Pediatric Severe Sepsis Prediction Using Machine Learning

Sepsis is a leading cause of death in the United States, with mortality highest among patients who develop septic shock. Early aggressive treatment decreases morbidity and mortality. Although automated screening tools can detect patients currently experiencing severe sepsis and septic shock, none predict those at greatest risk of developing shock. We analyzed routinely available physiological and laboratory data from intensive care unit patients and developed "TREWScore," a targeted real-time early warning score that predicts which patients will develop septic shock. TREWScore identified patients before the onset of septic shock with an area under the ROC (receiver operating characteristic) curve (AUC) of 0.83 [95% confidence interval (CI), 0.81 to 0.85]. At a specificity of 0.67, TREWScore achieved a sensitivity of 0.85 and identified patients a median of 28.2 [interquartile range (IQR), 10.6 to 94.2] hours before onset. Of those identified, two-thirds were identified before any sepsis-related organ dysfunction. In comparison, the Modified Early Warning Score, which has been used clinically for septic shock prediction, achieved a lower AUC of 0.73 (95% CI, 0.71 to 0.76). A routine screening protocol based on the presence of two of the systemic inflammatory response syndrome criteria, suspicion of infection, and either hypotension or hyperlactatemia achieved a lower sensitivity of 0.74 at a comparable specificity of 0.64. Continuous sampling of data from the electronic health records and calculation of TREWScore may allow clinicians to identify patients at risk for septic shock and provide earlier interventions that would prevent or mitigate the associated morbidity and mortality.

Experimental and clinical studies of septic shock support the concept that early resuscitation with fluid and inotropic therapies improves survival in a time-dependent manner. The new American College of Critical Care Medicine-Pediatric Advanced Life Support (ACCM-PALS) Guidelines for hemodynamic support of newborns and children in septic shock recommend this therapeutic approach. The objective of this study was to determine whether early septic shock reversal and use of resuscitation practice consistent with the new ACCM-PALS Guidelines by community physicians is associated with improved outcome. A 9-year (January 1993-December 2001) retrospective cohort study was conducted of 91 infants and children who presented to local community hospitals with septic shock and required transport to Children's Hospital of Pittsburgh. Shock reversal (defined by return of normal systolic blood pressure and capillary refill time), resuscitation practice concurrence with ACCM-PALS Guidelines, and hospital mortality were measured. Overall, 26 (29%) patients died. Community physicians successfully achieved shock reversal in 24 (26%) patients at a median time of 75 minutes (when the transport team arrived at the patient's bedside), which was associated with 96% survival and >9-fold increased odds of survival (9.49 [1.07-83.89]). Each additional hour of persistent shock was associated with >2-fold increased odds of mortality (2.29 [1.19-4.44]). Nonsurvivors, compared with survivors, were treated with more inotropic therapies (dopamine/dobutamine [42% vs 20%] and epinephrine/norepinephrine [42% vs 6%]) but not increased fluid therapy (median volume; 32.9 mL/kg vs 20.0 mL/kg). Resuscitation practice was consistent with ACCM-PALS Guidelines in only 27 (30%) patients; however, when practice was in agreement with guideline recommendations, a lower mortality was observed (8% vs 38%). Early recognition and aggressive resuscitation of pediatric-neonatal septic shock by community physicians can save lives. Educational programs that promote ACCM-PALS recommended rapid, stepwise escalations in fluid as well as inotropic therapies may have value in improving outcomes in these children.