Clinical diagnosis of infection in SICU: You are not as good you think!
Author(s):
Madhu Subramanian, UTSouthwestern; Sara Hennessy, University of Virginia; Malcolm MaCConmara, UT Southwestern Medical Center; Carol Hirschkorn, Parkland Hospital; Joseph Minei, UT Southwestern Medical Center; Robert Sawyer, University of Virginia HSC; Christian Minshall, UT Southwestern Medical Center; Tjasa Hranjec, UT Southwestern Medical Center
Background: The National Healthcare Safety Network and Centers for Medicare and Medicaid Services (CMS) require reporting of hospital-acquired infections. Accurate identification of infections by surgical intensivists is of great importance for patient care, monitoring and hospital reimbursement.
Hypothesis: We hypothesized that physicians are able to clinically differentiate between infected and non-infected patients leading to appropriate treatment.
Methods: Between 12/13-07/14, all potentially infected patients in surgical ICU were included in our prospective study. When patients were suspected of having an infection based on clinical grounds, cultures were obtained. Three most senior physicians were required to independently answer a questionnaire about possible infectious diagnosis, etiology of disease, and indications for treatment with antibiotics. Questionnaire and culture data were compared.
Results: Fifty-two critically ill patients presented with signs and symptoms of infection, with 156 (3 per patient) independent physician evaluations generated. Physicians correctly diagnosed patients as either infected or non-infected 53% of the time. Percent of overall agreement between reviewers on the correct diagnosis was poor (kappa= 0.33). Based on objective evidence, all 3 physicians were correct 27% of the time and incorrect 23% of the time. In infected patients, physicians correctly predicted the source of infection 60% of the time. In non-infected patients, physicians predicted negative culture results 36% of the time (Sensitivity=59.3%, Specificity=64.4%, PPV=80%, NPV=39.7%). For patients with more than one infection, a second source of infection was correctly diagnosed 26.7% of the time. Physicians were most accurate in predicting ventilator associated pneumonia, followed by blood stream and intra-abdominal, then urinary tract infections. Based on our ICU antibiotic protocols and physician action on culture results, antibiotics were appropriately initiated 78% of the time - 83% of the time in infected patients. When treatment was determined solely based on clinical judgment, antibiotics were initiated correctly 50% of the time, 39% in infected patients.
Conclusions: Even in the hands of the most experienced clinicians, infected patient can be accurately identified 50% of the time and furthermore there is poor agreement between clinicians. In the critically ill patient, an infectious diagnosis should always be accompanied by obtaining cultures, since clinical diagnosis is difficult and often inaccurate.