Development of a clinical prediction algorithm for Ebola Virus Disease
Study design
Retrospective analysis
Context
In its early stages, Ebola virus disease rapidly manifests itself as high fever and gastrointestinal symptoms that are virtually indistinguishable from other infectious diseases. Of the more than 60,000 admissions to an Ebola care center during the 10th epidemic in the DRC, 4% tested positive for the Ebola virus. As the testing protocol required two PCR tests 48 hours apart, patients remained in an Ebola health center for at least three days after admission.
To alleviate the burden this represents for Ebola patients and healthcare facilities, Epicentre developed and evaluated the performance and feasibility of a triage algorithm. The tool assessed clinical manifestations of Ebola, such as bleeding gums, coughing and diarrhea, as well as potential patient exposures to Ebola, to identify the most important predictors of Ebola infection.
Our role
Coordination
Partners
- Ministry of Health of the DRC
- INRB
Methods
In this retrospective analysis, we assessed routinely collected data from individuals with possible Ebola virus disease attending 30 Ebola health facilities in two provinces of the Democratic Republic of the Congo between Aug 1, 2018, and Aug 28, 2019. We used logistic regression analysis to model the probability of Ebola infection across 34 clinical variables and four types of possible Ebola virus disease exposures: contact with an individual known to have Ebola virus disease, attendance at any funeral, health facility consultation, or consultation with an informal health practitioner.
Findings
Data for 24 666 individuals were included. If a patient presented to care in the early symptomatic phase (ie, days 0-2), Ebola virus disease positivity was most associated with previous exposure to an individual with Ebola virus disease (odds ratio [OR] 11·9, 95% CI 9·1-15·8), funeral attendance (2·1, 1·6-2·7), or health facility consultations (2·1, 1·6-2·8), rather than clinical parameters. If presentation occurred on day 3 or later (after symptom onset), bleeding at an injection site (OR 33·9, 95% CI 12·7-101·3), bleeding gums (7·5, 3·7-15·4), conjunctivitis (2·4, 1·7-3·4), asthenia (1·9, 1·5-2·3), sore throat (1·8, 1·3-2·4), dysphagia (1·8, 1·4-2·3), and diarrhoea (1·6, 1·3-1·9) were additional strong predictors of Ebola virus disease. Some Ebola virus disease-specific signs were less prevalent among vaccinated individuals who were positive for Ebola virus disease when compared with the unvaccinated, such as dysphagia (-47%, p=0·0024), haematemesis (-90%, p=0·0131), and bleeding gums (-100%, p=0·0035).
Interpretation
Establishing the exact time an individual first had symptoms is essential to assessing their infection risk. An individual's exposure history remains of paramount importance, especially in the early phase. Ebola virus disease vaccination reduces symptom severity and should also be considered when assessing the likelihood of infection. These findings about symptomatology should be translated into practice during triage and should inform testing and quarantine procedures.
