52 - A Prediction Model for Serious Bacterial Infections Among Infants in the Emergency Department with Hypothermia

Emergency department (ED) evaluation and management of hypothermic infants represent a challenge for clinicians. There is insufficient evidence to guide the initial evaluation of hypothermic infants who may be at risk for serious bacterial infections (SBI). Appropriate risk stratification of hypothermic infants would allow for decreased testing variability and unnecessary interventions for lower-risk infants.

Objective:

To evaluate risk factors for SBI among hypothermic infants presenting to the ED.

Design/Methods:

We conducted a multicenter case-control study among infants ≤90 days presenting to one of four pediatric EDs within the United States between 2015-2019. Infants were included if they had an ED rectal temperature < 36.5°C and had a blood culture drawn in the ED. Each hypothermic infant found to have SBI was matched with two controls who did not have SBI. Covariables were determined based on the literature describing risk factors for infection in febrile and hypothermic infants and included historical, examination, and laboratory components.

Results:

There were 3,376 hypothermic infants identified. Among 934 (27%) hypothermic infants who had a blood culture drawn, 57 (6.1%) were found to have SBI. Infants with SBI were comprised of 41 (72%) with urinary tract infection (UTI), 8 (14%) with isolated bacteremia, 4 (7%) with UTI and bacteremia, and 4 (7%) with meningitis. In univariable analyses, the following were positively associated with presence of SBI: age >21 days, fever at home, maximum ED temperature ≥38.0°C, white blood cell count ≥12,000 per mm³, absolute neutrophil count (ANC) ≥4,500 per mm³, platelet count ≥328,000 per mm³, and abnormal urinalysis (Table, Figure 1). Prematurity, age ≤21 days, and hypothermia at home were negatively associated with this outcome. The full multivariable model exhibited a raw c-index of 0.960 (95% CI: 0.938, 0.983). Bootstrap backwards selection picked the following variables for a reduced model: abnormal urinalysis, abnormal ANC (≥4,500 per mm³), and minimum ED temperature ≤36°C (C index 0.895). The model calibrated fairly, with an average bootstrap-derived intercept of -0.13 and slope of 0.864 (Figure 2).

Conclusion(s):

Historical, examination and laboratory data show potential as variables to be used in risk stratification of young hypothermic infants. Although our model was a strong fit for SBI prediction, larger studies are needed to definitively risk stratify this cohort.