examination was performed on sputum or feces.

| "Systemic infection" pathogens
We used the term "systemic infections" to cover infections by several pathogens that are usually circulating in the blood and not commonly found in respiratory specimens, including dengue virus, Salmonella spp., Rickettsia typhi, Leptospira spp., and chikungunya virus. Given that dengue is endemic in Indonesia, all blood specimens, regardless of the clinical presentation, were tested for dengue virus by real-time polymerase chain reaction (rRT-PCR), NS1 antigen ELISA, dengue IgM, and IgG ELISA assays. As other "systemic infections" are prevalent in Indonesia, they were tested when dengue testing was negative, using molecular and serological assays.

| Real-time polymerase chain reaction for respiratory pathogens
Bacterial DNA was extracted from sputum or viral transport media (VTM) containing a respiratory swab using the QIAamp Bacterial DNA Mini Kit (Qiagen, Hilden) according to the manufacturer's protocol. Bacterial DNA was eluted in 100 µL of AE buffer and used as template for real-time PCR assay or stored at −80°C. RNA was extracted from VTM containing a respiratory swab or sputum using the QIAamp Viral RNA Mini Kit (Qiagen, Hilden) according to the manufacturer's protocol. Viral RNA was eluted in 60 µL of AVE buffer and stored at −80°C.

| Serology/ ELISA
Serologic testing was performed for influenza A, influenza B, respiratory syncytial virus (RSV), measles, and rubella. Influenza A (Cat#

ESR1231M) and B (Cat# ESR1232M), virus IgM kits and influenza A (Cat# ESR1231G) and B (Cat# ESR1232G), and virus IgG kits (SERION
ELISA classic kit Institut Virion/Serion GMBH-Germany) were used according to manufacturer instructions. ELISA kits (Serion) were used to test IgM and IgG antibodies of RSV, measles, and rubella viruses. Since we used a semi-quantitative ELISA method, we considered influenza/RSV/measles and rubella infection present when sero-conversion or at least twofold increase of IgM and/or IgG titers was observed, consistent with manufacturer specifications and "Quality Standards in Microbiological/Infectiological Diagnostics." 21

| Data analysis
Patient characteristics, clinical, hematology and chemistry data, and outcomes were collected via paper case report form, entered into the Open Clinica database, and tabulated according to SARI etiology.
Pathogens were evaluated according to clinical diagnoses of respiratory and non-respiratory conditions at discharge. Group comparisons were performed using chi-square. Comparisons between continuous variables were performed by t test using STATA 17. Logistic regression was used to explore univariate relationships between the vari- relative to the number of variables, we fit a multivariate logistic regression with LASSO penalty as an exploratory variable selection exercise. Statistical analyses were performed using SPSS version 22 (IBM Corporation) and R version 3.6.0.; the level of significance was set at P < .05. SARI criteria for diagnosing influenza were evaluated using randomly selected SARI and non-SARI cases. Influenza PCR and/or serology assays were considered the gold standard.

| RE SULTS
Of 1464 subjects enrolled in the parent AFIRE study, 420 met the SARI criteria. Eighty-nine had blood and respiratory specimens, and 331 had only blood specimens. The first screening for dengue and 4 other "systemic infections" (Salmonella spp., Rickettsia spp., chikungunya virus, and Leptospira spp.) contributed to 26% (109/420) of the total SARI cases. Influenza diagnostic tests were performed in all 420 SARI cases. Influenza was identified in 12.1% (51/420) cases, consisting of influenza A untypeable (32), influenza B (13), influenza A H3N2 (5), and influenza A H1N1 pdm (1). A total 43 out of 51 (83.7%) influenza cases were confirmed by serology only, including two cases that also had evidence of chikungunya virus and Rickettsia typhi infections, and 8 cases by RT-PCR and serology including one mixed infection with Moraxella catarrhalis. Among 200 randomly selected SARI cases, 24 influenza cases were identified and among 200 randomly selected non-SARI subjects, influenza was confirmed in 10 subjects, suggesting the sensitivity of SARI criteria to identify influenza cases was 70.6% (24/34), but the specificity was only 50.5% (190/366).
Other viral and bacterial respiratory pathogens contributed to 16.7% (70/420) of SARI cases, whereas non-respiratory pathogens other than the five pathogens denoted above contributed to 3% (13/420) of cases. The most common respiratory viruses after influenza virus were RSV and measles (2.6% (11/420) each), whereas the most common bacteria were Mycobacterium tuberculosis (2.9% (12/420)), Streptococcus pneumoniae, Pseudomonas aeruginosa, and Klebsiella pneumoniae(1.4% (6/420) each). Pathogens and the specimen type used for diagnostic tests are listed in Figure 1. A total 178 out of 420 (42.4%) cases had no pathogen identified. In addition, identified pathogens in SARI and non-SARI groups are listed in Table S2.

| Identified pathogens based on clinical diagnoses
Among the 420 subjects who met SARI criteria, 244 (58.1%) had a hospital discharge diagnosis of respiratory disease and 176 (41.9%) of non-respiratory disease. No subject had a discharge diagnosis of influenza. The three most common discharge diagnoses among patients considered to have respiratory diseases were pneumonia (124 (50.8%)), upper respiratory tract infection (68 (28%)), and pulmonary TB (24 (9.8%)). Dengue and typhoid fever were equally the most frequent discharge diagnosis among SARI patients with non-respiratory diseases (55 (31.3%)) ( Figure 2).
d The chi-square or ANOVA test were used to assess differences in frequency/mean amongst etiologies. Cells in bold show the highest or lowest percentage/value of parameters in each category when the mean difference was significant at the .05 level compared with two or more other categories by pairwise comparisons post-hoc tests. Other Viruses (Chikungunya virus (5), HIV (1)), Amoeba (4), and co-infection cases (3) were not included in the analysis since N < 10 in those each group.  (1) 12 (1) Leptospira spp.  (3) 38 (2) 50 (3) 17 (4) 19 (5) 178 (17 shortness of breath and leukocytosis in respiratory bacteria; and granulocytosis in non-respiratory bacteria. Details of the proportion of these parameters in each pathogen group are listed in Table 1. and logistic regression in Table S3.

| Comorbidities and outcomes
The proportion of SARI in hospitalized patients with fever was higher in ≤5 years old and ≥60 years old (52.7% and 48.6%) than in the 5 to ≤18 years old and 18 to <60 years old groups (25.5% and 22.3%, respectively) ( Three influenza cases were also fatal. The first was a 49-year-old woman with rheumatic heart disease, the second was a 49-year-old woman with bronchiectasis, and the third was 59-year-old woman with mediastinal tumor and heart disease. Among the SARI subjects, 173 (41.2%) had one or more comorbidities. These were more common in subjects diagnosed with respiratory diseases (132/244 (54.1%)) compared to subjects diagnosed as having non-respiratory diseases (39/176 (22.1%)). 22.5%-24.7% for specificity). In contrast, a study from Northern India reported a sensitivity and specificity of 37% and 78%, respectively. 14 These disparate results may be due to use of the previous SARI criteria by the latter study. The low specificity of SARI criteria seems related to its broad clinical criteria, resulting in 176 (41.9%) cases of non-respiratory illness meeting the definition of SARI.

| D ISCUSS I ON
"Systemic" viral and bacterial infections may manifest clinically with respiratory signs and/or may progress to complicated pneumonia. Dengue with respiratory manifestations has been reported in more than 60% of the schoolchildren in Colombia, 26 in 4 Taiwanese patients, and 1 returned traveler in the United States, whose nasopharyngeal swabs were also positive by RT-PCR. 5,6 Dengue was also the third leading cause of acute respiratory distress syndrome in pediatric patients. 27 Similarly, leptospirosis may present with severe pulmonary manifestations 10,28 and the bacteria can be detected from throat swabs. 10 Other "systemic" diseases such as chikungunya, rickettsiosis, and typhoid fever may also present with respiratory manifestations. 8,9,11 Although SARI patients with discharge diagnoses of such "systemic" diseases constituted a large proportion of cases, missed diagnoses only occurred in 25% (27/109)  Rapid influenza testing has low sensitivity (50%-70%), while molecular diagnosis requires trained technicians and specialized instruments. 31 We herein confirmed that serologic testing as an adjunct to RT-PCR was helpful for identification of viral pathogens in communi- A major strength of this study was testing for a complete panel of respiratory viral and bacterial pathogens by culture, molecular assays, and serological tests on acute and convalescent plasma. In addition, this study is the first to report the epidemiology of non-respiratory pathogens associated with SARI and the first to evaluate SARI criteria for identifying influenza cases in Indonesia.
In conclusion, influenza is often overlooked as an etiology of febrile illness in Indonesia. Implementation of the SARI criteria in tertiary referral hospitals would help identify potential influenza infections. Public health strategies to address the high burden of influenza should include vaccination programs, clinician education and awareness campaigns, improved diagnostic capacity, and support for effective point-of-care tests.

ACK N OWLED G EM ENTS
We are grateful for the support of the INA-RESPOND network, especially the research assistants. We also thank the patients at each