The Indian Journal of Pediatrics https://doi.org/10.1007/s12098-017-2591-y
Development of a Prognostic Prediction Model to Determine Severe Dengue in Children Priya Sreenivasan 1 & Geetha S 1,2 & Sasikala K 2 Received: 8 May 2017 / Accepted: 18 December 2017 # Dr. K C Chaudhuri Foundation 2018
Abstract Objective To develop a prognostic prediction model using the seven warning signs highlighted by WHO revised Dengue fever classification 2009 to determine severe dengue in children. Methods In this prospective analytical study conducted in a tertiary care centre, consecutive sampling of all children aged 1mo to 12y admitted with serologically confirmed Dengue was done from May 2015 through August 2016. After excluding 27 patients with co-infections and co-morbidities, 359 patients were followed up daily to assess clinical and laboratory progression till discharge/ death. Independent predictors were abdominal pain or tenderness, persistent vomiting, lethargy, mucosal bleed, clinical fluid accumulation, hepatomegaly >2 cm and rising hematocrit concurrent with platelet count <100 × 109/L. Outcome measure was severe dengue defined as per WHO guidelines 2009. Results Among 359 children, 93 progressed to severe dengue. In univariate analysis, significant predictors were clinical fluid accumulation (OR 4.773, p = 0.000, 95%CI 2.511–9.075), persistent vomiting (OR 1.944, p = 0.010, 95%CI 1.170–3.225), mucosal bleed (OR 2.045, p = 0.019, 95%CI 1.127–3.711) and hematocrit ≥0.40 concurrent with platelet count <100 × 109/L (OR 2.985, p = 0.000, 95%CI 1.783–4.997). The final multivariable model included clinical fluid accumulation (aOR 3.717, p = 0.000, 95%CI 1.901– 7.269), hematocrit ≥0.40 concurrent with platelet count <100 × 109/L (aOR 2.252, p = 0.004, 95%CI 1.302–3.894) and persistent vomiting (p = 0.056) as predictors of severe dengue. Conclusions Among seven WHO warning signs, predictors of severe dengue as suggested by the present multivariable prediction model include clinical fluid accumulation, persistent vomiting and hematocrit ≥0.40 concurrent with platelet count <100 × 109/L. Keywords Dengue fever . Pediatric . Prediction model . Severe dengue . WHO 2009 classification . Warning signs Abbreviations AP/T AST/ALT CFA DIC/MODS GCS HM
Abdominal Pain/ Tenderness Aspartate Transaminase/ Alanine Transaminase Clinical Fluid Accumulation Disseminated Intravascular Coagulation/ Multi Organ Dysfunction Syndrome Glasgow Coma Scale Hepatomegaly
Intravenous Fluids Lethargy/ Restlessness Mucosal Bleed Non Severe Dengue Platelet Count Packed Cell Volume Persistent Vomiting Severe Dengue Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis World Health OrganizationTropical Diseases Research Warning Signs
Department of Pediatrics, Government Medical College, Thiruvananthapuram, Thiruvananthapuram District, Kerala 695011, India
Clinical Epidemiology Research & Training Centre (CERTC), Government Medical College, Thiruvananthapuram, Kerala, India
Re-emergence of Dengue as a pandemic-prone infection is a major threat to global health. Its endemicity spans over 125
Indian J Pediatr
countries including Southeast Asian, Western Pacific and Latin American regions. Estimated 50–100 million infections occur yearly with 20,000 deaths worldwide . India is dengue hyper-endemic with all four serotypes in circulation; 1,29,166 cases and 245 deaths were reported in 2016 . Considering the clinically unpredictable course of illness, the WHO-TDR supported DENCO study group (2008) developed an evidence based classification comprising two entities, dengue and severe dengue (SD) . This large multicentre prospective study conducted in four Southeast Asian and three Latin American countries included both adults and children as participants. They also evaluated warning signs (WS) of disease progression to SD. In their model, abdominal pain/ tenderness, lethargy, mucosal bleed (MB) and low platelet count (PC) were associated with increased risk for SD; raised hematocrit was included though not statistically significant. Later in 2008, a global expert group convened by WHOTDR in Geneva added three more WS [persistent vomiting (PV), hepatomegaly >2 cm and clinical fluid accumulation] and proposed the revised classification model with a total of seven WS . Its applicability was assessed by a WHO-TDR funded multicentre study involving 18 countries (2009) . Though it found acceptance and user-friendliness for triage and management, need for further research on predictive value of WS were highlighted as they were too many in number and too subjective with unclear definitions [6, 7]. Over the years, studies failed to identify other simple clinical or lab predictors other than WS carefully chosen by WHO. Children acquire overt disease more and behave differently from adults in disease progression . But both these populations share the same classification model and warning signs. Progression to severity varies with region. Serum samples from patients admitted in authors’ hospital showed lineage shift in Indian strains of Dengue virus serotype 3 that attributed to dramatic increase in disease severity in their region during 2008–2011 . According to WHO 2009 guidelines, presence of any WS warrants admission to a health care facility followed by intravenous fluid administration. This leads to individual judgment, over-hospitalization and huge economic burden . Considering above said factors, the present study aims to develop a prognostic prediction model to determine SD in children using seven WS suggested by WHO as independent predictors.
Material and Methods A prospective analytical study was done in a tertiary care teaching centre in Kerala, a South- Indian state, from May 2015 through August 2016. Consecutive sampling of children 1mo-12y admitted with serologically proven dengue (defined as per WHO revised guidelines 2009) was done . Details of baseline demography, presence of each WS with
exact day and time of appearance, receipt of intravenous fluids (mL/kg/h) and available laboratory parameters were collected. Blood counts with hematocrit, dengue serology, liver and renal function tests were done at admission. Referring doctor was contacted for clarifications if needed. All patients with either NS1Ag ELISA positivity (done if admitted within 5d of onset of fever), IgM ELISA positivity (done if admitted after 5d of onset of fever) or both (done if admitted on 5th day of onset of fever) were followed up. Appearance of WS, progression to SD and need for intravenous fluids (mL/kg/h) were noted daily till discharge or death. Total count, hematocrit and platelet counts were done daily, frequency of which varied (4-24 h) depending on clinical progression. Additional investigations were done when required. Patients with co-infections or co-morbidities were excluded. Independent variables were abdominal pain/ tenderness, PV, lethargy/ restlessness, MB, clinical fluid accumulation (CFA), hepatomegaly >2 cm and a rising hematocrit with concurrent fall in PC <100 × 109/L. Operational definitions were given for abdominal pain/tenderness (severe enough to warrant medical attention), PV (≥2 episodes of vomiting that amounts to fatigue or requires intravenous fluids), lethargy /restlessness (without altered sensorium), CFA (either pleural effusion not severe enough to cause respiratory distress as evidenced by reduced intensity of breath sounds on auscultation or ascites as evidenced by shifting dullness) and MB (bleed from gastrointestinal / genitourinary mucosa, nose or conjunctiva). Rising hematocrit is vaguely defined in WHO revised classification 2009. High prevalence of anemia in India and unavailability of baseline hematocrits add to the problem. Hence univariate analysis was planned at different hematocrit cut-offs (0.36–0.41) with SD as outcome. Lowest hematocrit with statistical significance was included for modeling. Independent variables were coded as present/ absent. CFA and hepatomegaly >2 cm were considered present if consensus opinion was obtained after independent examination by two consultants. If progressing to severity, a WS was coded ‘present’ only if it appeared at least 2 h before the onset of SD. This gap is logically required to commence effective action or referral after prediction. Blinding predictor assessment for outcome information was inherent in the prospective study design. Outcome variable was SD defined (WHO guidelines 2009) as severe plasma leak leading to shock and/ or fluid accumulation with respiratory distress, severe bleeding or severe organ impairment . Shock was considered if tachycardia, cool extremities, capillary refill ≥3 s, weak pulse, narrow pulse pressure (≤ 20 mmHg) or low systolic blood pressure (<80 mmHg if ≤5 y and <90 mmHg if >5 y) were noted after controlling fever. SD included CFA with respiratory distress as evidenced by tachypnea/ dyspnea. Severe bleed meant persistent/ overt bleed with unstable hemodynamics,
Indian J Pediatr
hypotensive shock with low/normal hematocrit before fluid resuscitation, rapid fall in hematocrit from an initial documented value (if available) with shock, shock refractory to consecutive intravenous fluids of 60 ml/kg, clinical/ radiological intracranial bleed or any bleed associated with disseminated intravascular coagulation (DIC)/ multi organ dysfunction syndrome (MODS). Severe organ impairment included aspartate transaminase (AST) or alanine transaminase (ALT) ≥ 1000 IU/L, altered consciousness with low glasgow coma scale (GCS) or convulsions apart from typical febrile seizures, rhythm abnormality in ECG apart from sinus arrhythmia or cardiogenic shock, serum creatinine more than or equal to twice the upper limit of normal and presence of DIC/MODS. Intravenous fluid therapy can influence disease progression. Receipt of intravenous fluids <10 mL/kg/h before onset of SD was considered a probable confounder; receipt ≥10 mL/ kg/h was considered as SD. TRIPOD statement gives guidelines for sample size calculation in prediction modeling . Accordingly, events per variable are taken as ≥10 for binary outcome variable. Minimum of 80 cases of SD were aimed as independent variables were 8 in number, including the probable confounding variable. Dichotomous variables were used in statistical analysis. Univariate analysis with each predictor was planned with binary logistic regression keeping SD as outcome (α 5%, β 20%). Predictor significance level for inclusion in model was predetermined (α 20%). Multiple logistic regression was planned starting with enter method proceeding with backward elimination. Model performance/ goodness of fit were to be assessed with Hosmer Lemeshow statistics and deviance method. Statistical analysis was planned with SPSS version16. Informed written consents, Institutional Research Board and Ethical Committee clearances were obtained before commencement of the study.
Results Patient selection process, follow-up and outcome are depicted in Fig. 1 and baseline characteristics in Table 1. Among 21 infants, the youngest baby was 41d old. In case of SD, those with WS at any time of illness are contrasted with those having onset of WS before progression to severity (Table 2). Organs involved among those with severe isolated organ failure were central nervous system (CNS) (6 patients; encephalitis, status epilepticus apart from febrile seizures), heart (2 patients; myocarditis and heart block), liver (1 patient; elevated liver enzymes >1000 IU/L), kidney (1 patient; acute kidney injury) and muscle (1 patient; severe myositis with
creatine phosphokinase levels above 4000 IU/L). 6 patients had multiple organ failure. In univariate analysis, CFA, PV, MB and hematocrit ≥0.40 concurrent with PC <100 × 109/L were found statistically significant (Table 2). Abdominal pain/ tenderness, lethargy/ restlessness and hepatomegaly >2 cm were not statistically significant. One hundred forty nine patients received intravenous fluids (either hypotonic or isotonic fluids but no colloids) before admission to authors’ hospital. Receipt of intravenous fluids <10 mL/kg/h either before or after admission to authors’ hospital was not significantly associated with outcome. Table 3 shows the final multivariable prediction model. Goodness of fit assessed with Hosmer Lemeshow statistics showed a well-fitting model (p = 0.561). The scatter plot (Fig. 2) validated the model and depicted homogeneity of variance.
Discussion The present study shows a multivariable prediction model for SD in children with CFA, PVand hematocrit ≥0.40 concurrent with PC <100 × 109/L as predictors. The study had limitations. Even with best efforts, robustness of definitions of abdominal pain and lethargy could not be ensured. Their subjective nature made elicitation from history/ clinical examination difficult for child, parent and investigators. This problem was encountered in other studies also. Alexander et al. have defined ‘severe abdominal pain’ in their annexure; their model includes ‘abdominal pain’ . The authors came across studies where lethargy was excluded from analysis due to its subjective nature [12, 13]. Clinical details of illness before admission to authors’ center were collected from parental recall. Initial investigation reports, referral letter details and telephonic conversations with referring doctor in selected patients minimized this recall bias. Ascertainment of outcome was done without blinding to predictor information. Outcome ascertainment was done by researchers with concurrence from treating clinician in all cases to minimize this bias. Most of the attempts made to identify predictors of SD included adults and were retrospective [12, 14–16]. Prediction of disease progression was based on calculations of sensitivity, specificity and predictive values in some studies [12, 13]. Zhang et al. published a meta-analysis in 2014 that included 16 studies (pediatric population in 8) to identify predictors of SD . Here, retrospective and many of the prospective studies had missing data on WHO 2009 WS. Horstick et al. in their systematic review that compares usefulness of 1997 and 2009 WHO classifications, have only 2 studies among 34 that focus on WS . Protocol for a
Indian J Pediatr Fig. 1 Flowchart showing the patient selection process, follow up and outcome
Serologically confirmed Dengue among 1mo-12y age group _ 386 patients
Study population _ 359 serologically confirmed patients (NS1Ag positive _159, IgM positive 167, Both positive 33)
Severe dengue 93 patients
Non severe dengue 266 patients
Death 4 patients
prospective multicentre observational study in 8 countries was published by Jaenisch et al. to identify simple parameters that predict severity .
Baseline characteristics of patients
NSDb Total SDa (n = 359) (n = 93) (n = 266)
Median age (years) Male gender (number) Median day of admission to the hospital Median day of defervescence Median duration of hospitalization (days) Median platelet count (× 109/L) Median hematocrit (proportion of 1.0) ICU admissions (number) Compensated shock (number) Decompensated shock (number) Respiratory distress (number) Severe bleed (number) Organ failure (number) Death (number)
7.75 194 5 6 5 60 0.38 40 83 58 16 5 17 4
Non severe dengue
9.2 44 5 6 6 42 0.40 33 83 58 16 5 17 4
7.375 150 5 6 4 70 0.37 7
WS becomes a predictor only if it occurs sufficiently early before onset of SD. Thein et al., in their retrospective study showed that while WS occurred in 86% of 1507 adult dengue patients, only 42% occurred before onset of SD . In the present study the authors prospectively collected more convincing data about time of appearance of each WS and its relation to the onset of SD (Table 2). CFA as a WS was proposed by the expert group convened by TDR-WHO. A late WS, it manifests after reasonable plasma leak or excess intravenous fluid usage. Though radiological evidence precedes clinical signs, latter occurs sufficiently early before the onset of SD. Recognition warrants clinical skill and meticulous daily examination. In a study validating utility of WS, Thein et al. found that CFA had high specificity (0.98) to predict SD . A study by Thanachartwet et al. on adults showed CFA to have high sensitivity (0.75) and specificity (0.905) in diagnosing SD at admission . The present study demonstrated an OR 4.773 (95%CI 2.511–9.075) in univariate analysis and an adjusted OR 3.717 (95%CI 1.901–7.269) for CFA. Another evidence based WS is rising hematocrit with concurrent rapid fall in PC <100 × 109/L. This may be the earliest WS to appear and if so, marks the onset of critical phase.
Indian J Pediatr Table 2
Univariate analysis of each warning sign with severe dengue as outcome
Missing number (n = 359)
NSD (n = 266)
SD (n = 93) Entire course
HM >2 cm PCV ≥0.40 & PC < 100 × 109/L PCV ≥ 0.40 PC < 100 × 109/L
Receipt of IVF <10 mL/kg/h
NSD Non severe dengue; SD Severe dengue; AP/T Abdominal pain/ Tenderness; L/R Lethargy/ Restlessness; PV Persistent vomiting; CFA Clinical fluid accumulation; MB Mucosal bleed; HM Hepatomegaly; PCV Packed cell volume; PC Platelet count; IVF Intravenous fluids
Thrombocytopenia occurs due to antibody mediated destruction, marrow suppression, peripheral sequestration and DIC . Temporary alteration of endothelial glycocalyx and functional capillary leak causes hemoconcentration and raised hematocrit. Rapid fall in PC precedes plasma leak. Increasing thrombocytopenia was strongly associated with increasing severity of vascular leak but only weakly to bleed in a study conducted by Wills et al. . Krishnamurthi et al. has shown that although there was no correlation between bleeding scores and pleural effusion index (their measure of vascular leak) or bleeding scores and platelet counts, there was a correlation between pleural effusion index and PC . Hence, clubbing hematocrit ≥0.40 with concurrent fall in PC < 100 × 109 /L as one WS by WHO has definite scientific backup; the word ‘rapid’ needs clarification. Though the authors did univariate analysis at different hematocrit cut-offs (0.36–0.41) with SD as outcome, lowest hematocrit with statistical significance was obtained at hematocrit ≥0.40 (OR 2.262, 95%CI 1.385–3.694). Hematocrit ≥0.40 concurrent with PC <100 × 109/L had an OR 2.985 (95%CI 1.783–4.997) and an adjusted OR 2.252 (95%CI 1.302–3.894). In the present data, 4 children with PCV ≥ 0.40 without concurrent PC < 100 × 109 /L also progressed to SD. Thein et al. found that hematocrit rise
Table 3 Final multivariable prediction model that predicts severe dengue in children
with rapid PC drop had high specificity (0.94) in predicting SD . In a retrospective pediatric study by Pongpan et al. in 777 patients, significant predictive ability for severity included hematocrit ≥0.40 (OR1.34, p = 0.003, 95%CI 1.10–1.64) and PC <50 × 109/L (OR 3.95, p < 0.001, 95%CI 3.14–4.96) separately . MB, a WS put forward by DENCO study can occur due to thrombocytopenia, platelet dysfunction and abnormalities of coagulation/fibrinolytic system . Krishnamurti et al. has shown that hemorrhage in dengue without circulatory collapse is mostly due to activation of platelets and is well compensated; severity of coagulopathy is unrelated to bleeding score and most importantly, vascular alteration may be the principal factor involved in the association of thrombocytopenia and hemorrhage with disease severity . Severe bleed in children is mostly due to profound shock following capillary leak . Though significant in univariate analysis (OR 2.045, 95%CI 1.127–3.711), MB was not found significant in multiple logistic regression with CFA, PV and raised hematocrit with concurrent fall in PC <100 × 109 /L as other variables in the index study. Collinearity was suspected but could not be demonstrated in statistical analysis. On backward elimination, model properties became better with exclusion of mucosal bleed. In
CFA PCV ≥ 0.40 with PC < 100 × 109/L PV Constant
1.313 0.812 0.522 −1.651
0.000 0.004 0.056 0.000
3.717 2.252 1.686 0.192
1.901–7.269 1.302–3.894 0.987–2.881
CFA Clinical fluid accumulation; PCV Packed cell volume; PC Platelet count; PV Persistent vomiting; b Regression co-efficient; aOR Adjusted Odds ratio
Indian J Pediatr
Fig. 2 Validation of the multivariable prediction model using scatter plot
an effort to keep the number of predictors to minimum, it is not included in present model. In the WHO 2009 classification, probable dengue has nausea/ vomiting as a symptom and PVas a WS. Unless clearly defined, distinction may be difficult. Studies adopt different operational definitions [14, 20]. Vuong et al. conducted a study and suggested that PV should be vomiting two times or more per day . It may occur due to intestinal wall edema, hepatic involvement or electrolyte imbalance. Coupled with inadequate oral intake, it augments hemoconcentration. In the univariate analysis, PV was statistically significant (OR 1.994, 95%CI 1.172–3.225). This clinically significant easily identifiable variable is included in the present model though not significant in multiple logistic regression (95%CI 0.987–2.881). Carrasco et al. found PV to be statistically significant among other variables as predictors for severity . Thein et al. found that PV had high specificity (0.93) for predicting severity . Abdominal pain and lethargy were not statistically significant here probably because they were too subjective and lacked proper definitions in spite of authors’ best efforts. In a review by Samanta et al., it is stated that hepatomegaly in dengue is commoner in children (37.5–80.8%) than in adults (4–52%) . In the present study, statistical significance was not obtained with hepatomegaly probably because unlike in adults, it is very common for infants and children to manifest hepatomegaly in the febrile phase itself.
The present study has practice implications as its findings, after external validation, can be incorporated into pediatric dengue admission and triage policy of the region. The present findings are generalizable to children with dengue.
Conclusions It is concluded from the prediction model that among the seven WS highlighted by WHO in 2009, CFA, PVand hematocrit ≥0.40 with concurrent PC <100 × 109/L predict SD in children. Acknowledgements The authors deeply acknowledge Kerala University of Health Sciences (KUHS) for granting permission for publication of this article, prepared as part of thesis work done in partial fulfilment of the requirements for Mphil (Clinical Epidemiology) degree course. Contributions PS: Conceived the concept, designed the protocol, collected, analyzed & interpreted the data and prepared the manuscript; GS: Guided conduct of the study, critically revised & approved the manuscript; SK: Designed the study methodology, analyzed & interpreted the data and approved the version to be published. Dr. Santhoshkumar A, Professor & Head of Department of Pediatrics, Government Medical College, Thiruvananthapuram will act as guarantor for this paper.
Compliance with Ethical Standards Conflict of Interest None.
Indian J Pediatr
World Health Organization. Dengue: Dengue factsheet. Available at: www.searo.who.int/entity/vector_borne_tropical_diseases/data/ data_factsheet/en/. Accessed 30 April 2017. National Vector Borne Disease Control Programme. Dengue. Dengue cases and deaths in the country since 2010. Available at: www.nvbdcp.gov.in/den-cd.html. Accessed 30 April 2017. Alexander N, Balmaseda A, Coelho ICB, et al. Multicentre prospective study on dengue classification in four south-east Asian and three Latin American countries. Tropical Med Int Health. 2011;16:936–48. World Health Organization. Dengue guidelines for diagnosis, treatment, prevention and control: new edition 2009. Available at: www. who.int/rpc/guidelines/9789241547871/en/. Accessed 30 April 2017. Barniol J, Gaczkowski R, Barbato EV, et al. Usefulness and applicability of the revised dengue case classification by disease: multicentre study in 18 countries. BMC Infect Dis. 2011;11:106. Horstick O, Farrar J, Lum L, et al. Reviewing the development, evidence base and application of the revised dengue case classification. Pathog Glob Health. 2012;106:94–101. Hadenegoro SRS. The revised WHO dengue case classification: does the system need to be modified? Paediatr Int Child Health. 2012;32:33–8. Jain A, Chaturvedi UC. Dengue in infants: an overview. FEMS Immunol Med Microbiol. 2010;59:119–30. Manakkadan A, Joseph I, Prasanna RR, Kunju RI, Kailas L, Sreekumar E. Lineage shift in Indian strains of dengue virus serotype-3 (genotype III), evidenced by detection of lineage IV strains in clinical cases from Kerala. Virol J. 2013;10:37. Farrar JJ, Hien TT, Horstick O, et al. Dogma in classifying dengue disease. Am J Trop Med Hyg. 2013;89:198–201. Moons KGM, Altman DG, Reitsma JB, et al. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): explanation and elaboration. Ann Intern Med. 2015;162:W1–73. Thein TL, Gan VC, Lye DC, et al. Utilities and limitations of the World Health Organization 2009 warning signs for adult dengue severity. PLoS Negl Trop Dis. 2013;7:e2023. Leo YS, Gan VC, Ng EL, et al. Utility of warning signs in guiding admission and predicting severe disease in adult dengue. BMC Infect Dis. 2013;13:498.
Carrasco LR, Leo YS, Cook AR, et al. Predictive tool for severe dengue conforming to World Health Organization 2009 criteria. PLoS Negl Trop Dis. 2014;8:e2972. Jayaratne SD, Atukorale V, Gomes L, et al. Evaluation of the WHO revised criteria for classification of clinical disease severity in acute adult dengue infection. BMC Res Notes. 2012;5:645. Pongpan S, Wisitwong A, Tawichasri C, et al. Development of dengue infection severity score. ISRN Pediatr. 2013;2013:845876. Zhang H, Zhou YP, Peng HJ, et al. Predictive symptoms and signs of severe dengue disease for patients with dengue fever: a meta analysis. Biomed Res Int. 2014:articleID359308. Horstick O, Jaenisch T, Martinez E, et al. Comparing the usefulness of the 1997 and 2009 WHO dengue case classification: a systematic literature review. Am J Trop Med Hyg. 2014;91:621–34. Jaenisch T, Tam DTH, Kieu NTT, et al. Clinical evaluation of dengue and identification of risk factors for severe disease: protocol for a multicentre study in 8 countries. BMC Infect Dis. 2016;16:120. Thanachartwet V, Wattanathum A, Oer-areemitr N, et al. Diagnostic accuracy of peripheral venous lactate and the WHO 2009 warning signs for identifying severe dengue in Thai adults: a prospective observational study. BMC Infect Dis. 2016;16:46. National Vector Borne Disease Control Programme. National guidelines for clinical management of Dengue fever. Available at: http://pbhealth.gov.in/Dengue-National-Guidelines-2014% 20Compressed.pdf. Accessed 23 February 2017. Wills B, Ngoc TV, Van NTH, et al. Hemostatic changes in Vietnamese children with mild dengue correlate with the severity of vascular leakage rather than bleeding. Am J Trop Med Hyg. 2009;81:638–44. Krishnamurti C, Kalyanarooj S. Cutting MA, et al. mechanisms of hemorrhage in dengue without circulatory collapse. Am J Trop Med Hyg. 2001;65:840–7. Halstead SB. Dengue: the syndromic basis to pathogenesis research. Inutility of the 2009 WHO case definition. Am J Trop Med Hyg. 2013;88:212–5. Voung NL, Manh DH, Mai NT, et al. Criteria of Bpersistent vomiting^ in the WHO 2009 warning for dengue case classification. Trop Med Health. 2016;44:14. Samanta J, Sharma V. Dengue and its effects on liver. World J Clin Cases. 2015;16:125–31.