Early hepatitis is the strongest risk factor for the development of severe dengue infection: A points-based risk-score to predict critical disease in dengue fever

Abstract

INTRODUCTION AND OBJECTIVES: Some dengue fever (DF) patients develop plasma leakage [critical-phase (CP)], which may progress to multi-organ failure. We attempted to identify early predictors ofCP in DF. METHODS: This was a retrospective, case-record-based study. Clinical, laboratory features in the first three-days (Dl-3) of illness was used to formulate a risk prediction model (RPM). Patients with serologically confirmed DF, admitted to the University Medical Unit, Teaching Hospital, Ragama, from 01.01.2017-30.06.2017 were included. Patients were randomly assigned to training (TD) and validation datasets (VD) of equal size. Stepwise multivariate logistic regression (p<0.05) was used to identify risk factors in TD. Versions of RPM were compared using Akaike criteria and McFadden's adjusted R2. Coefficients from the best RPM were used to derive weighted risk scores. Best RPM was validated in VD using C-statistic. RESULTS: 697 patients were included (mean-age:34.7±16.1 years, females-48.8%, TD=350, VD=346). CP developed inJ27 (32.6%). Mortality was I%. Risk predictors (p<0.05) were: female gender (OR=2.l), diabetes (OR=l.8), vomiting (OR=l.9), platelets<l20,000/mm-3 (OR=2.8) and AST>60 IU/L (OR=3.3). In multivariate analysis, female gender (score=2), vomiting (score=3), platelets<120,000/mm-3 (score=3) and AST>60 IU/L 3 (score=4) were significant while diabetes was non-significant. Calculated RPM score ranged from 0-12. C-statistic for the TD was 0.78 and VD 0.77 (Hosmer-Lemeshow test: p=0.19 and 0.34 respectively). A cut-off of 5 was selected to maximise sensitivity (0.96), negative predictive value (0.95) with specificity of 0.44. CONCLUSION: This simple risk score seems useful in identifying those at risk ofCP within Dl-3 of the onset ofDF. The early presence of dengue hepatitis was the strongest predictor of CP.