Burden of fever and hospital mortality in patients admitted to the intensive care unit with isolated traumatic brain injury-A retrospective cohort study using continuous temperature data.

Background: Fever has been shown to be associated with poor outcomes in patients with traumatic brain injury. Earlier studies have used peak daily temperature to derive the burden of fever. The association between hospital mortality and fever burden calculated as the area under the temperature-time curve for the entire duration of intensive care unit (ICU) stay has not been studied before.

Objectives: The objective of this study was to investigate the association between the burden of fever and hospital mortality in patients with isolated traumatic brain injury admitted to the ICU.

Methods: We conducted this retrospective cohort study using an electronic database in a tertiary ICU in Sydney. We included all adult patients admitted to the ICU with isolated traumatic brain injury over 3 years from 1 July 2017 to 30 June 2020. We collected data on demographics, clinical characteristics, and interventions for all patients. We defined the burden of fever as an area under the temperature-time curve above 37 °C. The primary outcome was hospital mortality. We used multivariable logistic regression to determine the association between burden of fever and hospital mortality. We assessed the importance of the burden of fever in a predictive model using machine-learning methods (Bagging and Random Forest).

Results: A total of 88 patients (76% males, mean age: 54 ± 23 years, mean Acute Physiology and Chronic Health Evaluation [APACHE] II score: 15 ± 7) were included in the study, and 18 (20.5%) of the 88 patients died in hospital. Compared to survivors, the nonsurvivors had lower mean Glasgow Coma Scale (GCS) score at the scene, higher mean APACHE II and III scores, and higher rates of intracranial pressure monitoring, surgery, mechanical ventilation, use of vasopressors, and cooling. On multivariable logistic regression, age (odds ratio: 1.05, 95% confidence interval: 1.02-1.09, p = 0.01) was found to be an independent predictor of hospital mortality. A higher GCS score at the scene (odds ratio: 0.81, 95% confidence interval: 0.66-0.98, p = 0.03) was associated with survival. The burden of fever was not associated with hospital mortality. The top three important variables in the predictive model were APACHE III, GCS score at scene, and age.

Conclusion: The burden of fever was not an independent predictor of hospital mortality. The results of this study need to be confirmed in a large multicenter study.