Prediction of Sepsis for the Burn Intensive Care Unit Patient

Severely injured burn patients with greater than 30% total body surface area involvement have a high probability of developing sepsis during hospitalization, an event highly correlated with multiple organ failure and death.  Septicemia has been reported as the cause of death in 28 to 65% of fatal burn cases and a two-fold increase in mortality has been associated with a two day delay of antibiotic treatment for blood-stream infection.  The ability to identify and aggressively treat sepsis early in the clinical course may impact outcomes.  However, due to immunocompromise and hypermetabolism, diagnosis of infection in burn patients is difficult.  Widely used sepsis guidelines for critical care patients rely on the systemic inflammatory response score (SIRS), yet SIRS is actually the baseline for the severely burned patient and the definition is not used for this population.  The American Burn Association (ABA) sepsis guidelines incorporate changes in heart rate, respiratory effort, temperature, platelet count, feeding intolerance, and resistance to insulin as measures of sepsis in the burn patient.  Other clinical indicators of infection include hypotension with a requirement for vasopressors and elevated white blood cell count.  A combination of these markers present in burn patients is theorized to be more predictive of sepsis than traditional subjective clinical assessment by the provider. Nursing staff monitor and record hourly vital signs, obtain routine laboratory results, measure serum blood glucose to titrate insulin infusion, and manage continuous enteral feeds.  Taken individually minor changes in clinical parameters may be overlooked; but when evaluated collectively using computer algorithms to detect subtle shifts, the bedside nurse may detect the onset of infection much sooner than a clinician performing daily ICU rounds.  

The purpose of this preliminary study is to use available clinical data collected from medical record review to predict the presence of sepsis in the burned intensive care unit patient using a computerized statistical model. Early prediction is intended to trigger timely clinical intervention, resulting in the ultimate reduction of multiple organ failure or death. A retrospective medical record review will be conducted to collect identified clinical parameters associated with infection in the burn patient. Statistical modeling will be conducted to determine the most specific markers for infection. Validation of the model will be conducted on a different group of patients.

Final report is available on NTRL at: https://ntrl.ntis.gov/NTRL/dashboard/searchResults/titleDetail/ADA615150.xhtml