BACKGROUND: Risk-stratified management of febrile neutropenia (FN) allows intensive management of high-risk cases and early discharge of low-risk cases. Most risk stratification systems predicting severe infection from admission variables have been derived from childhood or adult populations and consequently their value in adolescents/young adults (AYA) may vary. Our objective was to determine their value in this population.
METHODS: Data from the 'predicting infectious complications in children with cancer' (PICNICC) individual participant data collaboration were used to evaluate six previously described risk stratification schema in the AYA population. Complete case analyses were undertaken for five 'paediatric' rules, with imputation for specific missing variables of the 'adult' rule. The predictive performance of the rules or the outcome microbiologically defined infection (sensitivity, specificity and predictive values) were compared.
RESULTS: Among the 5,127 episodes of FN in 3,504 patients in the PICNICC collaboration data set, 603 episodes of FN from 478 patients in 20 studies were of patients 16-25 years old. The six rules demonstrated variable sensitivity (33-96%) and specificity (13-83%). Their overall discriminatory ability was poor (area under the receiver operator curve estimates 0.514-0.593).
CONCLUSIONS: Both paediatric and adult FN risk stratification schema perform poorly in AYA with cancer. An alternative rule or clinical recognition of their limitations is required.