Early Machine-Human Interface around Sepsis Severity Identification: From Diagnosis to Improved Management?

Abstract

Objective: To investigate the statistical measures of the performance of 2 interventions: a) early sepsis identification by a computerized sepsis "sniffer" algorithm (CSSA) in the emergency department (ED) and b) human decision to activate a multidisciplinary early resuscitation sepsis and shock response team (SSRT).

Methods: This study used a prospective and historical cohort study design to evaluate the performance of two interventions.

Intervention: A computerized sepsis sniffer algorithm (CSSA) to aid in early diagnosis and a multidisciplinary sepsis and shock response team (SSRT) to improve patient care by increasing compliance with Surviving Sepsis Campaign (SSC) bundles.

Results: The CSSA yielded a sensitivity of 100% (95% CI, 99.13-100%) and a specificity of 96.2% (95% CI, 95.55-96.45%) to identifying sepsis in the ED (Table 1). The SSRT resource was activated appropriately in 34.1% (86/252) of patients meeting severe sepsis or septic shock criteria; the SSRT was inappropriately activated only three times in sepsis-only patients. In 53% (134/252) of cases meeting criteria for SSRT activation, the critical care team was consulted as opposed to activating the SSRT resource.

Conclusion: Our two-step machine-human interface approach to patients with sepsis utilized an outstandingly sensitive and specific electronic tool followed by more specific human decision-making.

Keywords: Algorithm; Computerized decision support; Sepsis.