Fever in non-neurological critically ill patients: a systematic review of observational studies

There is no recommendation on how increased body temperature should be treated in non-neurological critically ill patients. To understand the epidemiology of fever and its association with mortality, we conducted a systematic review of the literature to search for data related to the association between fever and mortality.

We searched MEDLINE and PUBMED related articles and reference lists from January 1978 to July 2011 to select observational studies for assessment of the association of fever with mortality in non-neurological critically ill patients.

We reviewed 1464 articles and found 9 relevant articles. We found that (1) there is no uniform definition of fever, (2) fever (37.5°C to >39.0°C) was not significantly associated with mortality (odds ratio, 1.22; P = .52), and (3) high fever (39.3°C to 39.5°C) was significantly associated with mortality (odds ratio, 2.95; P = .03). We also found that there has been no multicenter prospective observational study including important confounding factors, such as the use of antipyretic treatments, steroids, and extracorporeal circuits.

The limited evidence available suggests that the recommended definition of fever (38.3°C) might be too low to predict increased mortality. Because fever is common in the intensive care unit, there is an urgent need for more studies in this field.