Dengue is the most important arthropod-borne viral disease in humans. The World Health Organization (WHO) has estimated that 1.8 billion people, or more than 70% of the global at-risk population, live in the WHO Southeast Asia and the Western Pacific regions which account for nearly 75% of current global disease burden from dengue [1]. Singapore, a developed island city-state in Southeast Asia has experienced resurgent dengue epidemics since the 1990s after previous decades of vector control effectively reduced the Aedes house index [2]. In addition, dengue infections in Singapore in recent years have shifted from primarily a childhood disease to that of adults [2]. With this resurgence, the classification of dengue cases for hospitalization and clinical management has gained importance.

The WHO 1997 guideline on management of dengue hemorrhagic fever (DHF) differentiates dengue manifestation into dengue fever (DF) and DHF, where plasma leakage underlies the pathophysiology of DHF [3]. This differentiation of dengue disease severity relied on data gathered from pediatric studies. Severe dengue especially in adults not fulfilling DHF criteria are being increasingly recognized [4]. Both Singapore and Puerto Rico reported confirmed adult dengue fatalities not fulfilling DHF and dengue shock syndrome (DSS) classification [5–7]. Persistent vomiting and abdominal pain were identified in a Cuban study as the most frequent warning signs [8], and tachycardia on admission was the only predictor of death in a Singapore case-control study [5]. In 2009, a new WHO guideline emphasized prominent gastro-intestinal symptoms in defining probable dengue and warning signs as predictors of severe dengue, and re-defined severe dengue beyond DHF and DSS [1].

This study describes a large cohort of adult dengue deaths in Singapore confirmed by dengue polymerase chain reaction (PCR) assay and non-structural protein 1 (NS1), and evaluates the utility of the two WHO guidelines 1997 and 2009 in confirmed adult dengue fatalities. This is important to determine the effectiveness of these guidelines in classifying dengue and severe dengue cases.

We performed a multi-center retrospective observational study of adult dengue deaths at all five major adult public hospitals in Singapore from 1 January 2004 to 31 December 2008. Cases were identified through matching positive dengue diagnostic test results with hospital mortality records. Patients tested positive by dengue PCR or NS1 as previously described [9, 10], and died within the same hospital admission, were included. Probable dengue diagnosed with rapid dengue serology was excluded due to lack of specificity [11].

Demographic, epidemiological, co-morbidity, serial clinical and laboratory, radiological, treatment and outcome data were obtained, and principal and secondary discharge diagnoses were recorded. The duration of acute illness was determined from the onset of fever, and from the onset of symptoms preceding hospitalization if fever was absent.

Institutional review boards of National Healthcare Group (Tan Tock Seng Hospital, National University Hospital and Alexandra Hospital) and Singhealth Group (Singapore General Hospital and Changi General Hospital), Singapore, approved the study, which was funded by the National Medical Research Council, Singapore. All records examined were anonymized.

According to the WHO 1997 guideline, DF clinically required the presence of fever and two or more of headache, retro-orbital pain, myalgia, arthralgia, rash, hemorrhagic manifestations and leukopenia; DHF the presence of all of fever, thrombocytopenia ≤100 × 10^9/L, any bleeding, and plasma leakage manifesting as either hematocrit change of ≥20%, clinical fluid accumulation (e.g. pleural effusion or ascites), or hypoproteinemia; and DSS the presence of one of rapid and weak pulse with narrow pulse pressure <20 mmHg, or hypotension for age in a patient with DHF.

Minor modifications to the WHO 2009 criteria were made in this study. Probable dengue was defined as fever and two of nausea or vomiting, rash, aches and pain, leukopenia, and any warning sign. The tourniquet sign was not performed in Singapore. Warning signs were determined as: abdominal pain or tenderness, persistent vomiting on at least two consecutive hospital days, clinical fluid accumulation (pleural effusion or ascites), mucosal bleed, lethargy or restlessness, any hepatomegaly (instead of greater than 2 cm), and hematocrit change of ≥20% with concurrent platelet nadir < 20 × 10^9/L or 50 × 10^9/L (instead of the qualitative rapid change in hematocrit with rapid drop in platelet count).

Severe dengue was present if any one of the following was recorded:

(1) Plasma leakage evidenced by hematocrit change of ≥20%, pleural effusion or ascites, or hypoproteinemia, leading to

(a) DSS (tachycardia, cold and clammy extremities, capillary refill time greater than three seconds, weak or undetectable pulse, narrow pulse pressure ≤20 mmHg, or systolic blood pressure <90 mmHg or unrecordable blood pressure), or

(b) Fluid accumulation with respiratory distress (respiratory rate ≥30/minute, oxygen saturation ≤92% on room air, or mechanical ventilation); or

(2) Severe bleeding manifesting as gastro-intestinal bleeding or menorrhagia, or requirement for transfusion of packed red blood cells or whole blood; or

(3) Severe organ involvement as follows:

(a) Serum alanine or aspartate transaminase ≥1000 units/L;

(b) Impaired consciousness;

(c) Acute renal impairment defined as serum creatinine > two times upper limit of normal [12]; or

(d) Myocarditis or encephalopathy/encephalitis.

For the statistical analysis, Fisher's exact test was applied to detect statistical difference between dichotomous variables, while the Kruskal-Wallis test was used to test for significant trend for continuous data. The analysis was performed in SPSS version 16 (SPSS Inc., Chicago, IL, USA), with the level of significance set at a two-tailed p value of <0.05.

Singapore experienced large dengue epidemics in the past decade affecting predominantly adults [13]. This shift from childhood to adult illness was attributed to lower herd immunity and transmission outside home [2]. Epidemiological data showed adults were at lower risk of DHF than children [2, 14]. Identifying dengue cases, and severe dengue cases among them, are important to allow for targeted clinical management and close monitoring. Difficulties in applying the WHO 1997 DHF criteria reported by many countries led to a new WHO guideline in 2009 focussing on recognizing warning signs and re-defining severe dengue [1, 4, 15, 16]. We found that the diagnostic criteria for probable dengue from the WHO 2009 guideline identified 78.6% of our adult confirmed dengue deaths compared with 32.1% for the WHO 1997 criteria. In addition, throughout the hospital admissions, the WHO 1997 DHF criteria could only be fulfilled in 35.7% of cases, mainly due to the lack of clinical bleeding, while the WHO 2009 criteria classified all cases as severe dengue.

As for the importance of individual warning signs in dengue deaths, a retrospective study in Puerto Rico identified abdominal pain, persistent vomiting, abrupt temperature change and abnormal mental status as four major warning signs [7]. Similarly, in Cuba all twelve fatalities had headache, persistent vomiting, malaise, bleeding and shock [8]. In another Singapore study, abdominal pain, and nausea and vomiting occurred in half of adult dengue deaths [5]. In our study, the presence of any warning sign was the commonest diagnostic criterion at hospital presentation (67.9%) apart from fever, and within seven days of dengue diagnosis (27 [96.4%]). For individual warning signs, lethargy was the commonest at hospital presentation (39.3%) but within seven days of dengue diagnosis, hemoconcentration with concurrent thrombocytopenia <50 × 10^9/L became the commonest (71.4%).

Our study also showed that impaired consciousness increased from 7 (25.0%) at hospital presentation to 16 (57.1%) within seven days of dengue diagnosis. Rigau-Perez et al reported a high proportion of impaired consciousness at presentation and during disease progression [7]. Other studies showed that dengue encephalopathy was diagnosed in 0.5% of children with DHF, and that impaired consciousness was associated with high mortality and multi-organ involvement [17, 18]. The importance in recognizing this sign in potentially severe dengue should be emphasized.

Hemoconcentration was documented in 89.3% of our patients while Cuba reported a similarly high 91.6% [8]. The WHO 2009 criteria lack clarity in quantifying the degree and rate of change in hematocrit and platelet count [1]. The combination of hematocrit change ≥20% and concurrent thrombocytopenia <50 × 10^9/L and <20 × 10^9/L were present in 71.4% and 50.0% of our cases respectively. Our findings provide clinicians with additional quantitative criteria as part of the new WHO warning signs.

Our study also detailed the interval from illness onset to developing warning signs and progression to death. Persistent vomiting ≥ 2 days was the earliest warning sign in our study followed by lethargy and hemoconcentration. Notably, rapid progression to death in our cohort occurred once hemoconcentration with thrombocytopenia developed. In Cuba, vomiting was also one of the commonest warning signs, and deterioration occurred at a mean of 3.75 days from illness onset with total illness duration of 7 days [8]. Hence, clinicians should identify dengue patients with persistent vomiting ≥2 days for close monitoring for hematocrit change ≥20% concurrent with thrombocytopenia <20 × 10^9/L.

There were three earlier reports from Singapore on adult dengue deaths. The age of the deceased in our study and two recent studies [5, 6] was older than a much earlier report by Chan et al [19] with possible shift of severe disease to older population. In our study, 75% had co-morbidities similar to other studies [7, 8]. Asthma, diabetes mellitus and sickle cell disease have been reported risk factors for DHF [20] and await validation in well-designed case control studies. In Cuba white race was prevalent among fatal cases [8], but in multi-racial Singapore, we found no evidence of ethnic preponderance.

Our study included many elderly dengue deaths with a median age of 59 years. Two studies on dengue in older adults suggested fewer constitutional symptoms in these patients [21, 22]. Dengue shock syndrome was an independent risk factor for mortality among the elderly in Taiwan [21]. Multi-organ failure with renal impairment was common in older patients [23]. A study from Taiwan reported that concurrent bacteremia occurred in 7.3% of all study subjects compared with 17.4% in the elderly [21]. Bacteremia was detected in 14.3% in our study, possibly due to older patients and prolonged hospitalization (53.6% of our patients were hospitalized for more than a week and 25% more than two weeks). Recognizing and treating nosocomial infections in complicated dengue infection should therefore be emphasized.

There are some limitations of our retrospective study as it depended on the thoroughness of clinicians' documentation. Documentation on bleeding including petechiae was frequently missing, undermining the utility of WHO 1997 DHF criteria [24]. We did not perform the tourniquet test, which is part of WHO criteria for hemorrhagic tendency. A Vietnamese study found that tourniquet provided additional diagnostic information only in 5% of cases [25]. Information on dengue sero-type and primary versus secondary dengue infection were lacking. However, a prospective study may not be feasible given infrequent adult dengue death, and our study provides a reflection of the possible implications of use of the two WHO criteria, and additional warning signs to be emphasized. Our findings in a highly selected sample of adult dengue fatalities need to be validated in less severe form of dengue diseases in both adults and children. The predominance of older adults in this study raises the need to include older individuals in future vaccine studies.

In conclusion, our findings showed that the WHO 2009 guideline was effective in diagnosing probable and severe dengue in adult dengue deaths. We have also quantified the change in hematocrit and platelet count as part of WHO 2009 warning signs, and have suggested that impaired consciousness may be an important warning sign.

Word count = 2856