Predictors of mortality in patients with yellow fever: an observational cohort study

Yellow fever virus infection results in death in around 30% of symptomatic individuals. The aim of this study was to identify predictors of death measured at hospital admission in a cohort of patients admitted to hospital during the 2018 outbreak of yellow fever in the outskirts of São Paulo city, Brazil.

In this observational cohort study, we enrolled patients with yellow fever virus from two hospitals in São Paolo-the Hospital das Clínicas, University of São Paulo and the Infectious Diseases Institute "Emilio Ribas". Patients older than 18 years admitted to hospital with fever or myalgia, headache, arthralgia, oedema, rash, or conjunctivitis were consecutively screened for inclusion in the present study. Consenting patients were included if they had travelled to geographical areas in which yellow fever virus cases had been previously confirmed. Yellow fever infection was confirmed by real-time PCR in blood collected at admission or tissues at autopsy. We sequenced the complete genomes of yellow fever virus from infected individuals and evaluated demographic, clinical, and laboratory findings at admission and investigated whether any of these measurements correlated with patient outcome (death).

Between Jan 11, 2018, and May 10, 2018, 118 patients with suspected yellow fever were admitted to Hospital das Clínicas, and 113 patients with suspected yellow fever were admitted to Infectious Diseases Institute "Emilio Ribas". 95 patients with suspected yellow fever were included in the study, and 136 patients were excluded. Three (3%) of 95 patients with suspected yellow fever who were included in the study were excluded because they received a different diagnosis, and 16 patients with undetectable yellow fever virus RNA were excluded. Therefore, 76 patients with confirmed yellow fever virus infection, based on detectable yellow fever virus RNA in blood (74 patients) or yellow fever virus confirmed only at the autopsy report (two patients), were included in our analysis. 27 (36%) of 76 patients died during the 60 day period after hospital admission. We generated 14 complete yellow fever virus genomes from the first 15 viral load-detectable samples. The genomes belonged to a single monophyletic clade of the South America I genotype, sub-genotype E. Older age, male sex, higher leukocyte and neutrophil counts, higher alanine aminotransferase, aspartate transaminase (AST), bilirubin, and creatinine, prolonged prothrombin time, and higher yellow fever virus RNA plasma viral load were associated with higher mortality. In a multivariate regression model, older age, elevated neutrophil count, increased AST, and higher viral load remained independently associated with death. All 11 (100%) patients with neutrophil counts of 4000 cells per mL or greater and viral loads of 5·1 log₁₀ copies/mL or greater died (95% CI 72-100), compared with only three (11%) of 27 (95% CI 2-29) among patients with neutrophil counts of less than 4000 cells per mL and viral loads of less than 5·1 log₁₀ copies/mL.

We identified clinical and laboratory predictors of mortality at hospital admission that could aid in the care of patients with yellow fever virus. Identification of these prognostic markers in patients could help clinicians prioritise admission to the intensive care unit, as patients often deteriorate rapidly. Moreover, resource allocation could be improved to prioritise key laboratory examinations that might be more useful in determining whether a patient could have a better outcome. Our findings support the important role of the virus in disease pathogenesis, suggesting that an effective antiviral could alter the clinical course for patients with the most severe forms of yellow fever.

São Paulo Research Foundation (FAPESP).