Validation of a Screening Tool for the Early Identification of Sepsis

    loading  Checking for direct PDF access through Ovid



Sepsis is the leading cause of mortality in noncoronary intensive care units. Recent evidence based guidelines outline strategies for the management of sepsis and studies have shown that early implementation of these guidelines improves survival. We developed an extensive logic-based sepsis management protocol; however, we found that early recognition of sepsis was a major obstacle to protocol implementation. To improve this, we developed a three-step sepsis screening tool with escalating levels of decision making. We hypothesized that aggressive screening for sepsis would improve early recognition of sepsis and decrease sepsis-related mortality by insuring early appropriate interventions.


Patients admitted to the surgical intensive care unit were screened twice daily by our nursing staff. The initial screen assesses the systemic inflammatory response syndrome parameters (heart rate, temperature, white blood cell count, and respiratory rate) and assigns a numeric score (0–4) for each. Patients with a score of ≥4 screened positive proceed to the second step of the tool in which a midlevel provider attempts to identify the source of infection. If the patients screens positive for both systemic inflammatory response syndrome and an infection, the intensivist was notified to determine whether to implement our sepsis protocol.


Over 5 months, 4,991 screens were completed on 920 patients. The prevalence of sepsis was 12.2%. The screening tool yielded a sensitivity of 96.5%, specificity of 96.7%, a positive predictive value of 80.2%, and a negative predictive value of 99.5%. In addition, sepsis-related mortality decreased from 35.1% to 23.3%.


The three step sepsis screening tool is a valid tool for the early identification of sepsis. Implementation of this tool and our logic-based sepsis protocol has decreased sepsis-related mortality in our SICU by one third.

Related Topics

    loading  Loading Related Articles