Background The case defi nition for suspected Ebola virus disease is broad, so many negative children are isolated for testing, risking nosocomial infection. We collected data on children admitted to Ebola holding units in Sierra Leone to refi ne the case defi nition and describe outcomes of admitted children.
MethodsAll children aged less than 13 years admitted to 11 Ebola holding units in Sierra Leone between Aug 1, 2014, and March 31, 2015, were eligible for inclusion. Data were collected from paper-based clinical records, district-wide laboratory results, burial records, staff interviews, and follow-up telephone calls. The cohort was split into training and validation datasets. A model was developed with multivariable logistic regression and compared with laboratory results to explore the sensitivity and specifi city of the alternative case defi nition. Findings Of 1054 children admitted, 309 (29%) tested positive for Ebola virus disease and 697 (66%) tested negative (48 [5%] missing). The model had an area under receiver operating characteristic curve of 0•80 (high performance).A case defi nition of Ebola virus disease contact alone, fever (in children >2 years), or fever and conjunctivitis (<2 years) was 94% sensitive and 35% specifi c. Contact, fever, and conjunctivitis, or contact, fever, anorexia, and two of abdominal pain, diarrhoea, or male sex (>2 years) improved specifi city (97%), with sensitivity of 23%. Children testing negative had a case fatality rate of 8% versus 57% in those with Ebola virus disease (p<0•001).Interpretation Contact history, fever, conjunctivitis, abdominal pain, and diarrhoea are key characteristics for diagnosis of paediatric Ebola virus disease. The case defi nitions developed can be used fl exibly-for example, for triage into risk categories to reduce risk of nosocomial infection.