The Need for Risk Stratification Tools in the Pediatric Emergency Department

In this issue of Pediatrics, Romaine et al¹ describe the derivation and validation of a novel risk stratification score, the Liverpool quick Sequential Organ Failure Assessment (LqSOFA) score, in predicting critical care admission for febrile children in a single emergency department (ED) in the United Kingdom. The authors aimed to address the long-standing challenge of identifying potentially critically ill children from the sea of well children with fever who present for pediatric emergency care: identifying the proverbial “needle in a haystack.” Their team proposed the LqSOFA as a rapid, easily implemented score that can be applied during an ED visit to facilitate recognition of ill children, with a particular interest in identifying children with sepsis.

The elements of the LqSOFA were determined by consensus, and the score gives a point each for poor perfusion, altered mental status (measured by the Awake, Verbal, Pain, Unresponsive scale), and heart rate and respiratory rate above the 99th percentile for age. Overall, the authors report a high specificity and negative predictive value of >99% for an LqSOFA score ≥2 but a negative predictive value of only 85% for scores ≥1. The LqSOFA had low sensitivity (39% for an LqSOFA score ≥2 and 72% for an LqSOFA score ≥1). The balance of sensitivity and specificity is particularly tricky in sepsis, in which the cost is immeasurably high for a missed fatal case, yet these outcomes are rare when the common nature of febrile illnesses in children is considered such that purely maximizing the area under the receiver operating characteristic curve may not always be the true goal. Interestingly, the score with the highest area under the receiver operating characteristic curve in this study, the pediatric early warning score, also demonstrated a better balance of sensitivity (87%) and specificity (89%).

Before implementing the LqSOFA, one must consider whether the environment in which it was derived and validated generalizes to a setting, such as an ED in the United States.² As presented, the LqSOFA was studied in a single center in the United Kingdom, with significant differences in comparison with many US children’s hospitals. One particularly noteworthy element was that blood pressure measurement was lacking in a large majority of children (missing in >75% in the derivation cohort, which included only children sick enough to have blood work performed). It may be that in settings that routinely measure blood pressure, such as a US ED,¹ the addition of this element would be feasible and would substantially change the characteristics of the score. However, the proposed LqSOFA may be easier to operationalize in a prehospital setting, for example, where obtaining a blood pressure may be more difficult and less common. In addition, lactate was deemed poorly discriminative, yet it was measured in <4% of the study population. Lactate and blood pressure have been crucial elements of pediatric and adult septic shock definitions and are highly associated with mortality.^3–5

The predictive accuracy of any new score would benefit from comparison with these elements in data sets with lower rates of missing data before implementation.

There are several challenges in deriving scores that may be surmountable by taking advantage of both emerging machine learning techniques and human factors approaches. These include the ability to identify predictive elements and cutoffs in a data-driven fashion as well as to develop dynamic scores using techniques that collect and incorporate minute-to-minute changes in clinical data elements.^6,7 The context and setting in which a score was derived should be considered and may influence its performance in a different setting. Additionally, the importance of human decision-making in score implementation remains an understudied area. Many sepsis alert systems employed in children’s hospitals use a combination of vital signs measurement, physical examination data, and clinician judgment to inform treatment choices.⁸ Further exploration of these human factors will also likely inform and augment performance of both this proposed and future predictive scores. Although the LqSOFA adds to our understanding of pediatric risk for critical illness in the ED, more study is required before it is ready for broad implementation.

ABBREVIATIONS

ED

emergency department

LqSOFA

Liverpool quick Sequential Organ Failure Assessment