(531) Thresholds of Thrombocytopenia Associated With Mortality in Critically Ill Children

Introduction: Thrombocytopenia is associated with mortality in critically ill children, though optimal thresholds are unclear. This single-center study aims to define platelet count thresholds predictive of mortality in critically ill children with or without oncologic diagnosis or history of hematopoietic stem cell transplant (HSCT).

Methods: The study was IRB approved with a waiver of informed consent. Pts aged < 18 yrs, admitted from 2018 through 2022 were included. Exclusion criteria included: discharged alive from the PICU in < 48 hrs; no platelet count; transfer from outside hospital; or limitation of care order at the time of admission. Electronic health records data were collected electronically with manual chart review validation. Outcome was hospital mortality. Logistic regression models were built for first available or lowest platelet count during the PICU stay in pts with or without oncology diagnosis/ history of HSCT. ROC curves were plotted, and thresholds selected to optimize sensitivity and specificity (Youden’s J statistic).

Results: 4,416 pts were included. 57% male; median age 1.6 [IQR: 0.4, 8.6] yrs. 359 pts had an oncologic diagnosis or history of HSCT (mortality 9.7% vs. 4.9% in non-oncology/HSCT pts). In both cohorts, first available and lowest platelet counts were associated with mortality. In non-oncology/HSCT pts, AUC was 78.4% for lowest vs 64% for first platelet count. Optimal threshold for lowest platelet count was ≤ 197 (sens. 0.8 (0.74, 0.85); spec. 0.66 (0.65, 0.68); for first platelet count was ≤ 247 (sens. 0.57 (0.5, 0.64), spec. 0.68 (0.66, 0.69). For oncology/HSCT pts, AUC was 72.7% for lowest vs 66.3% for first platelet count. Optimal threshold for lowest platelet count was ≤ 120 (sens. 0.88 (0.73, 0.97); spec. 0.53 (0.47, 0.59); for first platelet count was ≤ 118 (sens. 0.65 (0.46, 0.8), spec. 0.69 (0.63, 0.74). Median time to lowest platelet count was 12 [3, 51] hrs for non-oncology/HSCT vs 229 [14, 175] hrs for the oncology/HSCT group, p < 0.001.

Conclusions: Platelet count thresholds associated with mortality in critically ill children likely differ by diagnosis and timing. Lowest platelet count may be more predictive of mortality than admission platelet count. Studies to validate findings and define age and diagnosis-based thresholds are ongoing.