Relationship Between Oral Temperature and Bacteremia in Hospitalized Patients

Background Early recognition and treatment of bacteremia can be lifesaving. Fever is a well-known marker of bacteremia, but the predictive value of temperature has not been fully explored. Objective To describe temperature as a predictor of bacteremia and other infections. Design Retrospective review of electronic health record data. Setting A single healthcare system comprising 13 hospitals in the United States. Patients Adult medical patients admitted in 2017 or 2018 without malignancy or immunosuppression. Main Measures Maximum temperature, bacteremia, influenza and skin and soft tissue (SSTI) infections based on blood cultures and ICD-10 coding. Key Results Of 97,174 patients, 1,518 (1.6%) had bacteremia, 1,392 (1.4%) had influenza, and 3,280 (3.3%) had an SSTI. There was no identifiable temperature threshold that provided adequate sensitivity and specificity for bacteremia. Only 45% of patients with bacteremia had a maximum temperature ≥ 100.4˚F (38˚C). Temperature showed a U-shaped relationship with bacteremia with highest risk above 103˚F (39.4˚C). Positive likelihood ratios for influenza and SSTI also increased with temperature but showed a threshold effect at ≥ 101.0 ˚F (38.3˚C). The effect of temperature was similar but blunted for patients aged ≥ 65 years, who frequently lacked fever despite bacteremia. Conclusions The majority of bacteremic patients had maximum temperatures below 100.4 ˚F (38.0˚C) and positive likelihood ratios for bacteremia increased with high temperatures above the traditional definition of fever. Efforts to predict bacteremia should incorporate temperature as a continuous variable.