Identifying Adult Dengue Patients at Low Risk for Clinically Significant Bleeding

Clinically significant bleeding is important for subsequent optimal case management in dengue patients, but most studies have focused on dengue severity as an outcome. Our study objective was to identify differences in admission parameters between patients who developed clinically significant bleeding and those that did not. We sought to develop a model for discriminating between these patients.

We conducted a retrospective study of 4,383 adults aged >18 years who were hospitalized with dengue infection at Tan Tock Seng Hospital, Singapore from 2005 to 2008. Patients were divided into those with clinically significant bleeding (n = 188), and those without (n = 4,195). Demographic, clinical, and laboratory variables on admission were compared between groups to determine factors associated with clinically significant bleeding during hospitalization.

On admission, female gender (p<0.001); temperature >38°C (p<0.001); nausea/vomiting (p = 0.009) and abdominal pain/tenderness (p = 0.005); lower systolic blood pressure (p<0.001); higher pulse rate (p<0.001); increased absolute neutrophil count (ANC; p<0.001); reduced absolute lymphocyte count (ALC; p<0.001), haematocrit percentage (p<0.001) and platelet count (p = 0.04), and increased prothrombin time (p = 0.003) were significantly associated with clinically significant bleeding on univariate analysis. Multivariate analysis showed that independent variables in the final model were female gender (aOR 2.85; 95% CI: 1.9-4.33); temperature >38°C (aOR 1.81; 95% CI: 1.27-2.61), nausea/vomiting (aOR 1.39; 95% CI: 0.94-2.12), ANC (aOR 1.3; 95% CI: 1.15-1.46), ALC (aOR 0.4; 95% CI: 0.25-0.64), hematocrit percentage (aOR 0.96; 95% CI: 0.92-1.002) and platelet count (aOR 0.993; 95% CI: 0.988-0.998). At the cutoff of -3.919, the model achieved an AUC of 0.758 (sensitivity:0.87, specificity: 0.38, PPV: 0.06, NPV: 0.98).

Clinical risk factors associated with clinically significant bleeding were identified. This model may be useful to complement clinical judgement in triaging adult dengue patients given the dynamic nature of acute dengue, particularly in pre-identifying those less likely to develop clinically significant bleeding.