Bayesian analysis of the epidemiology of bleeding in critically ill children

Abstract

Purpose: We updated our findings on the epidemiology of clinically relevant bleeding (CRB) in critically ill children. We also determined the concordance of CRB as defined by the International Society of Thrombosis and Haemostasis, i.e., ISTH definition, and characteristics identified by pediatric intensivists in a recent survey, i.e., survey definition.

Methods: In a prospective cohort study, we included children <18 years old who were admitted to the pediatric intensive care unit for >1 day. We followed them daily for bleeding. Bayesian inference was used as the primary analytic tool to incorporate our prior findings.

Results: Using the ISTH definition, the estimated frequency of CRB was 10.0% (95% credible interval, CrI: 7.6%, 12.8%) from 41 of 405 children who had CRB. The estimated frequency from 4 of 12 adolescents >13 years old who received mechanical ventilation or vasopressor support and had CRB was 32.9% (95% CrI: 12.0%, 58.8%). Using the survey definition, the estimated frequency of CRB for the entire cohort was 10.8% (95% CrI: 8.3%, 13.8%). Concordance between definitions for each bleeding event was 0.40 (95% confidence interval: 0.27, 0.52).

Conclusions: Our updated findings highlight the high frequency of CRB regardless of definition used for CRB.

Keywords: Adolescent; Intensive care unit; Mechanical ventilation; Risk factor; Transfusion.

Figure 1.

Posterior distributions of the estimated frequency of clinically relevant bleeding for the entire cohort (A and C) and among adolescents who received mechanical ventilation or vasopressor support (B and D). Clinically relevant bleeding was categorized using the definition by the International Society of Thrombosis and Haemostasis (A and B) or the characteristics identified by pediatric intensivists in a prior survey (C and D). The estimated frequency of clinically relevant bleeding in the current study (Greenway) was compared with previous studies based on age and definition of clinically relevant bleeding. Each curve represented the posterior distribution of the estimated frequency of clinically relevant bleeding of each study. A total of 4 studies, including the current study, was available for comparison for the entire cohort (A and C), but only 2 studies were available for adolescents (B and D).

Figure 2.

Concordance in the categorization of 174 bleeding events (A) and 100 children who had a bleeding event (B) using the definition by the International Society of Thrombosis and Haemostasis (ISTH Definition) or the characteristics identified by pediatric intensivists in a prior survey (Survey Definition). The number in each cell represents counts of bleeding events (A) or children (B). A total of 44 clinically relevant and 77 minor bleeding events were concordantly categorized using the ISTH and survey definitions (A). In terms of children, 34 were concordantly identified as having had a clinically relevant bleeding and 48 were concordantly identified as having had minor bleeding only (B).