Use of simple clinical and laboratory predictors to differentiate influenza from dengue and other febrile illnesses in the emergency room

Influenza is one of the most common infectious diseases worldwide [1]. The spectrum of clinical manifestations ranges from mild-form nonspecific febrile illness, such as cough, sore throat, headache, rhinorrhea, malaise, and muscle ache, to respiratory failure and death [2]-[4]. Its symptoms and signs can be similar to other viral illnesses such as dengue, making the infections difficult to distinguish. The clinical manifestations of dengue illness vary greatly, ranging from a mild, flu-like, and self-limited febrile illness to severe dengue [5],[6]. Influenza and dengue overlap geographically in tropical and subtropical regions of the world [1],[7]. Concurrent influenza and dengue outbreaks have been reported [8]-[11]. The overlapping clinical features of influenza and dengue created clinical diagnosis and management challenges during simultaneous influenza and dengue outbreaks in Puerto Rico in 1977 [11]. Dengue epidemics have occurred in Taiwan for decades [12],[13]. An outbreak of 2009 pandemic H1N1 occurred in Taiwan after the first case was identified on May 20, 2009 [14]. The overlapping symptoms and signs of influenza and dengue created a confusing clinical situation and challenges in etiology identification. Although rapid diagnostic tests can help confirm an influenza diagnosis, their sensitivity ranges from 40-70% depending on the day of illness and specimen type [15]. Definitive diagnosis of dengue is made using serology tests, but these tests are not always readily available in most clinical laboratories [7],[16]. Our study aimed to identify clinical and laboratory features that distinguish influenza from dengue and other febrile illnesses (OFI) in dengue and non-dengue endemic settings. We applied decision tree analysis to our dataset to discriminate patients with influenza from those with dengue. Our findings may be valuable for clinicians working in crowded emergency rooms (ER) in countries with limited medical resources.

Our study investigated simple clinical and laboratory features to identify patients with influenza among children and adults with acute febrile illness in ER in dengue and non-dengue endemic areas. In our series, the mean time lapse between onset of illness and ER presentation was 1.7, 3.7, and 3 days for influenza, dengue, and OFI, respectively. This permits a detailed comparison of early clinical and laboratory characteristics between influenza and dengue or OFI.

The symptoms of influenza overlap substantially with dengue and OFI (Table 2). We found that symptoms resulting from respiratory tract infections such as cough, rhinorrhea, sore throat, and dyspnea were more prevalent in patients with influenza than with dengue. In contrast, non-respiratory tract symptoms including rashes, headache, abdominal pain, joint pain, and orbital pain were less common in the patients with influenza compared to patients with dengue. The influenza virus primarily infects epithelial cells of the respiratory tract and causes upper respiratory symptoms [25]. Damage to infected cells results in release of inflammatory mediators, leading to a systemic response (i.e., fever, headache, joint pain, malaise, and myalgia) with symptoms similar to other viral illnesses such as dengue [26],[27].

The overlapping geographic range of influenza and dengue as well as simultaneous dengue and influenza outbreaks [8]-[11], often in resource-limited countries, leads to diagnostic difficulties, as nonspecific symptoms are common to both infections. Early detection of influenza and dengue is especially important because influenza may be prevented through proper isolation and treated using antiviral agents [28],[29], whereas failure to make a timely dengue diagnosis with adequate fluid replacement can lead to severe dengue [30]. Our study demonstrated that the presence of leukopenia (white cell count <3.0 × 10⁹ cells/L) and marked thrombocytopenia (platelet count <100 × 10⁹ cells/L) are useful for differentiating dengue from influenza in both adults and pediatric patients. The absence of rash further discriminated influenza from dengue in pediatric patients. Rash associated with influenza is not a common manifestation, occurring in only 2-8% of patients, usually children [31],[32]. In contrast, the reported frequency of rash in dengue cases ranges from 46-68%, particularly in children less than 15 years who usually have a nonspecific febrile syndrome accompanied by rash [33],[34]. Mild leukopenia and relative lymphopenia are typical findings of influenza; thrombocytopenia may be present in complicated cases [4],[14],[25]. Notably, our study and others found leukopenia and thrombocytopenia to be common laboratory findings in dengue infection, with platelet counts below 20 × 10⁹ per liter often observed in severe dengue [7],[30]. A study examining predictors of diagnosis in 1,962 febrile travelers returning from the tropics found the main predictors of dengue infection compared with other fevers, excluding malaria, to include skin rash, thrombocytopenia, and leukopenia [35]. This finding was consistent with our study findings.

Our data suggest that skin rash, leukopenia (white cell count <3.0 × 10⁹ cells/L), and platelet counts <100 × 10⁹ cells/L were useful to predict negative influenza results during influenza and dengue epidemics. However, it is unlikely that any single indicator will be useful in clinical practice because these symptoms and laboratory findings can be present in both infections [11]. In our series, we developed two simple and practical diagnostic algorithms using clinical and laboratory indicators to distinguish influenza from dengue in adult and pediatric patients, respectively. We found that the diagnostic algorithm correctly classified 94.2% of pediatric patients with influenza in the "high probability" group with only one misclassified dengue patient (Figure 1). Analysis of our data after excluding pediatric patients showed similar results, except that skin rash that was no longer associated with influenza infection in adult patients; however, the diagnostic tree (Figure 2) still correctly classified 87.8% of adult influenza cases as "high probability". These findings underscored that rash, leukopenia, and marked thrombocytopenia could help to establish a diagnostic algorithm to distinguish influenza from dengue patients during outbreaks of both diseases. Additional prospective studies are needed to validate this predictive model in other dengue-endemic regions and in populations with different ethnicities.

The wide range of influenza-associated symptoms often makes it difficult to distinguish from other febrile or respiratory illnesses [25]. A crowded ER can make it challenging for physicians to differentiate between influenza and OFI [36]. Early antiviral treatment (≤48 h) of influenza is especially important for patients with underlying risk factors to avoid otherwise preventable morbidity and mortality [37]. Our results demonstrate that rhinorrhea, malaise, sore throat, and dyspnea in addition to a slightly low platelet count (100-149 × 10⁹ cells/L) in children, as well as rhinorrhea and malaise in adults were valuable predictors during the evaluation of the likelihood of influenza versus OFI in a non-dengue endemic setting. However, apart from rhinorrhea, which lacked adequate sensitivity and specificity, we found that all other variables were specific (>70%) but not sensitive enough in distinguishing influenza from OFI. This is not surprising as more than 60% of OFI cases included in this series were individuals with pharyngitis or bronchitis (Table 4), another viral infection commonly encountered in the ER [38]. The information of this study are most helpful for clinicians to facilitate diagnosis of influenza during periods of high influenza activity in non-dengue endemic setting.

In the present study, patients with influenza had a shorter fever duration than the dengue cases (mean 2.8 ± 1.6 vs. 4.0 ± 2.3 days). This finding is similar to a previous study that found a fever duration of <4 days for patients with influenza [25], while the average length of fever in dengue patients was approximately 5 days [39], coinciding with the disappearance of viremia.

Although influenza primarily causes upper respiratory tract infections, pulmonary and extra-pulmonary complications have also been reported [2],[40],[41]. Previous studies from the United States and Australia of critically ill patients with 2009 pandemic H1N1 infections found ARDS complication in 35.8% and 48.8% of cases and a 45% and 14.3% hospital mortality, respectively [40],[41]. ARDS was noted in 5 influenza cases in our study (Table 3); 3 were fatal. Remarkably, delayed oseltamivir therapy was found in all fatal cases in our series. The importance of a timely anti-viral therapy for severe influenza should therefore be emphasized.

Earlier studies describe neurological complications of influenza including aseptic meningitis, encephalopathy/encephalitis, Guillain-Barré syndrome, and transverse myelitis [42]-[44]. Neurological complications have been reported for patients with 2009 pandemic H1N1 infections [43]. Of 447 patients with 2009 pandemic H1N1 infection in this study, meningoencephalitis was found in 6 (1.3%), with presenting symptoms of altered mental status and seizures (Table 3). Other influenza-associated extrapulmonary complications among our patients included myocarditis, rhabdomyolysis, gastrointestinal bleeding, and intracranial hemorrhage (Table 3). Physicians should be aware of influenza-associated extrapulmonary complications when caring for influenza patients and manage them accordingly.

Gastrointestinal bleeding was the most common complication among dengue patients in this study. Previous studies have reported gastrointestinal bleeding to be a warning sign of severe dengue; timely management and intensive monitoring of patients with gastrointestinal bleeding is therefore necessary [45].

This study has several limitations. First, because it was conducted in a single medical center, disease severity may have been biased by referral patterns. Second, as a retrospective study, missing laboratory data was inevitable. Third, the decision to perform diagnostic tests for influenza and dengue infections was based on individual physicians' experience; therefore, patients included in our series may have been biased by individual physicians' personal judgments. In addition, our results were based on data from early stages of illness; future studies are necessary to validate these findings in different course of illness for better generalization of their utility.

This study demonstrates substantial overlap in clinical presentation between influenza and dengue as well as OFI. This study also identifies simple and useful clinical and laboratory data to enable identification and facilitate diagnosis of influenza in different clinical settings (dengue and non-dengue endemic areas). We provide two decision tree algorithms using simple clinical and laboratory data that can be easily implemented in resource-limited countries to differentiate patients with influenza from those with dengue. This information is especially important to clinicians in countries where medical resources are sparse and the burden of influenza and dengue is high.

SYH and IKL made major contributions to sample collection and wrote the manuscript. IKL conceived and designed the study, analysis of the data and preparation of the manuscript. SYH, SCH, CCC, TYC, WCH, JWL and LW collected clinical data. All authors read, commented on, and approved the final manuscript.