Early clinical determinants of adverse outcomes in paediatric febrile neutropenia: a prospective cohort study.

Background: Febrile neutropenia (FN) remains a major cause of morbidity and mortality in paediatric haemato-oncology, but routinely available determinants of adverse in-hospital outcome remain incompletely characterised. We aimed to identify baseline clinical, haematological, biochemical, and microbiological factors associated with in-hospital mortality and prolonged fever. Methods: We conducted a prospective single-centre cohort study of 90 consecutive FN admissions in paediatric haemato-oncology patients. Thirty-nine prespecified baseline predictors were evaluated using Firth penalised-likelihood logistic regression. Multivariable models were prespecified on clinical grounds, and discrimination was internally validated using 500 bootstrap replicates with optimism correction. Prolonged fever was defined as time to defervescence >5 days, with death before defervescence counted as an event. Results: Of 86 mortality-evaluable episodes, 16 children died in hospital (18.6%). In the multivariable mortality model, any respiratory support (adjusted odds ratio [aOR], 11.9; 95% CI, 2.5-56.9) and lower admission albumin (aOR, 0.29 per 1 g/dL increase; 95% CI, 0.11-0.77) showed the strongest associations with death. Active disease (aOR 3.5, 95% CI 0.87-14.2) and gram-negative bacteraemia (aOR 4.2, 95% CI 0.93-18.9) showed directionally adverse but imprecise associations. The bootstrap-corrected AUC was 0.88. Of 60 classifiable episodes for the secondary outcome, 31 (51.7%) had prolonged fever. Prolonged fever was associated with older age (aOR 1.24 per year, 95% CI 1.05-1.48), lower albumin (aOR 0.31 per 1 g/dL increase, 95% CI 0.12-0.82), any respiratory support (aOR 10.0, 95% CI 0.98-102.0), and gram-negative bacteraemia (aOR 5.5, 95% CI 1.01-30.3). The bootstrap-corrected AUC was 0.85. Neutropenia depth and platelet count were not independently associated with either outcome. Conclusion: In paediatric haemato-oncology patients with FN, adverse in-hospital outcomes were associated mainly with early or evolving respiratory compromise, lower admission albumin, active disease, and gram-negative bacteraemia. Three of these factors, together with age, were associated with prolonged fever or death before defervescence among classifiable episodes. Albumin should be interpreted as a dynamic marker of acute inflammatory burden, capillary leak, and host reserve rather than as a standalone nutritional measure. These routinely available variables may support exploratory bedside risk appraisal, but external validation is required before clinical implementation.