key: cord-0043298-yvek5vjz
authors: Althaus, T.; Thaipadungpanit, J.; Greer, R.C; Swe, M.M.M; Dittrich, S.; Peerawaranun, P.; Smit, P.W; Wangrangsimakul, T.; Blacksell, S.; Winchell, J.M.; Diaz, M.H.; Day, N.P.J; Smithuis, F.; Turner, P.; Lubell, Y.
title: Causes of fever in primary care in Southeast Asia and the performance of C-reactive protein in discriminating bacterial from viral pathogens
date: 2020-05-11
journal: Int J Infect Dis
DOI: 10.1016/j.ijid.2020.05.016
sha: e8ac4d7d375a25793807c5c4f600319100efb483
doc_id: 43298
cord_uid: yvek5vjz

OBJECTIVES: We investigated causes of fever in the primary levels of care in Southeast Asia, and evaluated whether C-reactive protein (CRP) could distinguish bacterial from viral pathogens. METHODS: Blood and nasopharyngeal swab specimens were taken from children and adults with fever (>37.5˚C) or history of fever (<14 days) in Thailand and Myanmar. RESULTS: Of 773 patients with at least one blood or nasopharyngeal swab specimen collected, 227 (29.4%) had a target organism detected. Influenza virus type A was detected in 85/227 cases (37.5%), followed by dengue virus (30 cases, 13.2%), respiratory syncytial virus (24 cases, 10.6%) and Leptospira spp. (9 cases, 4.0%). Clinical outcome was similar between patients with a bacterial or a viral organism, regardless of antibiotic prescription. CRP was higher among patients with a bacterial organism compared to those with a viral organism (median 18 mg/L, interquartile range [10-49] versus 10 mg/L [≤8-22], p-value 0.003), with an area under the curve of 0.65, 95% confidence interval (0.55-0.75). CONCLUSIONS: Serious bacterial infections requiring antibiotics are exceptions rather than the rule in the first lines of care. CRP-testing could assist in ruling out such cases in settings where diagnostic uncertainty is high and routine antibiotic prescription is common. The original CRP randomised-controlled trial (RCT) was registered with ClinicalTrials.gov, number NCT02758821.

Fever is a common reason for seeking healthcare in Southeast Asia and as malaria incidence declines, bacteria and viruses now represent the main contributors to acute febrile illness [1] [2] [3] [4] [5] . Identifying these pathogens is challenging, even in well-resourced laboratories with specialised staff, and most aetiological data for febrile illness originate in tertiary hospitals [6] . Hospitalised patients, however, are by definition more severely ill, often with important comorbidities, implying that findings may not be applicable to febrile patients attending primary levels of care.

Primary care in low-middle income countries (LMICs) is typically characterised by a shortage in human resources, diagnostics and evidence-based guidelines [7] . Studies investigating causes of fever in this environment are few and frequently of poor quality: enrolment is often limited to a single clinical presentation and specific age category, and microbiological investigations rarely use goldstandard methods [8] [9] [10] . Additionally, most primary care patients attend early after symptom onset with non-severe presentations, lowering the chances of detecting a pathogen [8] [9] [10] [11] [12] [13] [14] [15] [16] . Empiric treatment guidelines are therefore based on limited epidemiological evidence and are often implemented by insufficient and poorly trained staff, contributing to irrational antibiotic prescription practices [17] [18] [19] .

High prescription levels are partly driven by frequent clinical overlap between bacterial and viral infections, challenging the identification of patients who might benefit from antibiotics [2, 3, 20] .

Given these limitations in clinical judgment and laboratory structures, point-of-care testing (POCT) to guide fever management could be beneficial in primary care settings [21] . Pathogen-specific tests represent one such option but several barriers undermine their potential use: these only exist for a small number of pathogens, with inconsistent performance, and most are antibody detection-based that might preclude the distinction between active infection and past-exposure [22, 23] . A few antigen detection-based POCTs exist but their integration in low-level care is unrealistic: a Salmonella Typhi rapid test requires laboratory infrastructure with poor detection in blood even at high concentrations and results are not available before 24-48 hours [24, 25] ; test sensitivities for influenza virus A, respiratory syncytial virus (RSV) and group A Streptococcus antigen-based POCTs are inconsistent J o u r n a l P r e -p r o o f 5 [26] [27] [28] [29] ; and accurate dengue antigen-based RDTs have not been found to be cost-effective in resource-poor settings [30, 31] .

Non-specific host biomarkers measure the host-response to stimuli, and have been evaluated in the context of fever to discriminate between bacterial and viral pathogens [32] . C-reactive protein (CRP) is one of the most studied host-response biomarkers of bacterial infection, consistently showing high sensitivity and moderate specificity, and CRP POCTs have been shown to be cost-effective in resource-poor environments [30, 32] . However, 80% studies evaluating CRP performance originate from high income countries [32] . In Southeast Asia, these evaluations are mainly hospital-based [33] [34] [35] other than a single community-based study [36] . Good diagnostic performance of CRP in identifying bacterial infections was observed but generalisability was limited due to demographic, clinical and diagnostic heterogeneity of these studies.

In this study, we aim to identify key organisms among acutely febrile children and adults attending primary health care in Southeast Asia, and to evaluate the performance of CRP for discriminating between bacteria and viruses.

Chiang Rai province is the northernmost province in Thailand bordering Myanmar and Lao People's Democratic Republic. The majority of the population are Thai, with approximately 15% ethnic minorities and hill-tribes. The six participating primary care sites were located within a 30-kilometre radius of Chiang Rai city centre, covering rural and peri-urban as well as mountainous and plateau areas.

Hlaing Tha Yar, Lower Myanmar, is a peri-urban township on the west side of Yangon. The township has the highest rates of diseases related to hygiene and environmental conditions (e.g. diarrhoea, dysentery, and tuberculosis) in Yangon [37] . Four sites were included: three primary care clinics and one outpatient department from a public governmental hospital.

Both Chiang Rai and Hlaing Tha Yar are defined by a tropical climate.

Specimens were collected from febrile patients recruited into a previously described multi-centre randomised-controlled trial evaluating the impact of C-reactive protein (CRP) testing on antibiotic prescription in primary care [38] . Febrile children and adults (defined as ≥12 years of age) were recruited between June 2016 and August 2017. Inclusion criteria were age ≥1 year with a documented fever (defined as a tympanic temperature >37.5˚C) or a chief complaint of acute fever (<14 days), regardless of previous antibiotic intake and co-morbidities other than malignancies. Exclusion criteria were symptoms requiring hospital referral defined as either impaired consciousness; inability to take oral medication or convulsions; a positive malaria test; the main complaint being trauma and/or injury; suspicion of either tuberculosis, urinary tract infection, local skin infection or dental abscess; any symptom present for more than 14 days; any bleeding; and inability to comply with the follow-up visit at day 5.

On the day of enrolment, all patients had demographic information collected and underwent a routine clinical examination including vital signs (blood pressure, pulse, respiratory rate, temperature).

Patients were followed-up after their enrolment both at day 5 and day 14.

Of the 2,392 febrile children and adults recruited, 799 were randomly allocated to the control group with blood specimens collected for off-site CRP testing (as compared with the intervention groups that had CRP tests performed on-site). Details are illustrated in Figure 1 . Antibiotics were prescribed to this group according to routine clinical practice, clinicians were not informed of the CRP results or any aetiological findings. In case of co-detection in blood and NP swabs, target organisms detected in blood were assumed to be the primary cause of illness.

Bacterial and viral aetiological groups included all cases where any bacteria or viruses were detected in blood, respectively, and only target viruses and bacteria detected in NP swabs.

We described organism distribution among children and adults (defined as ≥12 years of age)

separately. Descriptive analysis for continuous variables with normal distribution used means and standard deviations (SD) and medians with inter-quartile ranges (IQR) for non-normally distributed continuous variables. Comparison between groups used t-tests for normally distributed variables, the Mann-Whitney test for non-normally distributed variables, and chi-squared test for categorical variables.

CRP values were compared across aetiological groups and clinical syndrome using the Mann-Whitney U test for two-group comparisons and the Kruskal-Wallis test for multi-group comparisons.

Non-parametric receiver operating characteristic (ROC) curves were plotted and the Wald test was used to compare areas under the curve. Covariates included the following factors: patient age and prior use of antibiotics. Diagnostic accuracy was assessed by calculating the areas under the ROC J o u r n a l P r e -p r o o f 9 curves (AUC). An AUC of >0.9 was considered excellent; 0.8-0.9, very good; 0.7-0.8, good; 0.6-0.7, average; <0.6, poor [39, 40] . Sensitivity, specificity, and percentage of correctly classified cases were also assessed for the two CRP cut-off points used in the original trial: 20 mg/L and 40 mg/L and these were compared with the accuracy of routine prescribing practice [38] .

Data analyses were performed with STATA version 15 (College Station, Texas, USA).

The protocol, informed consent form and case record forms were reviewed and approved by the 

Of the 799 patients prospectively enrolled and randomised into the trial control group, 773 (96.8%) had at least one blood or NP swab specimen collected, including 371 (48.0%) children and 402 adults (52.0%). Out of these 773 patients, 33 had only a NP swab and no blood collected, while 146 had only a blood specimen collected without a NP swab. Among these 146 patients, 80 had a blood specimen obtained on a dried blood spot (DBS). As shown in Table 2 , children presented significantly earlier after symptom onset than adults, with fewer comorbidities and less self-reported medication (p <0.05). Antibiotic intake declaration was similar in children and adults (p = 0.237).

Page 10 of 31 J o u r n a l P r e -p r o o f 10 Clinically, respiratory syndrome was the most prevalent presentation both in children and adults.

Within patients with a respiratory syndrome, the most frequent symptoms were localised in the upper respiratory tract including common cold, diagnosed in 48.2% (120/249) of children and 45.0%

(108/240) of adults (p = 0.479). Gastrointestinal syndrome was the second most prevalent presentation, and there were no differences between children and adults. 630 blood specimens tested for bacterial screening, using Taqman array card (n=601) and bacterial singleplex polymerase chain reaction (n=626) ** 634 blood specimens tested for Leptospira screening, using the Taqman array card (n=601), the bacterial singleplex polymerase chain reaction (n=626) and the microagglutination test (n=134) *** 658 blood specimens tested for Orientia tsutsugamushi and Rickettsia spp. screening, using Taqman array card (n=601), the bacterial singleplex polymerase chain reaction (n=626), and the indirect immmunofluoresence assay (n=656) **** 601 blood specimens tested using the Taqman array card only (n=601) ***** 686 blood specimens tested for dengue, chikungunya and zika virus screening, using Taqman array card (n=601), the viral singleplex polymerase chain reaction on fresh blood (n=626) and dried blood spot (n=245)

No evidence for a difference in antibiotic prescription was observed between the bacterial and viral groups at day 0, and clinical outcomes were also not significantly different between the two groups (Table 4 ).

Outcome characteristics by aetiological group in Chiang Rai, northern Thailand and Hlaing Tha Yar, Lower Myanmar, 2016-2017.

The prescription of antibiotics at the facility was considered between the enrolment at day 0 until day 14 of the follow-up Severity was ranked from 1-4 with severity=1 as the less severe presentation CRP: C-reactive protein Elevated CRP defined as ≥50 mg/L in children and ≥100 mg/L in adults SAE: serious adverse event, defined as admission to hospital or death within 14 days of enrolment Broad-spectrum antibiotics include ceftriaxone, cefixime, ciprofloxacin, levofloxacin, azithromycin, and amoxicillin with clavulanic acid.

Among patients with a bacterial organism, two-thirds did not receive any antibiotic ( Occurrence of SAE, n (%) 0 (0) 0 (0) 1.000

Unscheduled visits, n (%) 0 (0) 6 (3.1) 0.281 antibiotic. No evidence for a difference in clinical outcomes was observed after 14 days of follow-up, regardless of whether an antibiotic was prescribed.

Of 

We investigated the spectrum of organisms among febrile children and adults in the community and evaluated the performance of CRP in distinguishing bacteria from viruses including its potential impact on antibiotic prescription compared with current practice. Patients were recruited prospectively across ten sites in Thailand and Myanmar including urban, semi-urban and rural areas spanning over a full calendar year.

In our study, Leptospira spp., influenza virus and dengue virus were the leading organisms identified, which is consistent with previous reports in the region [8, 36] . The broad inclusion criteria, allowing for enrolment of all patients over 1 year old regardless of previous antibiotic intake, comorbidities, or clinical presentation, make our findings more generalisable than previous studies.

Investigating non-malarial acute febrile illness remains challenging in resource-poor areas [8] , and despite screening for multiple organisms on blood and respiratory specimens, we were only able to identify a probable cause of fever in 227 (29.4%) of patients. This low detection may be explained by the inclusion of only non-severe outpatients [14, 41, 42] , while other studies in Southeast Asia recruiting more severe and hospitalised patients identified an organism in around 50% of cases [2] [3] [4] 35] . Only 15.9% (36/227) of organisms detected were bacteria, which may be explained by the lower risk of bacterial infections in non-severely ill patients, and where present, characterised by lower bacterial loads [42] .

Most bacteria were identified using a singleplex PCR and not the TAC assay, while viruses were equally detected by these two molecular methods. This lower sensitivity in the TAC assay for the detection of bacteria has been described in previous studies using multi-pathogen molecular detection platforms [43, 44] . The trade-off between advantages for screening multiple organisms at the same time with a simplified molecular platform should be weighed against potentially lower sensitivity, especially for bacteria such as O. tsutsugamushi or Leptospira spp., which are considered important drivers of acute febrile illness in Southeast Asia [45, 46] . the TAC assay to identify infections among neonates in South Asia, but detected the presence of certain organisms among both controls and cases [47] . A 2019 multi-country study into causes of severe pneumonia also excluded molecular assay results positive for K. pneumoniae because of poor assay specificity [48] .

Furthermore, most of our patients presented with low CRP regardless of whether a bacterial or viral organism was detected, and recovered regardless of whether an antibiotic was prescribed. It is likely that invasive bacterial infections requiring an antibiotic are exceptions while most primary care patients present with a self-limiting infection [49] . Other primary care-based studies have recently supported restriction of antibiotic prescription to a small minority of patients: in Tanzania, a clinical trial using a 80 mg/L threshold lowered antibiotic reduction to 2.3% without affecting outcomes, while a U.S. study concluded that 72% of outpatients attending a general practice for a respiratory presentation should not even require a medical consultation, let alone an antibiotic prescription [50, 51] . Strategies whereby testing for CRP as a predictor of clinical outcome rather than determining aetiology have been evaluated in primary care: a cluster-randomised controlled trial in Belgium showed CRP to rule-out serious infection using a 5 mg/L threshold, while a systematic review found CRP-testing to be useful in identifying serious infections among febrile children [52, 53] .

In our study, CRP performance in distinguishing bacteria from viruses was average (AUC 0.65) and

lower than another study from the region which found (AUC 0.83 among 1,372 patients with a microbiologically-confirmed diagnosis from Thailand, Cambodia and Lao PDR [36] Our study has several limitations, mostly relating to the limited scope and accuracy of the reference diagnostic tests, and the impact of even slightly less than perfect "gold-standard" reference tests on the evaluation of new diagnostic and biomarker tests can be profound [56, 57] . As mentioned above, the sensitivity and specificity of the multiplex TAC assay was not optimal for bacteria detection, and the absence of convalescence specimens impeded our ability to diagnose patients based on serology, particularly with respect to bacterial zoonoses. Even genuine detection of bacterial and viral DNA in normally sterile sites cannot be used to conclusively determine causality, as this has been reported among healthy individuals, and in patients even weeks after recovery from infections, challenging the interpretation of molecular assays [58, 59] . Blood culture was not available and this further limited our aetiological investigation. We did not recruit a concomitant control group, which precludes robust attribution of causality in the organisms we detected, particularly in NP swabs. In a paediatric study in Asia and Africa, the inclusion of controls matched with pneumonia cases weakened the evidence of causality for almost all organisms detected [48] , and only RSV, hMPV, influenza virus A and B, parainfluenza virus type 1 and B. pertussis were considered pathogenic, consistent with other LMICbased studies [60, 61] . These organisms, however, are sometimes present in healthy individuals, with a prevalence of influenza virus among healthy children between 0-6%, RSV at 0-9% and hMPV between 0-7% [62] . On the other hand, we did not regard other organisms detected in NP swabs such as rhinovirus, parainfluenza virus or S. pneumoniae as pathogenic, because these are commonly detected among healthy individuals [48, 63, 64] . All these limitations in reference diagnostic tests might explain the average performance of CRP in our analysis.

We presented the key organisms detected among febrile children and adults attending primary healthcare in Southeast Asia. The performance of CRP in distinguishing between bacterial and viral organisms was limited, although the current findings suggest that CRP-guided treatment would increase the appropriate use of antibiotics with respect to aetiology. This is supported by the overall reduction in prescribing compared with current practice demonstrated in the original trial. Our findings also support conclusions from previous studies that even in the presence of bacterial organisms, very few ambulatory patients are likely to benefit from the extensive and poorly targeted antibiotic prescribing practices that currently prevail in most Southeast Asian primary care settings. 

The funders had no role in study design, data collection, data interpretation or writing the manuscript.

The corresponding author had full access to all the data and took the final decision to submit for publication.

 Point-of-care diagnostic tools could guide health workers' antibiotic prescription  C-reactive protein was significantly increased in case of bacterial infections  Most primary care patients recovered regardless of antibiotic prescription  Antibiotic prescription should be an exception in the primary levels of care

Influenza virus type B with 22 cases (3.5%), hMPV with 16 cases (2.6%), and

World Health Organization. Key points: World malaria report

A prospective study of the causes of febrile illness requiring hospitalization in children in Cambodia

Causes of fever in rural Southern Laos

Southeast Asia Infectious Disease Clinical Research Network. Causes and outcomes of sepsis in southeast Asia: a multinational multicentre cross-sectional study. The Lancet Global health

Diversity of infectious aetiologies of acute undifferentiated febrile illnesses in south and Southeast Asia: a systematic review

Febrile illness in Asia: gaps in epidemiology, diagnosis and management for informing health policy. Clinical microbiology and infection : the official publication of the European Society of Clinical Microbiology and Infectious Diseases

Maximising access to achieve appropriate human antimicrobial use in low-income and middle-income countries

Acute undifferentiated febrile illness in rural Cambodia: a 3-year prospective observational study

Infectious etiologies of acute febrile illness among patients seeking health care in south-central Cambodia

Dengue and other common causes of acute febrile illness in Asia: an active surveillance study in children

The fifth dimension of innate immunity

The importance of pathogen load

Why should we measure bacterial load when treating community-acquired pneumonia? Current opinion in infectious diseases

Meningococcal bacterial DNA load at presentation correlates with disease severity. Archives of disease in childhood

Beyond malaria--causes of fever in outpatient Tanzanian children. The New England journal of medicine

Causes of nonmalarial febrile illness in outpatients in Tanzania. Tropical medicine & international health

World Health Organization. The world health report 2006: working together for health: World Health Organization

Human resources for health: overcoming the crisis

Human resources for health in southeast Asia: shortages, distributional challenges, and international trade in health services

Diagnosis of scrub typhus. The American journal of tropical medicine and hygiene

World Health Organization. Rapid advice: diagnosis, prevention and management of cryptococcal disease in HIV-infected adults, adolescents and children

Scrub typhus point-of-care testing: A systematic review and meta-analysis

Rapid tests for diagnosis of leptospirosis: current tools and emerging technologies. Diagnostic microbiology and infectious disease

Evaluation of a simple blood culture amplification and antigen detection method for diagnosis of Salmonella enterica serovar typhi bacteremia

Diagnostic accuracy of antigen-based immunochromatographic rapid diagnostic tests for the detection of Salmonella in blood culture broth

Poor clinical sensitivity of rapid antigen test for influenza A pandemic (H1N1) 2009 virus. Emerging infectious diseases

Are Rapid Influenza Antigen Tests Still Clinically Useful in Today's Molecular Diagnostics World? Hawai'i

Diagnostic Accuracy of Rapid Antigen Detection Tests for Respiratory Syncytial Virus Infection: Systematic Review and Meta-analysis

Rapid antigen detection test for group A streptococcus in children with pharyngitis. The Cochrane database of systematic reviews

Modelling the Impact and Cost-Effectiveness of Biomarker Tests as Compared with Pathogen-Specific Diagnostics in the Management of Undifferentiated Fever in Remote Tropical Settings

A systematic review of the economic impact of rapid diagnostic tests for dengue. BMC health services research

Host biomarkers for distinguishing bacterial from non-bacterial causes of acute febrile illness: a comprehensive review

Etiology of central nervous system infections in the Philippines and the role of serum C-reactive protein in excluding acute bacterial meningitis. International journal of infectious diseases : IJID : official publication of the International Society for Infectious Diseases

Serum C-reactive protein concentrations in Malaysian children with enteric fever

Causes of acute undifferentiated fever and the utility of biomarkers in Chiangrai, northern Thailand

Performance of C-reactive protein and procalcitonin to distinguish viral from bacterial and malarial causes of fever in Southeast Asia

Poverty among households living in slum area of Hlaing Tharyar Township

Effect of point-ofcare C-reactive protein testing on antibiotic prescription in febrile patients attending primary care in Thailand and Myanmar: an open-label, randomised, controlled trial. The Lancet Global health

ROC analysis of the accuracy of Noncycloplegic retinoscopy, Retinomax Autorefractor, and SureSight Vision Screener for preschool vision screening. Investigative ophthalmology & visual science

Advances in statistical methodology for the evaluation of diagnostic and laboratory tests

Detection of viral respiratory pathogens in mild and severe acute respiratory infections in Singapore

Update on detection of bacteremia and fungemia

Diagnostic accuracy of SeptiFast multi-pathogen real-time PCR in the setting of suspected healthcareassociated bloodstream infection

Evaluation of two multiplex real-time PCR screening capabilities for the detection of Bacillus anthracis, Francisella tularensis and Yersinia pestis in blood samples generated from murine infection models

A review of the global epidemiology of scrub typhus. PLoS neglected tropical diseases

The importance of leptospirosis in Southeast Asia. The American journal of tropical medicine and hygiene

Causes and incidence of community-acquired serious infections among young children in south Asia (ANISA): an observational cohort study

Causes of severe pneumonia requiring hospital admission in children without HIV infection from Africa and Asia: the PERCH multi-country case-control study. The Lancet. 2019. 49. Van den Bruel A

Safety and Efficacy of C-reactive Protein-guided Antibiotic Use to Treat Acute Respiratory Infections in Tanzanian Children: A Planned Subgroup Analysis of a Randomized Controlled Noninferiority Trial Evaluating a Novel Electronic Clinical Decision Algorithm (ePOCT)

Necessity of office visits for acute respiratory infections in primary care

Diagnostic value of laboratory tests in identifying serious infections in febrile children: systematic review

Point-of-care C-reactive protein testing to reduce inappropriate use of antibiotics for non-severe acute respiratory infections in Vietnamese primary health care: a randomised controlled trial

The effect of uncertainty in patient classification on diagnostic performance estimations

Imperfect gold standards for biomarker evaluation

Van Der Pol B. Duration of polymerase chain reaction-detectable DNA after treatment of Chlamydia trachomatis, Neisseria gonorrhoeae, and Trichomonas vaginalis infections in women. Sexually transmitted diseases

Detection of the latent form of Epstein-Barr virus DNA in the peripheral blood of healthy individuals

Microorganisms associated with pneumonia in children< 5 years of age in developing and emerging countries: the GABRIEL pneumonia multicenter, prospective, case-control study

The role of influenza, RSV and other common respiratory viruses in severe acute respiratory infections and influenza-like illness in a population with a high HIV sero-prevalence

Viral and bacterial interactions in the upper respiratory tract

Frequent detection of respiratory viruses without symptoms: toward defining clinically relevant cutoff values

Asymptomatic Shedding of Respiratory Virus among an Ambulatory Population across Seasons. mSphere

We thank all primary care patients and health workers from Chiang Rai primary care centres and Hlaing Tha Yar public hospital and MAM clinics for taking part in the study. We are very grateful to the study staff and to the Clinical Trial Support Group at Mahidol-Oxford Tropical MedicineResearch Unit for ensuring the successful completion of the study. We thank Heiman Wertheim, Arjen Dondorp, Direk Limmathurotsakul, Christopher Parry, and Paul Newton for guidance on the study design; Clare Ling, Toni Whistler, Ampai Tanganuchitcharncha, Areerat Thaiprakhong, Nattapon Pinthong, Prapaporn Srilohasin and Kyaw Soe for laboratory support, as well as Duangjai Suwancharoen from the national institute for animal health (NIAH) for carrying out the Leptospira MAT. Finally, we thank Elizabeth Ashley, Jeroen Bok, Joshua Cohen, Ni Ni Tun, and Khin Yupar Soe for their assistance in study site coordination.