EFB Dengue Literature Review

Dengue Severity Classification

8 min read

Dengue Severity Classification

Overview

Dengue disease severity has been graded by two successive WHO frameworks, and the difference between them is structural, not cosmetic. (The framing in this section is background; the evaluation numbers are reserved for Key Points.)

Traditional WHO (1997). This scheme sorts symptomatic dengue into three tiers:

  • Dengue Fever (DF): an acute febrile illness with ≥2 of: headache, retro-orbital pain, myalgia, arthralgia, leukopenia, rash, hemorrhagic manifestations, supportive serology (or occurrence at the same place/time as confirmed cases).
  • Dengue Hemorrhagic Fever (DHF): requires all four of — (1) fever or history of acute fever lasting 2–7 days; (2) hemorrhagic manifestations (positive tourniquet test; petechiae/ecchymosis/purpura or mucosal/GI/injection-site bleeding; or hematemesis/melena); (3) thrombocytopenia (<100,000 platelets/mm³); and (4) evidence of plasma leakage from increased vascular permeability (pleural effusion, ascites, hemoconcentration ≥20%, or hypoproteinemia). Graded I–IV.
  • Dengue Shock Syndrome (DSS): DHF (grades III–IV) plus hypotension for age or narrow pulse pressure (<20 mmHg) plus signs of shock (rapid weak pulse; cold clammy skin; restlessness).

The defining feature of DHF/DSS is that it is built on the conjunction of four criteria. It is therefore a relatively specific clinical entity — essentially the plasma-leakage syndrome — but a case that is clinically severe yet missing any one criterion (e.g. shock without documented plasma leakage or without thrombocytopenia <100,000) falls outside the DHF/DSS net.

Revised WHO (2009). Prompted by the difficulty of applying DHF at the bedside and by severe cases the 1997 scheme missed, this scheme reorganizes dengue into:

  • Dengue without Warning Signs: fever plus 2 of: nausea/vomiting, rash, aches and pains, leukopenia, positive tourniquet test.
  • Dengue with Warning Signs: dengue plus any of: abdominal pain or tenderness, persistent vomiting, clinical fluid accumulation, mucosal bleeding, lethargy/restlessness, liver enlargement >2 cm, or a laboratory finding (rising hematocrit concurrent with a rapid platelet drop). The warning signs function as triage flags for patients at risk of deterioration.
  • Severe Dengue: dengue with ≥1 of: severe plasma leakage (leading to shock [DSS] or to fluid accumulation with respiratory distress); severe bleeding (clinician-assessed); or severe organ involvement (liver AST/ALT ≥1000; CNS impaired consciousness; heart or other organ failure).

The structural pivot is that 2009 flags severity on any single criterion rather than a conjunction. This makes it far more sensitive as a triage instrument, but also broader — “severe dengue” is no longer coextensive with the specific plasma-leakage syndrome that DHF was built around.

Key Points from Literature

  • Severity findings are classification-scheme-dependent, and cross-scheme comparability is limited — this is the most wiki-relevant point. The two WHO schemes agree only fairly and capture overlapping-but-different patient sets, so a “severity association” reported in one paper cannot be assumed to mean the same thing as one reported under another scheme (see Narvaez2011 - Evaluating WHO Dengue Severity Classifications, n=544 cross-sectional, pediatric). Concretely, the wiki’s dengue sources do not use a common scheme: GodoyLozano2016 and Ansari2025 stratify by WHO-2009 (dengue with/without warning signs, severe dengue), GarciaBates2013 uses Brazil’s national DF/DFC (complicated dengue) criteria, and the WHO-1997 DHF/DSS framing common in the older literature is different again. Severity-stratified findings across these papers are therefore not directly comparable and must each be read relative to the scheme used.
  • Sensitivity/specificity contrast against a clinical gold standard. Using clinical intervention level as the gold standard (Category III = ICU / inotropes / ventilation / organ failure = “severe”), DHF/DSS captured Cat III with sensitivity 39.0%, specificity 75.5%, PPV 43.4%, NPV 71.9%; the 2009 “Severe Dengue” category captured the same Cat III with sensitivity 92.1%, specificity 78.5%, PPV 67.4%, NPV 95.4% (see Narvaez2011 - Evaluating WHO Dengue Severity Classifications, n=544 cross-sectional, pediatric). The revised scheme is dramatically more sensitive and has a high NPV, i.e. it rarely misses a child who will need intensive intervention.
  • Inter-scheme agreement is only fair. The two schemes agree at kappa = 0.25 (fair) on the same cohort — direct quantitative support for the comparability caveat above (see Narvaez2011 - Evaluating WHO Dengue Severity Classifications, n=544 cross-sectional, pediatric).
  • Clinician usability favors the revised scheme. Agreement between clinician judgment and the classification algorithm was kappa 0.46 (moderate) for the traditional scheme — DHF was hard to apply, with only 12.5% of cases classified correctly — versus kappa 0.62 (substantial) for the revised scheme (see Narvaez2011 - Evaluating WHO Dengue Severity Classifications, n=544 cross-sectional, pediatric).
  • Category distributions differ markedly. Traditional: DF 70.8% / DHF 19.5% / DSS 9.7%. Revised: Dengue without warning signs 6.6% / Dengue with warning signs 48.9% / Severe Dengue 44.5% (see Narvaez2011 - Evaluating WHO Dengue Severity Classifications, n=544 cross-sectional, pediatric). The bulk of patients shift into the “with warning signs” and “severe” bins under 2009.
  • Serotype × scheme flip. In the same cohort, DENV-2 was significantly associated with DHF/DSS under the traditional scheme (p<0.001) but there was no significant serotype–severity association under the revised scheme (p=0.104) (see Narvaez2011 - Evaluating WHO Dengue Severity Classifications, n=544 cross-sectional, pediatric). A determinant-of-severity signal can thus appear or vanish purely as a function of which classification is applied.
  • The trade-off, stated plainly. The revised scheme is the better triage tool (high sensitivity and NPV for intervention-level disease) but loses pathogenic-entity specificity — it dissolves the distinct plasma-leakage syndrome into a broad “severe” bin (see Narvaez2011 - Evaluating WHO Dengue Severity Classifications, n=544 cross-sectional, pediatric).
  • Within-scheme definitional heterogeneity — “WHO-2009” is not one operational definition. A PRISMA systematic review of how warning-sign and severe-dengue signs are operationally defined across the WHO-2009 literature found that, of 16 signs, only 2 had a consensus definition across studies: “liver enlargement” (warning sign) and “liver involvement” (severe, = AST/ALT >1000 IU/L in 94.7% of studies) — and both are the signs WHO-2009 already pre-defines. The other 14 vary widely: “shock” alone was defined via 23 distinct parameter-combinations across studies; hematocrit-rise cutoffs ranged >20% vs >15% vs gender-adjusted; platelet thresholds spanned <20k / <50k / <100k / <150k; respiratory-rate cutoffs ran 24 / 30 / 40 / 60 breaths/min (see Morra2018 - Defining Warning Signs and Severe Dengue, systematic review of 44 studies). This compounds the between-scheme caveat above: two papers can both report “WHO-2009 severe dengue” yet have applied materially different operational definitions, so a shared scheme label does not guarantee a shared case definition.
  • Specificity trade-off, independently echoed (attributed). A pediatric cohort comparing the two schemes reported specificity 73.0% for WHO-2009 vs 93.4% for WHO-1997 (Macedo et al, cited in Morra2018 - Defining Warning Signs and Severe Dengue) — consistent with the broader point that the revised scheme trades specificity for sensitivity. Note this is a different cohort and a different gold standard than Narvaez’s intervention-anchored 78.5% specificity (Cat III clinical-intervention level) above; the two numbers are not comparable and the juxtaposition is not a contradiction. Morra performs no diagnostic-accuracy meta-analysis of its own — its contribution is documenting definitional practice, not accuracy statistics.

Contradictions & Debates

The central tension is purpose-dependent and should not be collapsed into a verdict that one scheme is simply “better.”

  • Triage vs. research pull in opposite directions. For clinical management, the 2009 scheme wins: it is more sensitive, has a high NPV, and is easier for clinicians to apply (kappa 0.62 vs 0.46). For pathogenesis research, the same breadth is a liability — by flagging severity on any single criterion, “severe dengue” no longer isolates the specific plasma-leakage syndrome that DHF was designed to capture. Narvaez et al. explicitly argue that studies of viral, host, and immunological determinants of severity need a more specific definition than “severe dengue” to avoid lumping mechanistically distinct presentations together.
  • The serotype example makes the stakes concrete. Because a DENV-2 severity signal is significant under the traditional scheme and absent under the revised one in the same patients, any immunological or virological “severity association” is partly an artifact of the classification chosen, not a free-standing biological fact.
  • Comparability caveat (restated). Mixing DHF/DSS-classified cohorts with WHO-2009-classified cohorts (or with national criteria such as Brazil’s DF/DFC) in a synthesis risks false equivalence: at kappa 0.25 the two WHO schemes are only fairly concordant, so “severe” in one paper and “severe” in another may denote substantially different patient populations. When this wiki juxtaposes severity-stratified flow-cytometry or repertoire findings, the scheme each source used must travel with the finding.
  • Two axes of heterogeneity that stack. The comparability problem operates on two independent levels, and they compound rather than substitute. (1) Between-scheme heterogeneity (Narvaez): WHO-1997 and WHO-2009 agree only at kappa 0.25 and bin different patient fractions, so a finding’s meaning depends on which scheme produced it. (2) Within-scheme heterogeneity (Morra): even two studies both labelled “WHO-2009” may have operationalized the constituent signs differently (e.g., 23 distinct definitions of “shock”; varying hematocrit/platelet/respiratory-rate cutoffs), so a shared scheme label is not a shared case definition. The practical consequence: comparability of any severity-stratified result is limited both by which scheme a study used and by which operational definitions it applied within that scheme. A severity-stratified claim ideally carries both — the scheme (Narvaez) and, where available, the sign definitions (Morra).

Sources

Graph View

Backlinks