Fever in adult ICUs: an interrupted time series analysis*

Fever is common and associated with increased mortality among patients admitted to adult ICUs, yet recent literature suggests that the incidence of fever may be decreasing. The objective of this study was to determine whether the incidence of fever in adult ICUs changed over time and the factors responsible for the observed change.

Interrupted time series analysis. The primary outcome was the cumulative incidence of fever (temperature ≥ 38.3 °C). Secondary outcomes included the cumulative rate of blood cultures ordered, and the cumulative incidence of bloodstream infections and ventilator-associated pneumonia. Data were analyzed with segmented linear regression and adjusted for important confounding variables.

Calgary zone of Alberta Health Services between January 1, 2004, and June 30, 2009.

Adults (age ≥ 18 yr) admitted to ICUs.

None.

There were 18,989 ICU admissions among 17,153 patients. The cumulative incidence of fever during ICU admission decreased from 50.1% of all patients to 25.5% over the 5.5-year study period. Implementation of a new noninvasive thermometer was associated with a 5.1% (95% CI, 1.4-8.9%, p = 0.01) absolute decrease in fever incidence; however, the decrease in fever incidence was predominantly a function of a constant baseline decrease of 1.1% per quarter (95% CI, 0.8-1.5%, p < 0.0001). Multivariate logistic time series regression found that time and thermometer change were the only independent predictors of the changing incidence of fever. The ordering of blood cultures, bloodstream infection incidence, and ICU mortality were unchanged throughout the study period.

The incidence of fever in adult ICUs decreased considerably over time. The lack of change in the ordering of blood cultures and the incidence of bloodstream infections calls into question the importance of fever during the diagnostic evaluation of critically ill patients.