Mosquito-borne dengue viral fever is endemic to India. Every year thousands of cases are detected throughout India despite mosquito control measures. ¹ In 2020, India recorded 44585 cases of dengue and in 2021, it increased to 123106 cases. ¹ Dengue progresses from a stage of fever without warning signs to dengue with warning signs and then to severe dengue. About 5-15% of patients with dengue fever progress to severe dengue. ² This febrile illness has the potential to transform into severe dengue and result in mortality, even in young healthy population. ³

It’s clinically difficult to identify those patients with dengue fever who later on progress to severe dengue. ^{3 , 4} World Health Organisation has identified a set of seven clinical and laboratory parameters as warning signs (abdominal pain, persistent vomiting, fluid accumulation, mucosal bleeding, lethargy, liver enlargement, increasing hematocrit with decreasing platelets) which are extensively used in clinical practice for early identification of progression to severe dengue. ² There have been many studies evaluating the diagnostic accuracy of these parameters in dengue fever with mixed results. ^{2 , 5 – 7} In general these parameters have high specificity and high negative predictive value but a lower sensitivity and lower positive predictive value. ^{2 , 7}

Capillary leak is the hallmark of onset of severe dengue. In clinical practice, capillary leak is identified by hemoconcentration, a significant rise in hematocrit, and evidence of fluid collection in cavities. Rise in hematocrit and concurrent significant drop in platelet count is an important warning sign. To identify this parameter, patients are subjected to repeated blood investigations. Availability of good laboratories, manpower, and money for repeating these tests frequently would be a challenge in rural areas. Moreover, interpretation of rising hematocrit requires a baseline hematocrit value which is unavailable in most of the patients when they get admitted in resource-limited settings. ^{8 , 9} Assuming baseline hematocrit with population data is advised when baseline values are not available. ⁴ This is not reliable, because of the wide variability of hemoglobin and hematocrit in third-world countries. It also was noted earlier that many patients did not have a significant rise in hematocrit even when they develop shock in dengue fever. ¹⁰

Capillary leak leads to hypovolemia and varying degree of hypotension. ¹¹ Most of the clinical manifestations of dengue complications can be attributed to capillary leak syndrome. ¹¹ Orthostatic blood pressure changes are commonly used to assess intravascular volume. ¹² Hence, postural fall in blood pressure could be an indicator of significant capillary leak in early part of dengue fever. CDC guidelines and Pan American Health Organisation (PAHO) guidelines include Postural hypotension as one of the warning signs. ^{13 , 14} Clinical criteria used in dengue management at the Queen Sirikit National Institute of Child Health, Bangkok also include postural hypotension as one of the parameters for admission to the hospital. ¹⁵ Postural hypotension is not included as warning sign in WHO guidelines. Conventional warning signs (defined by WHO) have been studied extensively as predictors of severe dengue. However, data regarding postural hypotension in this condition is very limited. Hence this study was planned to evaluate the accuracy of postural hypotension in identifying patients with severe dengue and develop a simple model for identifying the subsequent development of severe dengue.

Clinical and laboratory data of 150 dengue patients admitted to the hospital were analysed. ²² All 150 subjects who were included were available for evaluation till the end of study except for one patient who succumbed to the illness. 74% were men and 26% were women. The mean age of the subjects was 37.9±15.3 years. 23 patients with dengue developed severe dengue as per the World Health Organisation classification. The median duration of fever before admission was four days (IQ range 3–6). There was a statistically significant difference in hemoglobin, haematocrit, and postural fall in the systolic blood pressure (SBP) on admission between the groups which developed severe dengue subsequently and the group which remained in the non-severe category ( Table 1).

Warning signs showed good specificity and positive predictive value for severe dengue, but in general, sensitivity and negative predictive values were low, except for pleural effusion/ascites which had good sensitivity and specificity. Postural fall in SBP also showed higher sensitivity and specificity ( Table 2).

Receiver operator curve (ROC) analysis showed an area under the curve of ROC 0.86; p<0.001 (95% CI: 0.77-0.95) and analysis of the coordinates showed that postural fall of more than 10.33% in systolic BP was the ideal cut-off. Multiple logistic regression analysis revealed that except for postural fall in the SBP on admission and the presence of pleural effusion/ascites, none of the other parameters had a statistically significant adjusted odds ratio ( Table 3).

Hence pleural effusion/ascites and postural fall in the systolic BP were taken as independent parameters for the decision tree analysis. Results of decision tree analysis showed that the model had a very high accuracy of 93.7% (95% CI: 88.9-98.5) for the training data set and 86.7% (95% CI: 76.5-96.9) for the test data set (validation data set). The cross-tabulation of observed and predicted values is mentioned in Table 4.

Patients with less than 10.33% fall on standing SBP were very unlikely to develop severe dengue than those with more than 10.33% fall in the systolic BP. Patients who did not have pleural effusion/ascites were also unlikely to develop severe dengue ( Figure 1).

Specificity, sensitivity, negative likelihood ratio, positive likelihood ratio, negative predictive value, positive predictive value, and diagnostic odds ratio of the model for the entire data set is shown in Table 5.

Three subjects succumbed to the illness and all of them had postural SBP fall of more than 10.33% and had pleural effusion/ascites in the initial screening and were found to have postural SBP fall of more than 18%.

We have studied 150 inpatients with dengue and evaluated the diagnostic performance of warning signs and postural hypotension. Fluid accumulation in cavities and postural fall in SBP of >10.33% were found to have good sensitivity and specificity. Multiple logistic regression showed that other warning signs did not have significant adjusted odds ratio for the development of severe dengue compared to ascites/pleural effusion and postural fall in SBP. Based on decision tree analysis, we developed a model to identify patients with high risk of progression to severe dengue utilizing simple clinical tools (postural hypotension and pleural effusion/ascites). Diagnostic odds ratio of the model is 53.9, making the decision tree a very good diagnostic tool for ruling out the risk of progression to severe dengue. This model performed well both in the development and validation cohort confirming good internal validity.

Severe dengue develops in about 5-15% of patients with dengue fever. ² Capillary leak is a unique and major pathogenetic mechanism in the development of severe dengue. Hence, markers of capillary leak like rising hematocrit and fluid accumulation in cavities have been part of dengue management protocols. WHO has identified seven warning signs for severe dengue: abdominal pain, persistent vomiting, fluid accumulation, mucosal bleeding, lethargy, liver enlargement, increasing hematocrit with decreasing platelets. These markers have been used extensively to triage patients with dengue fever to identify those at higher risk of developing severe dengue. Severe dengue has high mortality of nearly 20% which can be brought down to nearly 1% of patients if identified early and managed appropriately. ⁶ Recently, many studies have been published evaluating the performance of these parameters in predicting severe dengue. In a study published from Malaysia, 700 patients were studied to evaluate the diagnostic performance of warning signs for association with severe dengue. Though specificity was good, sensitivity was suboptimal. Sensitivity of rise in hematocrit was 0.29 which was similar to our study (0.13) and specificity was above 90%. ²

Capillary leak occurs secondary to damage of endothelial cell-to-cell junction. It can occur as a primary event (Clarkson’s disease) or secondary to a variety of infections like dengue or other inflammatory conditions. ¹¹ Capillary leak can lead to intravascular changes (hemoconcentration, hypovolemia- postural hypotension), extravascular changes (fluid accumulation in cavities) and organ involvement (abdominal pain, vomiting, lethargy, liver involvement). Intravascular changes have been typically assessed with hemoconcentration-rise in hematocrit. Importance of this phenomenon was identified around 1970 in Thailand and management protocols were developed by WHO subsequently based on this concept. ¹⁵ However, requirement of basal level to calculate rise in hematocrit is a major challenge in third world countries. Other parameters like postural hypotension would be an appropriate alternative to hematocrit in these situations since sensitivity and specificity for postural hypotension were very high for diagnosing severe dengue.

Evaluation of the rise in hematocrit as a predictor of severe dengue has yielded mixed results. In a meta-analysis of 87 studies consisting of 35,184 dengue fever and 8,173 severe dengue, 34 clinical/biochemical factors were associated with severe dengue out of which 9(including plasma leakage) were relevant within the 7 days window. ⁶ The odds ratio for pleural effusion for association with severe dengue was 15.83, p<001. Hematocrit was strongly associated with severe dengue (standardized mean difference=0.327, 95% CI: 0.019-0.546, p=0.003). ⁶ However, these observations were not confirmed in a meta-analysis involving a larger number of studies. ⁵ Sangkaew et al. performed a systematic review and meta-analysis of studies that focussed on predicting severe dengue using clinical and biochemical parameters during the febrile phase of the illness. ⁵ 150 research papers were included. Hemoconcentration was not found to be significantly different between the group which developed severe dengue compared to the group which remained non-severe. The standardised mean difference was 0.07 (95% CI: -0.11 to 0.26) which was statistically not significant (p= 0.59). ⁵ Moreover, 53.8% of patients with severe dengue did not have hemoconcentration during an epidemic of dengue in Brazil in 2008. ⁷ Difficulties in identifying plasma leakage by clinicians and the need for clinical parameters instead of laboratory parameters, especially in resource-limited settings, were also highlighted by Horstick et al. ⁹

There have also been a few studies in the past attempting to create a decision tree to help identify patients with a risk of progression to severe dengue, earlier on in the illness. ^{19 – 21} These studies either used biochemical parameters, had a lower diagnostic accuracy compared to the decision tree in the current study, or did not have an internal validation. A study by Tamibmaniam et al. tried to find the factors associated with severe dengue infection and to create a decision tree similar to our study. ²⁰ Simple logistic regression analysis found many factors to have the ability to predict severity, but multiple logistic regression narrowed it down to pleural effusion, vomiting, and low systolic blood pressure. Using these variables, they created a decision tree with a sensitivity of 0.81, specificity of 0.54, PPV of 0.16, and NPV of 0.96. Our decision tree has a comparable sensitivity and NPV while having a much higher specificity and PPV. Also, the study by Tamibmaniam et al. did not include an internal validation of their decision tree, unlike in the current study where the internal validation provided by the training sample further validates the results of the test sample.

Dengue patients can be easily and effectively triaged by checking for postural hypotension on admission and daily thereafter. If SBP fall is more than 10.33%, evaluation for pleural effusion/ascites can be done. Patients with pleural effusion/ascites and postural hypotension are the ones who need close monitoring and complete evaluation with a full battery of investigations as per WHO protocol to confirm the high risk of progression to severe dengue. Other patients may not require this aggressive approach since the model has a negative predictive value of about 94%. Postural hypotension represents intravascular changes and ascites/pleural effusion represents extravascular changes.

Identifying patients likely to develop severe dengue is a complex clinical decision-making process. Warning signs defined by WHO help the clinician in this process. No single parameter of these signs is definitive. Findings of this study confirm that postural hypotension is an important and useful warning sign in dengue fever. CDC and PAHO have included postural hypotension as one of the warning signs in their guidelines. ^{13 , 14} WHO guidelines do not include postural hypotension as warning sign. There is a strong case for inclusion of postural hypotension in other guidelines also. Our clinical decision tool could help clinicians in remote places make effective and appropriate clinical judgments when baseline hematocrit is unavailable.