Defining Fever in Critically Injured Patients: Test Characteristics of 3 Different Thresholds
Omeed Sizar, OMS II1; Abid Farooq; Fahim Habib2
1Nova Southeastern University College of Osteopathic Medicine, Fort Lauderdale, Florida; 2Broward General Hospital, Fort Lauderdale, Florida
Introduction: Fever remains the most common sign that prompts the workup for a possible infectious origin in critically injured trauma patients admitted to the intensive care unit. Yet, the definition of fever is highly variable, and the test characteristics of the various cutoff temperatures used have not been clearly defined. An accurate cutoff would allow for more precise and cost-effective management of the febrile trauma patient.
Hypothesis: The commonly accepted standards for fever do not accurately predict an infectious origin in febrile patients.
Methods: Medical records for 621 trauma patients at an urban level-I trauma center were retrospectively evaluated for fever and culture results. The maximum oral temperature during the 24-hour period before obtaining culture samples was used. Temperatures were correlated with positive or negative culture results to determine sensitivity, specificity, positive and negative predictive values, positive and negative likelihood ratios, and area under the curve.
Results: Sensitivity and specificity were calculated using using cutoff values of 100.4F°, 101F°, and 101.5°F. Receiver operator curve cutoffs identified 99.75°F as the temperature with the best test characteristics. Sensitivity showed an inverse relationship with temperature. 99.75°F exhibited a maximum value of 75.30% (CI, 70.27-79.88), with 101.5°F exhibiting the minimum value of 25% (CI, 20.87-29.50). Specificity had a direct relationship to temperature, with 99.75°F having a minimum specificity of 59.46% (CI, 51.00-61.00) and 101.5°F having a maximum specificity of 92.96% (CI, 88.65-96.00). Positive likelihood ratio had a lowest value of 1.86 (CI, 1.51-2.28) at the lowest temperature of 99.75°F, and the highest value of 3.35 (CI, 2.12-5.95) at a temperature of 101.5°F. Negative likelihood ratio was also lowest at 99.75°F with a value of 0.42 (CI, 0.33-0.52), and highest at 101.5°F with a value of 0.81 (CI, 0.75-0.86). Positive predictive value was lowest at a temperature of 99.75°F at 80.46% (CI, 75.57-84.74) and highest at a temperature of 101.5°F at 39.29% (CI, 35.00-43.70). Negative predictive value was highest at 99.75°F with a value of 52.07% (CI, 44.27-59.80) and lowest at 101.5°F with a value of 39.29% (CI, 35.00-43.70). Area under the curve was inversely related to temperature with a maximum value of 0.32 (CI, 0.690-0.774) at 99.75°F and a minimum value of 0.498 (CI, 0.450-0.546) for 101.5°F.
Conclusion: These results suggest that none of the current cutoffs used to define fever accurately predict an infectious origin in febrile patients. Although a temperature of 99.75°F demonstrated the best test characteristics, none of the commonly accepted standards of fever showed a strong correlation to culture results. Further research is warranted to identify biomarkers that accurately identify the presence of infectious processes in trauma patients.