939. Predicting Attributable Mortality in Pediatric Patients with Cancer Admitted to the Intensive Care Unit for Suspected Infection

Background

Infection and sepsis are important contributors to mortality in children with cancer. Although pediatric risk prediction scores have improved identification of children at high risk of death in the PICU, the value of these tests in immunocompromised children is unknown.

Methods

In this IRB-approved retrospective study performed at St. Jude Children’s Research Hospital, we evaluated the performance of 4 pediatric risk scores, the Pediatric Risk of Mortality (PRISM), Pediatric Sequential Organ Failure Assessment (pSOFA), Quick Sequential Organ Failure Assessment (qSOFA) scores (using data available at 1, 6, 12 and 24 hours) and the Paediatric Index of Mortality 3 (PIM-3) score (at 1 hour), to predict attributable mortality (death ≤ 60 days without organ dysfunction recovery). Inclusion criteria: Age < 24 years, active cancer therapy (other than bone marrow transplantation), and admission to PICU between 2013 and 2019 with suspected infection (collection of a blood culture and initiation of antibiotic therapy). Scores were calculated using the worst value obtained for each variable. Score distributions were compared by the Mann-Whitney U test, and optimal cutoffs selected by maximizing Youden’s index. An unadjusted p-value < 0.05 was considered statistically significant.

Results

Of 202 episodes of PICU admission for suspected infection in 168 participants, there were 12 attributable (6%) and 4 unrelated (2%) deaths. Demographic and cancer-related characteristics were not associated with mortality (Table 1). Of the 4 prediction scores, only the PRISM score at 24 hours was associated with mortality (P = 0.012; Table 2). For PRISM score ≥ 18, sensitivity was 58.3%, specificity was 81.6%, positive predictive value was 16.7%, and negative predictive value was 96.9% for attributable mortality.

Table 1. Risk factors for attributable mortality in pediatric patients with cancer admitted to the intensive care unit with suspected infection.

Conclusion

In children with cancer admitted to PICU with suspected infection, early pediatric risk prediction scores did not predict mortality. The PRISM score calculated at 24 hours did predict mortality but was relatively insensitive. Further research is needed to develop a risk score for immunocompromised children and to validate the 24 hour PRISM score in this population.

Disclosures

Joshua Wolf, MBBS, PhD, FRACP, Karius Inc. (Research Grant or Support) Joshua Wolf, MBBS, PhD, FRACP, Nothing to disclose

2178. Sensitivity of Blood Cultures in Detection of Bacteremia in Febrile Neutropenia

Background

Febrile neutropenia (FN) secondary to bacteremia is a treatable complication of chemotherapy that increases mortality if not promptly recognized and managed.

Methods

The sensitivity of blood cultures collected in pediatric oncology patients with FN was assessed and stratified based on the day of FN episode, culture media type, and the source of blood culture draw at a single US center between 2013 and 2018. Paired aerobic and lytic media bottles were inoculated with each culture draw using a weight-based volume of blood; anaerobic cultures were included with initial cultures starting in September of 2015.

Results

In a retrospective analysis of 10,596 patients, a total of 3,039 episodes of FN were identified. Of the FN episodes, 17.7% had at least one positive blood culture; 84.5%, 1.3%, 0.9% and 13.3% of positive cultures were collected on day 0, day 1, day 2 and ≥ day 3 of a febrile episode. Among the positive day 0 cultures, the median time to detection of an organism was 14.1 hours. Host characteristics of blood culture-positive FN episodes are summarized in Table 1. Bacteremia was identified in 537 FN cases; 18.1%, 11.9% and 2.6% of cultures were positive in only aerobic, lytic or anaerobic media cultures, respectively. The most commonly isolated organisms were Escherichia coli, coagulase-negative Staphylococcus, viridans group streptococcus, Klebsiella pneumoniae and Pseudomonas aeruginosa. Fifteen percent of infectious episodes with a positive blood culture were polymicrobial.

Conclusion

In summary, the study findings have important clinical implications such as emphasizing the value of day 0 cultures and highlighting the importance of routinely collecting blood cultures in more than one media type. Despite an optimized blood culture approach, less than a fifth of FN episodes had a blood culture-based diagnosis.

Disclosures

Randall Hayden, MD, Abbott Molecular: Advisory Board; Quidel: Advisory Board; Roche Diagnostics: Advisory Board.