The influence of different fever definitions on diagnostics and treatment after diagnosis of fever in chemotherapy-induced neutropenia in children with cancer.

Background: There is no evidence-based definition of the temperature limit defining fever (TLDF) in children with neutropenia. Lowering the TLDF is known to increase the number of episodes of fever in neutropenia (FN). This study aimed to investigate the influence of a lower versus standard TLDF on diagnostics and therapy.

Methods: In a single pediatric cancer center using a high standard TLDF (39°C tympanic-temperature) patients were observed prospectively (NCT01683370). The effect of applying lower TLDFs (range 37.5°C to 38.9°C) versus 39.0°C on these measures was simulated in silicon.

Results: In reality, 45 FN episodes were diagnosed. Of 3391 temperatures measured, 193 were ≥39.0°C, and 937 ≥38.0°C. For persisting fever ≥24 hours, additional blood cultures were taken in 31 (69%) episodes in reality. This number decreased to 22 (49%) when applying 39.0°C, and increased to 33 for 38.0°C (73%; plus 11 episodes; plus 24%). For persisting fever ≥48 hours, i.v.-antibiotics were escalated in 25 (56%) episodes. This number decreased to 15 (33%) when applying 39.0°C, and increased to 26 for 38.0°C (58%; plus 11 episodes; plus 24%). For persisting fever ≥120 hours, i.v.-antifungals were added in 4 (9%) episodes. This number increased to 6 (13%) by virtually applying 39.0°C, and to 11 for 38.0°C (24%; plus 5 episodes; plus 11%). The median length of stay was 5.7 days (range, 0.8 to 43.4). In 43 episodes with hospital discharge beyond 24 hours, applying 38.0°C led to discharge delay by ≥12 hours in 24 episodes (56%; 95% CI, 40 to 71), with a median delay of 13 hours, and a cumulative delay of 68 days.

Conclusion: Applying a low versus standard TLDF led to relevant increases of diagnostics, antimicrobial therapy, and length of stay. The differences between management in reality versus simply applying 39.0° as TLDF reflect the important impact of clinical assessment.