Background: Sepsis is a leading cause of morbidity and mortality among children, with early recognition and treatment impacting long-term health outcomes. While studies show screening interventions embedded in electronic health records can improve outcomes, evaluating such interventions is challenged by the lack of a universal reference standard of pediatric sepsis. Objective: Evaluate the test characteristics of an electronic sepsis screening algorithm (Figure 1) implemented in a pediatric emergency department (PED) using two reference standards for defining pediatric sepsis Design/Methods: Evaluation of a high-risk pediatric cohort treated in a PED, March-November 2017. High-risk was defined as patients with 1 or more of: 1) PED intravenous antibiotic, normal saline bolus, and blood culture, 2) PED vasopressor, 3) PED sepsis bundle orders, 4) positive sepsis screen, 5) death or ICU admission within 24 hrs, or 6) sepsis ICD10 diagnosis. High-risk cohort patient records were manually reviewed for systematic inflammatory response syndrome (SIRS) criteria during PED stay; if SIRS criteria met, indicators of organ dysfunction were abstracted. Primary outcomes included development of sepsis and/or septic shock within the first 2 days of care, adapted from Goldstein International Sepsis and Pediatric Sequential Organ Failure Assessment (pSOFA) criteria. Test characteristics of the screening algorithm for these outcomes were evaluated. Results: During the study period, 1489 children met the high-risk cohort definition; 82% were admitted, and 1% died. Of this cohort, 171 (11%) met the Goldstein criteria for sepsis, 430 (29%) for pSOFA sepsis (difference of 18%, 95% CI 15-21%); 126 (8%) met Goldstein septic shock criteria, and 15 (1%) for pSOFA septic shock (difference 7%, 95% CI 6-9%). There were no significant differences in screening intervention sensitivity or specificity defined by Goldstein or pSOFA criteria (Table 1). The likelihood ratios for the screen in determining Goldstein septic shock were significantly better than in determining pSOFA septic shock (LR+ 1.74 (1.60, 1.88) vs. 0.91 (0.83, 1.00). Conclusions: Reference standard definition impacted the population defined with sepsis and septic shock with a higher proportion of patients meeting criteria for pSOFA sepsis compared to Goldstein sepsis, but lower proportion for pSOFA shock than Goldstein shock. These differences in outcome definition impacted the test characteristics of the screening measure and influence interpretation of screening accuracy.

Figure 1

Electronic Sepsis Detection Algorithm BP = blood pressure. Abnormal capillary refill defined as flash (<1 second) or delayed (>3 seconds). Altered mental status defined as agitated, inconsolable, drowsy, lethargic, listless, or non-responsive to stimulation.

Figure 1

Electronic Sepsis Detection Algorithm BP = blood pressure. Abnormal capillary refill defined as flash (<1 second) or delayed (>3 seconds). Altered mental status defined as agitated, inconsolable, drowsy, lethargic, listless, or non-responsive to stimulation.

Close modal
Table 1

Test Characteristics of PED Sepsis Screening Algorithm by Two Outcome Measures PPV = positive predictive value. NPV = negative predictive value. LR+ = positive likelihood ratio. LR- = negative likelihood ratio. *Indicates significant differences (p < .05) between pSOFA and Goldstein test characteristics for either sepsis or shock outcomes.

Table 1

Test Characteristics of PED Sepsis Screening Algorithm by Two Outcome Measures PPV = positive predictive value. NPV = negative predictive value. LR+ = positive likelihood ratio. LR- = negative likelihood ratio. *Indicates significant differences (p < .05) between pSOFA and Goldstein test characteristics for either sepsis or shock outcomes.

Close modal