- Email: [email protected]
Epidemiology of Bleeding in Critically Ill Children
ARTICLE IN PRESS THE JOURNAL OF PEDIATRICS • www.jpeds.com
ORIGINAL ARTICLES
Epidemiology of Bleeding in Critically Ill Children Lauren J. White, MD1,*, Ryan Fredericks, MD1,*, Candace N. Mannarino, MD1, Stephen Janofsky, MD1, and Edward Vincent S. Faustino, MD, MHS2 Objective To determine the epidemiology of bleeding in critically ill children. Study design We conducted a cohort study of children <18 years old admitted to the pediatric intensive care
unit for >24 hours and without clinically relevant bleed (CRB) on admission. CRB was defined as resulting in severe physiologic derangements, occurring at a critical site or requiring major therapeutic interventions. Using a novel bleeding assessment tool that we developed, characteristics of the CRB were abstracted from the medical records independently and in duplicate. From the cohort, we matched each child with CRB to 4 children without CRB based on onset of CRB. Risk factors and complications of CRB were identified from this matched group of children. Results We analyzed 405 children with a median age of 35 months (IQR 7-130 months). A total of 37 (9.1%) children developed CRB. The median number of days with CRB was 1 day (IQR 1-2 days). Invasive ventilation (OR 61.35; 95% CI 6.27-600.24), stress ulcer prophylaxis (OR 2.70; 95% CI 1.08-6.74), surgical admission (OR 0.29; 95% CI 0.10-0.84), and aspirin (OR 0.04; 95% CI 0.002-0.58) were associated with CRB. CRB was associated with longer time to discharge from the unit (hazard ratio 0.20; 95% CI 0.13-0.33) and the hospital (hazard ratio 0.49; 95% CI 0.33-0.73). Children with CRB were on vasopressor longer and transfused more red blood cells after the CRB than those without CRB. Conclusions Our findings suggest that bleeding complicates critical illness in children. (J Pediatr 2017;■■:■■-■■). leeding complicates critical illness in adults.1 Approximately 6%-20% of adults admitted to the intensive care unit (ICU) develop major bleed, which is a clinically relevant bleed (CRB) resulting in severe physiologic derangements, occurring at a critical site or requiring major therapeutic intervention.1,2 An additional 8%-70% of them develop a minor bleed, which is an overt bleed that is less severe than a major bleed. Decrease in platelet count, prolongation of activated partial thromboplastin time, heparin at therapeutic doses, antiplatelet agents, renal replacement therapy, and recent surgery are associated with a major bleed. Compared with those with no bleeds, critically ill adults with major bleeds are transfused with larger amounts of blood products, stay longer in the ICU, and have higher risks of mortality. Bleeding may also complicate critical illness in children. However, the epidemiology (ie, incidence, risk factors and complications) of bleeding in these children is unclear. Studies have focused on selected children admitted to the ICU with illnesses or therapies that increase bleeding risk. These studies report that 3%-52% of the children bleed while admitted.3-9 The uncertainty in the estimates is partly due to inconsistencies in the definitions used for bleeding. Recently, the International Society on Thrombosis and Hemostasis (ISTH) defined categories of bleeding to standardize outcomes in clinical trials of anticoagulation.10 Aside from major and minor bleeds, as previously defined, the category of clinically relevant nonmajor bleeds was added for bleeding that does not fulfill the criteria for a major bleed but required transfusion of blood products, or medical or surgical intervention other than in an operating suite to restore hemostasis. In this study, we developed a novel bleeding assessment tool based on the definitions recommended by ISTH. Using this tool, we aimed to determine the epidemiology of bleeding in an unselected cohort of critically ill children.
B
Methods The study was composed of 2 phases. In phase 1, we developed the bleeding assessment tool, and then pilot tested it in a crosssectional study. In phase 2, we performed a cohort study to determine the incidence of bleeding. Risk factors and complications associated with bleeding were identified in a nested case-control study. The study was approved by the Human Investigation Committee at Yale University, which waived the need for signed consent. From the 1Department of Pediatrics, Yale-New Haven Children’s Hospital, New Haven, CT; and 2Department of Pediatrics, Yale School of Medicine, New Haven, CT *Contributed equally.
CRB HR ICU ISTH
Clinically relevant bleed Hazard ratio Intensive care unit International Society on Thrombosis and Hemostasis
Supported in part by the American Heart Association (14CRP20490002 to Dr. E.F.). The authors declare no conflicts of interest. 0022-3476/$ - see front matter. © 2017 Elsevier Inc. All rights reserved. http://dx.doi.org10.1016/j.jpeds.2017.01.026
1 FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
THE JOURNAL OF PEDIATRICS • www.jpeds.com Phase 1 Development of the Bleeding Assessment Tool. One of the investigators created the electronic tool that contained the definitions recommended by ISTH for major, clinically relevant nonmajor, and minor bleed (Appendix; available at www.jpeds.com). Items used to categorize a bleed were described in sufficient detail in the tool to maximize consistency in its use. The tool also contained the site and timing of the bleed. It was revised after discussion among all investigators. Pilot Testing of the Bleeding Assessment Tool. Using an electronic patient database, children admitted to the pediatric ICU at Yale-New Haven Children’s Hospital from January 1, 2013, to December 31, 2013, with a diagnosis of a bleed were randomly selected. The investigators blindly and independently reviewed the medical records of 10 children per cycle. The first bleed for every child was categorized using the bleeding assessment tool. In between cycles, the entire investigative team discussed the cases and resolved discrepancies. The process was repeated until the chance-corrected inter-rater agreement (k) was ≥0.80. Phase 2 Subjects. A cohort of children less than 18 years old who were admitted to the pediatric ICU at Yale-New Haven Children’s Hospital from January 1, 2014, to December 31, 2014, were randomly selected using the electronic patient database. Excluded were those who stayed in the ICU for less than 24 hours because they were mostly admitted for observation. For those with multiple admissions, 1 admission was randomly selected. Children who had a CRB, whether major or nonmajor, upon admission were excluded after review of medical records. We analyzed major and nonmajor CRBs together because both categories, by definition, are clinically significant and have the potential to impact the child’s health and outcome. Each child with a CRB (case) was randomly matched to 4 children without a CRB (control). The day of stay in the ICU when the CRB first occurred in cases were matched to a similar time point in controls (match day). Matching based on time leads to complete adjustment for the effect of duration of stay in the ICU and its determinants.11 Procedures. The medical records of each eligible child were blindly and independently reviewed by 2 randomly assigned investigators. The bleeding assessment tool was completed by each investigator for each bleed. A third randomly assigned investigator adjudicated discrepancies in the categorization of the bleed. The investigator primarily assigned to each child also collected from the medical records data on demographics; medications, interventions, and worst laboratory tests at any time while admitted to the ICU; and, outcomes. Collection of daily data was censored at 14 days, which represented the 95th percentile of the duration of stay in our ICU. Predicted risk of mortality was calculated based on the Pediatric Index of Mortality 2.12 Anticoagulants included unfractionated heparin
Volume ■■ except at doses to maintain patency of a vascular catheter, low molecular weight heparin, and warfarin, whereas antifibrinolytics included tranexamic and aminocaproic acid. Stress ulcer prophylaxis included proton pump and histamine H2 receptor inhibitors. Vasopressor use was defined as dopamine ≥5 mcg/kg/min, or any dose of dobutamine, epinephrine, norepinephrine, phenylephrine, milrinone, and vasopressin. Respiratory support was categorized as invasive ventilation via endotracheal tube or tracheostomy, noninvasive ventilation (ie, continuous or bilevel positive airway pressure, or high flow nasal cannula greater than 5 L/min for children less than 10 years old and greater than 8 L/min for children at least 10 years old) or none, which included any support that did not qualify for invasive or noninvasive ventilation. Extracorporeal support included extracorporeal membrane oxygenation, renal replacement therapy, and erythro- or leukapheresis. Cryoprecipitate was included in plasma transfusion. Statistical Analyses In phase 1, crude inter-rater agreement and k were calculated. Bleeds were dichotomously categorized based on the presence of a CRB in calculating k. Furthermore, agreement was defined as a majority of investigators assigning the same category. In a sensitivity analysis, k was calculated based on the presence of any bleed. In phase 2, the incidence of bleeding was expressed as proportion of children and days in the ICU with bleed. Conditional logistic regression with stepwise backward elimination at an exit threshold of P > .05 was used to identify factors associated with a CRB. Data on potential risk factors were limited to a specific number of days prior to match day. When applicable, the time period was based on pharmacokinetics (ie, 7 days for aspirin, 5 days for warfarin, and 1 day for unfractionated heparin, low molecular weight heparin, tranexamic acid, aminocaproic acid, vasopressor, and heparin for extracorporeal support). For others, it was arbitrarily set at 2 days based on consensus among investigators. Multiple imputations using multivariate normal distribution was used for missing laboratory data. Times to outcome, adjusted for age, and predicted risk of mortality, were compared between cases and controls using Cox regression with shared frailty model to account for the matching. Number of days on vasopressors and volume of blood products transfused from match day were compared between cases and controls using linear mixed effects model with matching as random effect. Use of vasopressors or blood product transfusion prior to match day, and duration of stay in the ICU from match day were entered as fixed effects. Data was expressed as median (IQR), mean difference (95% CI), count (percentage), OR (95% CI), and hazard ratio (HR; 95% CI). In the absence of reliable estimates of the proportion of critically ill children with bleed, 405 children were analyzed assuming a true proportion of 0.50. This provided the largest sample size for a 95% CI of ±0.05. All analyses were performed in Stata v 14 (StataCorp, College Station, Texas). A 2-sided P value of <.05 was considered statistically significant.
2
White et al FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
■■ 2017
ORIGINAL ARTICLES Results
Bleeding Assessment Tool A total of 3 cycles were performed before k ≥ 80 was achieved. In the presence of a CRB, the crude inter-rater agreement increased from 0.80 to 0.90 to 1.00, whereas k increased from 0.48 to 0.67 to 1.00. In between cycles, specific scenarios in which discrepancies occurred were discussed, which improved consistency in coding. No revisions on the tool were required. The overall k for the 30 children was 0.71. When applied to the cohort of 405 children, the crude inter-rater agreement was 0.92 but k was only 0.47. In the presence of any bleed, the crude inter-rater agreement also increased from 0.80 to 0.90 to 1.00 in phase 1. The k at each cycle was 0.25, 0.18, and 1.00. For the entire cohort, the crude inter-rater agreement was 0.81 with k of 0.57. Incidence of Bleeding There were 1069 potentially eligible admissions, of which 508 were hierarchically excluded for duration of stay in the ICU
of less than 24 hours (n = 366), multiple admissions (n = 119), and age on admission of at least 18 years (n = 23). Of the remaining 561 admissions, medical record review of the first randomly selected 450 excluded 45 because of the presence of a CRB on admission to the ICU. The most common causes of a CRB on admission were trauma (n = 27) and surgery (n = 10). The mean age of the final cohort of 405 children was 35 months (IQR 7-130 months) (Table I). The predicted risk of mortality was 0.008 (IQR 0.002-0.015). Only 117 (28.9%) were admitted after surgery. The median durations of stay in the ICU and the hospital were 2.0 days (IQR 1.4-3.3 days) and 4 days (IQR 3-8 days), respectively. A total of 4 (1.0%) died while admitted to the hospital. A total of 151 (37.3%) children had 197 bleeding events in 305 days while admitted to the ICU (Table I). Of those with bleeds, most had only 1 event (79.6%) (Figure 1; available at www.jpeds.com). The worst bleed was a CRB in 37 children (major in 30 and nonmajor in 7) for an incidence of 9.1%, and minor in 114 children (28.1%). Of the 197 bleeding events,
Table I. Characteristics of children analyzed* Cohort (n = 405) Demographics Age, mo Male Diagnostic category Cardiovascular Endocrinologic Gastrointestinal Genetic Infectious Injury/poisoning Neurologic Oncologic Orthopedic Respiratory Others Predicted risk of mortality Surgical admission Laboratory tests‡ Lowest hemoglobin Lowest platelet count Highest international normalized ratio Highest activated partial thromboplastin time Medications Antithrombotic Antifibrinolytic Aspirin Steroid Nonsteroidal anti-inflammatory drug Stress ulcer prophylaxis Vasopressor Transfusions Red blood cell Plasma Platelets Other interventions Worst level of respiratory support None Noninvasive Invasive Extracorporeal organ support
35 (7-130) 229 (56.5) 66 24 11 18 15 18 42 19 14 138 40 0.008 117
(16.3) (5.9) (2.7) (4.4) (3.7) (4.4) (10.4) (4.7) (3.5) (34.1) (9.9) (0.002-0.015) (28.9)
10.8 242 1.2 26.7
(9.2-12.4) (157-345) (1.0-1.3) (24.0-34.6)
44 31 33 161 126 207 53
(10.9) (7.7) (8.2) (39.8) (31.1) (51.1) (13.1)
81 (20.0) 27 (6.7) 16 (4.0)
Cases (n = 37) 56.4 (3-102) 25 (67.6) 8 0 1 3 3 0 5 0 2 9 6 0.042 11 9.7 181 1.3 32.3 2 2 1 14 10 24 13
(21.6) (0) (2.7) (8.1) (8.1) (0) (13.5) (0) (5.4) (24.3) (16.7) (0.008-0.039) (29.7) (7.7-11.6) (108-257) (1.1-1.6) (24.9-40.7) (5.4) (5.4) (2.7) (37.8) (27.0) (64.8) (35.1)
Controls (n = 148) 64.9 (6-135.5) 81 (54.7) 27 7 4 8 6 7 12 7 2 56 12 0.019 40 11.1 259 1.1 26.9 18 5 12 49 36 60 13
P value† .53 .15 .06
(18.4) (4.7) (2.7) (5.4) (4.1) (4.7) (8.1) (4.7) (1.4) (37.8) (8.1) (0.002-0.014) (27.0)
.07 .74
(9.4-12.5) (163-351) (0.8-1.3) (18.0-36.8)
.005 .04 .07 .15
(12.2) (3.4) (8.1) (33.1) (24.3) (40.5) (8.8)
.23 .56 .25 .58 .74 .01 <.001
19 (51.4) 5 (13.5) 6 (16.2)
15 (10.1) 4 (2.7) 3 (2.0)
<.001 .02 .003
1 3 33 2
46 40 62 4
<.001 113 94 198 16
(27.9) (23.2) (48.9) (4.0)
(2.7) (8.1) (89.2) (5.4)
(31.1) (27.0) (41.9) (2.7)
.42
*Data is expressed as median (IQR) or count (percentage). †P values refer to comparisons between cases and controls. ‡Multiple imputations were used to analyze the laboratory tests.
Epidemiology of Bleeding in Critically Ill Children FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
3
THE JOURNAL OF PEDIATRICS • www.jpeds.com Table II. Sites of bleeding* Clinically relevant (n = 47) Respiratory including thoracic cavity Surgical site Vascular catheter or insertion site Gastrointestinal Other skin bleeding not catheter-related Genitourinary Central nervous system Pericardium Ear/eye/nose/mouth
24 12 4 3 1 1 1 1 0
(51.1) (25.5) (8.5) (6.4) (2.1) (2.1) (2.1) (2.1) (0)
Minor (n = 150) 0 104 11 10 10 8 0 0 7
(0) (69.3) (7.3) (6.7) (6.7) (5.3) (0) (0) (4.7)
*Data are expressed as counts (percentages).
47 (23.9%) were CRBs (36 major and 11 nonmajor bleeds) and 150 were minor. Among children with any bleed, the median number of days in the ICU that a child had a bleed was 2 days (IQR 1-2 days). The median number of days with CRB was 1 day (IQR 1-2 days), whereas the median time to the first CRB was 2 days (IQR: 1-3 days). The most common site for a CRB was the respiratory system (51.1%) and that for minor bleeds was the surgical site (69.3%; Table II). Risk Factors for Clinically Relevant Bleed The 37 children with a CRB were matched to 148 children without a CRB. Cases had lower hemoglobin and platelet count prior to the CRB than their controls (Table I). Although only 125, 123, 65, and 65 matched children had values for hemoglobin, platelet count, international normalized ratio, and activated partial thromboplastin time, respectively, the comparisons were similar with and without multiple imputation. The proportion of cases on stress ulcer prophylaxis, vasopressor, transfused with red blood cells, plasma or platelets, or invasive ventilation prior to the CRB were higher than that of controls. There was no significant difference in the propor-
Volume ■■ tion of children with a minor bleed prior to the CRB between cases and controls. In the multivariable analysis, invasive ventilation (OR vs no respiratory support: 61.35; 95% CI 6.27600.24) and stress ulcer prophylaxis (OR 2.70; 95% CI 1.086.74) were associated with increased risk of CRB, whereas surgical admission (OR 0.29; 95% CI 0.10-0.84) and aspirin (OR 0.04; 95% CI 0.002-0.58) were associated with decreased risk (Table III). Complications of Clinically Relevant Bleed Children with a CRB had longer times to discharge from the ICU (HR 0.21; 95% CI 0.13-0.33) and from the hospital (HR 0.49; 95% CI 0.33-0.73) after adjusting for age and predicted risk of mortality (Figure 2, A and B). The median length of stay in the ICU was 4.9 days (IQR 2.1-9.6 days) for cases and 2.1 days (IQR 1.5-4.0 days) for controls, whereas the median length of stay in the hospital was 8 days (IQR 6-25 days) for cases and 4 days (IQR 2-9 days) for controls. Mortality was not significantly different between cases and controls (HR 4.76; 95% CI 0.40-56.58). After adjusting for use of vasopressor prior to match day and remaining days in the ICU, cases were on vasopressors longer than controls by 0.4 days (95% CI 0.20.6 days). Cases also received more red blood cells (mean difference: 5 mL/kg; 95% CI 1-9 mL/kg), but not plasma (mean difference 0.2 mL/kg; 95% CI −0.4 to 0.7 mL/kg) or platelets (mean difference 0.6 mL/kg; 95% CI −1.3 to 2.5 mL/kg) than controls, after adjusting for transfusion of blood products prior to match day and remaining days in the ICU.
Discussion Using the definitions proposed by the ISTH, we showed that CRBs seem to be common in this patient population with an incidence of 9.1%. Invasive ventilation and stress ulcer prophylaxis were associated with increased risk of a CRB, whereas
Table III. Association of selected patient characteristics and therapies with CRB Unadjusted OR Age per 12-mo increase Predicted risk of mortality per 1% increase Surgical admission Lowest platelet count per 10 000 increase Highest international normalized ratio per 1.0 increase Highest activated partial thromboplastin time per 1 s increase Antithrombotic Antifibrinolytic Aspirin Steroid Nonsteroidal anti-inflammatory drug Stress ulcer prophylaxis Vasopressor Plasma transfusion Platelet transfusion Worst level of respiratory support Noninvasive vs none Invasive vs none Extracorporeal organ support Minor bleed
95% CI
0.99 1.06 1.15 0.97 3.16 1.02 0.39 1.67 0.29 1.23 1.15 2.66 6.70 5.00 8.00
0.99-1.00 1.00-1.13 0.51-2.56 0.93-1.00 0.93-10.74 0.99-1.06 0.08-1.83 0.30-9.39 0.03-2.42 0.58-2.63 0.51-2.57 1.26-5.61 2.49-18.06 1.34-18.62 2.00-31.99
5.52 25.91 2.00 0.53
0.60-51.22 3.40-197.73 0.37-10.92 0.21-1.35
4
Adjusted OR
95% CI
0.29
0.10-0.84
0.04
0.002-0.58
2.70
1.08-6.74
4.95 61.35
0.50-49.12 6.27-600.24
White et al FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
■■ 2017
ORIGINAL ARTICLES
Figure 2. Kaplan-Meier curves of time to discharge from A, the ICU and B, the hospital.
surgical admission and aspirin were associated with decreased risk. Complications associated with CRBs were prolonged stay in both the ICU and the hospital, prolonged use of vasopressor, and increased transfusion of red blood cells. Our findings suggest that, like adults, bleeding complicates critical illness in children. Bleeding is perceived to occur commonly in certain children in the ICU.13 Yet, few studies have investigated bleeding in this population. The incidence of a CRB in our study is consistent with that in critically ill adults and in selected children admitted to the ICU with illnesses or therapies that increase the risk of bleeding.1-9 However, in general, bleeding seemed less common in critically ill children than in adults. In contrast to adults in whom as many as 90% have at least 1 bleed while admitted to the ICU, only 37.3% of children in our cohort had a bleed.2 Among those who had a bleed, the median number of bleeds in our cohort was only 1 compared with 3 in adults. CRBs also tended to be shorter in duration in children with a median of 1 day vs 4 days in adults. The differences in the characteristics of the bleed probably reflect the differences in underlying illnesses and therapies between the 2 age groups, and physiological changes in the hemostatic system that occur with age.1,14 We found invasive ventilation to be associated with increased risk of a CRB. It is unclear whether this risk factor is modifiable. Although bleeding from the lungs may be due to suctioning or ventilator-induced lung injury, it is also possible that the bleeding is due to the underlying lung disease. The associations of stress ulcer prophylaxis and aspirin with a CRB likely represent confounding by indication. Stress ulcer prophylaxis is commonly provided to invasively ventilated children in our ICU. Children who are at increased risk of bleeding, on the contrary, are unlikely to be given aspirin. Surgical admission seemed to be protective of a CRB. However, this is likely the result of excluding children who already had a CRB
upon admission to the ICU. Surgical site was the most common site of minor bleed, which was not an exclusion criterion. Unlike in adults, we did not find platelet count or activated partial thromboplastin to be associated with a CRB.1 The effects of these derangements in coagulation on the risk of a CRB may have been negated by the opposite effects of transfusion of platelets and plasma. Similar to adults, we demonstrated that a CRB was associated with complications.2 The prolonged durations of stay in both the ICU and the hospital translate to increased resource utilization and cost with a CRB. We also found increased vasopressor use and transfusion of red blood cells in children with CRBs. Matching by time, adjusting for these therapies prior to match day and only counting the days on vasopressor and volume of blood products transfused from match day minimized confounding and suggests that the association is real. However, although the associations were statistically significant, it is unclear whether they are physiologically significant or simply reflect our local practice.15 Multicenter studies are needed to confirm our results. We were unable to find a difference in mortality with a CRB, as was shown in adults, because of the low mortality rate in our cohort. In children on extracorporeal organ support with higher risks of mortality, bleeding was associated with a nearly 2-fold increased risk of death.6 Our findings have implications for research. The novel bleeding assessment tool we developed can be used for future studies, particularly on anticoagulation. Excellent agreement can be achieved with the tool. However, intensive training prior to the study and regular re-training, called calibration exercises, will be needed to maintain acceptable agreement throughout the study.16 Although we initially achieved a k of 1.0 during phase 1 of our study, k decreased to 0.47-0.57 during phase 2. The incidence of bleeding that we report can inform the design of studies with bleeding as an outcome. In particular, the
Epidemiology of Bleeding in Critically Ill Children FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
5
THE JOURNAL OF PEDIATRICS • www.jpeds.com baseline incidence of bleeding is essential in calculating sample sizes. Certain patient populations may also need to be excluded from studies that can potentially increase the risk of bleeding based on their baseline risks. Certain limitations should be noted. It is possible that we may have overestimated the incidence of CRBs because of the definitions we used. In some cases, bleeds were automatically categorized as a CRB based on site. For example, any bleed from the respiratory system was considered a CRB, which was not the case in adults.1,2 The critical nature of each site should be re-assessed with respect to the resulting physiologic derangement if a bleed were to occur. In other cases, a bleed that would otherwise have been minor would be considered a CRB because blood was transfused. It was not always clear from the medical records whether the transfusion was in response to the bleed. Therapy provided would need to be separated from physiologic derangements to provide a more accurate assessment of the severity of a bleed.16 We may have underestimated the incidence of minor bleeds. Unlike CRBs, minor bleeds were less likely to be documented. Documentation of bleeds were not consistent with multiple medical providers recording contrasting details that could have affected the agreement between investigators. Prospective studies in which details of the bleed can be confirmed will likely result in improved agreement. We performed the study in a single center. Our findings should be confirmed in other centers. Our sample size was not powered to detect specific associations between risk factors and CRBs. It is possible that we missed other important risk factors because of our sample size.6 Our sample size also resulted in lack of precision in the magnitude of the statistically significant associations that we detected. Lastly, we had to rely on multiple imputation because of missing data, particularly on the laboratory tests. Consistent measurement of these tests may further characterize the association between derangements in coagulation and CRBs. The novel bleeding assessment tool that we developed can be used in future studies in children with bleeding as an outcome. In addition, revising the definition proposed by ISTH to focus primarily on the physiologic derangements resulting from a bleed may provide a more accurate assessment of the severity of a bleed. ■ Submitted for publication Oct 17, 2016; last revision received Dec 14, 2016; accepted Jan 10, 2017 Reprint requests: Edward Vincent S. Faustino, MD, MHS, Yale School of Medicine, 333 Cedar Street, New Haven, CT 06520. E-mail: [email protected]
Volume ■■
References 1. Lauzier F, Arnold DM, Rabbat C, Heels-Ansdell D, Zarychanski R, Dodek P, et al. Risk factors and impact of major bleeding in critically ill patients receiving heparin thromboprophylaxis. Intensive Care Med 2013;39:2135-43. 2. Arnold D, Donahoe L, Clarke F, Tkaczyk A, Heels-Ansdell D, Zytaruk N, et al. Bleeding during critical illness: a prospective cohort study using a new measurement tool. Clin Invest Med 2007;30:E93-102. 3. Kuhle S, Eulmesekian P, Kavanagh B, Massicotte P, Vegh P, Mitchell LG. A clinically significant incidence of bleeding in critically ill children receiving therapeutic doses of unfractionated heparin: a prospective cohort study. Haematologica 2007;92:244-7. 4. Schroeder AR, Axelrod DM, Silverman NH, Rubesova E, Merkel E, Roth SJ. A continuous heparin infusion does not prevent catheter-related thrombosis in infants after cardiac surgery. Pediatr Crit Care Med 2010;11:48995. 5. Nadel S, Goldstein B, Williams MD, Dalton H, Peters M, Macias WL, et al. Drotrecogin alfa (activated) in children with severe sepsis: a multicentre phase III randomised controlled trial. Lancet 2007;369:836-43. 6. Dalton HJ, Garcia-Filion P, Holubkov R, Moler FW, Shanley T, Heidemann S, et al. Association of bleeding and thrombosis with outcome in extracorporeal life support. Pediatr Crit Care Med 2015;16:167-74. 7. Sahin S, Ayar G, Yazici MU, Koksal T, Akman AO, Gunduz R, et al. Stress induced gastrointestinal bleeding in a pediatric intensive care unit: which risk factors should necessitate prophilaxis? Minerva Pediatr 2014;68:1926. 8. Longmuir SQ, McConnell L, Oral R, Dumitrescu A, Kamath S, Erkonen G. Retinal hemorrhages in intubated pediatric intensive care patients. J AAPOS 2014;18:129-33. 9. Deerojanawong J, Peongsujarit D, Vivatvakin B, Prapphal N. Incidence and risk factors of upper gastrointestinal bleeding in mechanically ventilated children. Pediatr Crit Care Med 2009;10:91-5. 10. Mitchell LG, Goldenberg NA, Male C, Kenet G, Monagle P, NowakGottl U, et al. Definition of clinical efficacy and safety outcomes for clinical trials in deep venous thrombosis and pulmonary embolism in children. J Thromb Haemost 2011;9:1856-8. 11. Vriesendorp TM, van Santen S, DeVries JH, de Jonge E, Rosendaal FR, Schultz MJ, et al. Predisposing factors for hypoglycemia in the intensive care unit. Crit Care Med 2006;34:96-101. 12. Slater A, Shann F, Pearson G. PIM2: a revised version of the Paediatric Index of Mortality. Intensive Care Med 2003;29:278-85. 13. Kenet G, Strauss T, Kaplinsky C, Paret G. Hemostasis and thrombosis in critically ill children. Semin Thromb Hemost 2008;34:451-8. 14. Monagle P, Chan AK, Goldenberg NA, Ichord RN, Journeycake JM, NowakGottl U, et al. Antithrombotic therapy in neonates and children: antithrombotic Therapy and Prevention of Thrombosis, 9th ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines. Chest 2012;141:e737S-801S. 15. Bateman ST, Lacroix J, Boven K, Forbes P, Barton R, Thomas NJ, et al. Anemia, blood loss, and blood transfusions in North American children in the intensive care unit. Am J Respir Crit Care Med 2008;178:26-33. 16. Arnold DM, Lauzier F, Rabbat C, Zytaruk N, Barlow Cash B, Clarke F, et al. Adjudication of bleeding outcomes in an international thromboprophylaxis trial in critical illness. Thromb Res 2013;131:2049.
6
White et al FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
■■ 2017
ORIGINAL ARTICLES
Bleeding in the PICU Instructions for Assessors: Complete this form daily while the patient was admitted to the PICU. USE A SEPARATE FORM FOR EACH UNIQUE SITE OF BLEEDING. * Required
1. Patient Study ID *
2. Study Collaborator *
3. Day of Admission in the PICU * Day of admission to the PICU = Day 1. Please do NOT include days admitted to other units in the hospital.
4. Bleeding Site * Choose the site of the bleeding event on this day. If the patient did not have a bleeding event on this day, select "No Bleeding Event." Mark only one oval.
Stop filling out this form.
No Bleeding Event
Vascular Catheter or Insertion Site
Skip to question 5.
Other Skin Bleeding (NOT CVC-Related) Skip to question 8.
Gastrointestinal
Respiratory (Including the Thoracic Cavity) Surgical Site Genitourinary
Skip to question 7.
Skip to question 9.
Skip to question 10. Skip to question 11. Skip to question 12.
Intra/Retroperitoneal Central Nervous System
Skip to question 13.
Eye, Ear, Nose, or Mouth
Skip to question 14.
Pericardium Joint
Skip to question 17.
Skip to question 15.
Other (Describe)
Skip to question 16.
Bleeding Site - Vascular or Catheter Insertion Site 5. Vascular or Catheter Insertion Site * Mark only one oval.
Peripheral Vein - Arm Peripheral Vein - Leg Femoral Vein Internal Jugular Vein Subclavian Vein
Other:
Appendix. Bleeding assessment tool. (Continues)
Epidemiology of Bleeding in Critically Ill Children FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
6.e1
THE JOURNAL OF PEDIATRICS • www.jpeds.com
Volume ■■
6. Specify Side * Mark only one oval.
Left Right
Skip to question 17.
Bleeding Site - Other Skin Bleeding (NOT CVC-Related) 7. Other Skin Bleeding Site(s) - NOT CVC-Related * Please select all that apply. Check all that apply.
Bruising Petechiae Non-Surgical Wound
Other: Skip to question 17.
Bleeding Site - Gastrointestinal 8. Gastrointestinal Bleeding Site * Please select all that apply. Check all that apply. Hematemesis Melena Hematochezia
Other: Skip to question 17.
Bleeding Site - Respiratory (Including the Thoracic Cavity) 9. Respiratory/Thoracic Cavity Bleeding Site * Please select all that apply. Check all that apply. ETT/Tracheostomy Aspirate Hemoptysis Hemothorax Other: Skip to question 17.
Appendix. Continues.
6.e2
White et al FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
■■ 2017
ORIGINAL ARTICLES
Bleeding Site - Surgical Site 10. Surgical Bleeding Site * Please select all that apply. Check all that apply. Incision Drain/Tube
Other: Skip to question 17.
Bleeding Site - Genitourinary 11. Genitourinary Bleeding Site * Please select all that apply. Check all that apply. Gross Hematuria Bleeding Around a Urinary Catheter Menstrual Bleed (that required medical consultation or intervention) Other: Skip to question 17.
Bleeding Site - Intra/Retroperitoneal 12. Intra/Retroperitoneal Bleed * Please describe the location of the bleed.
Skip to question 17.
Bleeding Site - Central Nervous System 13. Central Nervous System Bleed * Please select all that apply. Check all that apply.
Epidural Bleed Subdural Bleed Subarachnoid Bleed
Intraparenchymal Bleed Intraventricular Bleed Intraspinal Bleed (any site within the spinal column)
Other: Skip to question 17.
Appendix. Continues.
Epidemiology of Bleeding in Critically Ill Children FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
6.e3
THE JOURNAL OF PEDIATRICS • www.jpeds.com
Volume ■■
Bleeding Site - Eye, Ear, Nose, or Mouth 14. Eye, Ear, Nose, or Mouth Bleed * Please select all that apply. Check all that apply. Subconjunctival Bleed
Retinal Bleed Hemotympanum Epistaxis Oropharyngeal Bleed Other: Skip to question 17.
Bleeding Site - Joint 15. Bleeding in the Joint * Please specify which joint(s) is/are involved. Check all that apply. Shoulder Elbow Wrist Finger
Hip Knee Ankle Toe Other: Skip to question 17.
Bleeding Site - Other (Describe) 16. Other Bleeding Site * Please describe the location of the bleed (if possible).
Skip to question 17.
Appendix. Continues.
6.e4
White et al FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
■■ 2017
ORIGINAL ARTICLES
Bleeding Severity Please select AT LEAST one item from ONLY ONE of the bleeding severity categories below. Please select the most severe category as it applies to the bleeding event documented on this current day of admission. Remember to use a different form for another bleeding event on this same day but at a different site.
17. Major Bleeding Please check all that apply. Check all that apply.
Fatal Bleeding Clinically Overt Bleeding with Decrease in Hgb of >= 2g/dL in a 24-hr Period Bleed that is Intra/Retroperitoneal, Pulmonary, or Involves the CNS Bleeding Requiring Surgical Intervention in the Operating Room 18. Clinically Relevant Non-Major Bleeding
Please check all that apply. Check all that apply. Overt Bleeding Requiring Blood Product Administration Bleeding that Requires Medical or Surgical Intervention NOT in the Operating Room (and
Does NOT Include Menstrual Bleeding) 19. Minor Bleeding
Please check all that apply. Check all that apply. Overt or Macroscopic Evidence of Bleeding that Does NOT Fulfill Above Criteria (for Major Bleeding or Clinically Relevant Non-Major Bleeding)
Menstrual Bleeding Resulting in Medical Consult and/or Intervention Skip to question 20.
Timing 20. Type of Bleed * Mark only one oval. New Bleed Ongoing Bleed
Recurrent Bleed at the Same Site 21. Bleeding Start Date Please type which day of admission the bleed started.
22. Bleeding End Date
Please type which day of admission the bleed ended.
23. Started Prior to Today's Study Day? *
Mark only one oval. Yes No
Appendix. Continued.
Epidemiology of Bleeding in Critically Ill Children FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
6.e5
THE JOURNAL OF PEDIATRICS • www.jpeds.com
Volume ■■
Figure 1. Frequency of bleeding based on severity.
6.e6
White et al FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
ORIGINAL ARTICLES
Epidemiology of Bleeding in Critically Ill Children Lauren J. White, MD1,*, Ryan Fredericks, MD1,*, Candace N. Mannarino, MD1, Stephen Janofsky, MD1, and Edward Vincent S. Faustino, MD, MHS2 Objective To determine the epidemiology of bleeding in critically ill children. Study design We conducted a cohort study of children <18 years old admitted to the pediatric intensive care
unit for >24 hours and without clinically relevant bleed (CRB) on admission. CRB was defined as resulting in severe physiologic derangements, occurring at a critical site or requiring major therapeutic interventions. Using a novel bleeding assessment tool that we developed, characteristics of the CRB were abstracted from the medical records independently and in duplicate. From the cohort, we matched each child with CRB to 4 children without CRB based on onset of CRB. Risk factors and complications of CRB were identified from this matched group of children. Results We analyzed 405 children with a median age of 35 months (IQR 7-130 months). A total of 37 (9.1%) children developed CRB. The median number of days with CRB was 1 day (IQR 1-2 days). Invasive ventilation (OR 61.35; 95% CI 6.27-600.24), stress ulcer prophylaxis (OR 2.70; 95% CI 1.08-6.74), surgical admission (OR 0.29; 95% CI 0.10-0.84), and aspirin (OR 0.04; 95% CI 0.002-0.58) were associated with CRB. CRB was associated with longer time to discharge from the unit (hazard ratio 0.20; 95% CI 0.13-0.33) and the hospital (hazard ratio 0.49; 95% CI 0.33-0.73). Children with CRB were on vasopressor longer and transfused more red blood cells after the CRB than those without CRB. Conclusions Our findings suggest that bleeding complicates critical illness in children. (J Pediatr 2017;■■:■■-■■). leeding complicates critical illness in adults.1 Approximately 6%-20% of adults admitted to the intensive care unit (ICU) develop major bleed, which is a clinically relevant bleed (CRB) resulting in severe physiologic derangements, occurring at a critical site or requiring major therapeutic intervention.1,2 An additional 8%-70% of them develop a minor bleed, which is an overt bleed that is less severe than a major bleed. Decrease in platelet count, prolongation of activated partial thromboplastin time, heparin at therapeutic doses, antiplatelet agents, renal replacement therapy, and recent surgery are associated with a major bleed. Compared with those with no bleeds, critically ill adults with major bleeds are transfused with larger amounts of blood products, stay longer in the ICU, and have higher risks of mortality. Bleeding may also complicate critical illness in children. However, the epidemiology (ie, incidence, risk factors and complications) of bleeding in these children is unclear. Studies have focused on selected children admitted to the ICU with illnesses or therapies that increase bleeding risk. These studies report that 3%-52% of the children bleed while admitted.3-9 The uncertainty in the estimates is partly due to inconsistencies in the definitions used for bleeding. Recently, the International Society on Thrombosis and Hemostasis (ISTH) defined categories of bleeding to standardize outcomes in clinical trials of anticoagulation.10 Aside from major and minor bleeds, as previously defined, the category of clinically relevant nonmajor bleeds was added for bleeding that does not fulfill the criteria for a major bleed but required transfusion of blood products, or medical or surgical intervention other than in an operating suite to restore hemostasis. In this study, we developed a novel bleeding assessment tool based on the definitions recommended by ISTH. Using this tool, we aimed to determine the epidemiology of bleeding in an unselected cohort of critically ill children.
B
Methods The study was composed of 2 phases. In phase 1, we developed the bleeding assessment tool, and then pilot tested it in a crosssectional study. In phase 2, we performed a cohort study to determine the incidence of bleeding. Risk factors and complications associated with bleeding were identified in a nested case-control study. The study was approved by the Human Investigation Committee at Yale University, which waived the need for signed consent. From the 1Department of Pediatrics, Yale-New Haven Children’s Hospital, New Haven, CT; and 2Department of Pediatrics, Yale School of Medicine, New Haven, CT *Contributed equally.
CRB HR ICU ISTH
Clinically relevant bleed Hazard ratio Intensive care unit International Society on Thrombosis and Hemostasis
Supported in part by the American Heart Association (14CRP20490002 to Dr. E.F.). The authors declare no conflicts of interest. 0022-3476/$ - see front matter. © 2017 Elsevier Inc. All rights reserved. http://dx.doi.org10.1016/j.jpeds.2017.01.026
1 FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
THE JOURNAL OF PEDIATRICS • www.jpeds.com Phase 1 Development of the Bleeding Assessment Tool. One of the investigators created the electronic tool that contained the definitions recommended by ISTH for major, clinically relevant nonmajor, and minor bleed (Appendix; available at www.jpeds.com). Items used to categorize a bleed were described in sufficient detail in the tool to maximize consistency in its use. The tool also contained the site and timing of the bleed. It was revised after discussion among all investigators. Pilot Testing of the Bleeding Assessment Tool. Using an electronic patient database, children admitted to the pediatric ICU at Yale-New Haven Children’s Hospital from January 1, 2013, to December 31, 2013, with a diagnosis of a bleed were randomly selected. The investigators blindly and independently reviewed the medical records of 10 children per cycle. The first bleed for every child was categorized using the bleeding assessment tool. In between cycles, the entire investigative team discussed the cases and resolved discrepancies. The process was repeated until the chance-corrected inter-rater agreement (k) was ≥0.80. Phase 2 Subjects. A cohort of children less than 18 years old who were admitted to the pediatric ICU at Yale-New Haven Children’s Hospital from January 1, 2014, to December 31, 2014, were randomly selected using the electronic patient database. Excluded were those who stayed in the ICU for less than 24 hours because they were mostly admitted for observation. For those with multiple admissions, 1 admission was randomly selected. Children who had a CRB, whether major or nonmajor, upon admission were excluded after review of medical records. We analyzed major and nonmajor CRBs together because both categories, by definition, are clinically significant and have the potential to impact the child’s health and outcome. Each child with a CRB (case) was randomly matched to 4 children without a CRB (control). The day of stay in the ICU when the CRB first occurred in cases were matched to a similar time point in controls (match day). Matching based on time leads to complete adjustment for the effect of duration of stay in the ICU and its determinants.11 Procedures. The medical records of each eligible child were blindly and independently reviewed by 2 randomly assigned investigators. The bleeding assessment tool was completed by each investigator for each bleed. A third randomly assigned investigator adjudicated discrepancies in the categorization of the bleed. The investigator primarily assigned to each child also collected from the medical records data on demographics; medications, interventions, and worst laboratory tests at any time while admitted to the ICU; and, outcomes. Collection of daily data was censored at 14 days, which represented the 95th percentile of the duration of stay in our ICU. Predicted risk of mortality was calculated based on the Pediatric Index of Mortality 2.12 Anticoagulants included unfractionated heparin
Volume ■■ except at doses to maintain patency of a vascular catheter, low molecular weight heparin, and warfarin, whereas antifibrinolytics included tranexamic and aminocaproic acid. Stress ulcer prophylaxis included proton pump and histamine H2 receptor inhibitors. Vasopressor use was defined as dopamine ≥5 mcg/kg/min, or any dose of dobutamine, epinephrine, norepinephrine, phenylephrine, milrinone, and vasopressin. Respiratory support was categorized as invasive ventilation via endotracheal tube or tracheostomy, noninvasive ventilation (ie, continuous or bilevel positive airway pressure, or high flow nasal cannula greater than 5 L/min for children less than 10 years old and greater than 8 L/min for children at least 10 years old) or none, which included any support that did not qualify for invasive or noninvasive ventilation. Extracorporeal support included extracorporeal membrane oxygenation, renal replacement therapy, and erythro- or leukapheresis. Cryoprecipitate was included in plasma transfusion. Statistical Analyses In phase 1, crude inter-rater agreement and k were calculated. Bleeds were dichotomously categorized based on the presence of a CRB in calculating k. Furthermore, agreement was defined as a majority of investigators assigning the same category. In a sensitivity analysis, k was calculated based on the presence of any bleed. In phase 2, the incidence of bleeding was expressed as proportion of children and days in the ICU with bleed. Conditional logistic regression with stepwise backward elimination at an exit threshold of P > .05 was used to identify factors associated with a CRB. Data on potential risk factors were limited to a specific number of days prior to match day. When applicable, the time period was based on pharmacokinetics (ie, 7 days for aspirin, 5 days for warfarin, and 1 day for unfractionated heparin, low molecular weight heparin, tranexamic acid, aminocaproic acid, vasopressor, and heparin for extracorporeal support). For others, it was arbitrarily set at 2 days based on consensus among investigators. Multiple imputations using multivariate normal distribution was used for missing laboratory data. Times to outcome, adjusted for age, and predicted risk of mortality, were compared between cases and controls using Cox regression with shared frailty model to account for the matching. Number of days on vasopressors and volume of blood products transfused from match day were compared between cases and controls using linear mixed effects model with matching as random effect. Use of vasopressors or blood product transfusion prior to match day, and duration of stay in the ICU from match day were entered as fixed effects. Data was expressed as median (IQR), mean difference (95% CI), count (percentage), OR (95% CI), and hazard ratio (HR; 95% CI). In the absence of reliable estimates of the proportion of critically ill children with bleed, 405 children were analyzed assuming a true proportion of 0.50. This provided the largest sample size for a 95% CI of ±0.05. All analyses were performed in Stata v 14 (StataCorp, College Station, Texas). A 2-sided P value of <.05 was considered statistically significant.
2
White et al FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
■■ 2017
ORIGINAL ARTICLES Results
Bleeding Assessment Tool A total of 3 cycles were performed before k ≥ 80 was achieved. In the presence of a CRB, the crude inter-rater agreement increased from 0.80 to 0.90 to 1.00, whereas k increased from 0.48 to 0.67 to 1.00. In between cycles, specific scenarios in which discrepancies occurred were discussed, which improved consistency in coding. No revisions on the tool were required. The overall k for the 30 children was 0.71. When applied to the cohort of 405 children, the crude inter-rater agreement was 0.92 but k was only 0.47. In the presence of any bleed, the crude inter-rater agreement also increased from 0.80 to 0.90 to 1.00 in phase 1. The k at each cycle was 0.25, 0.18, and 1.00. For the entire cohort, the crude inter-rater agreement was 0.81 with k of 0.57. Incidence of Bleeding There were 1069 potentially eligible admissions, of which 508 were hierarchically excluded for duration of stay in the ICU
of less than 24 hours (n = 366), multiple admissions (n = 119), and age on admission of at least 18 years (n = 23). Of the remaining 561 admissions, medical record review of the first randomly selected 450 excluded 45 because of the presence of a CRB on admission to the ICU. The most common causes of a CRB on admission were trauma (n = 27) and surgery (n = 10). The mean age of the final cohort of 405 children was 35 months (IQR 7-130 months) (Table I). The predicted risk of mortality was 0.008 (IQR 0.002-0.015). Only 117 (28.9%) were admitted after surgery. The median durations of stay in the ICU and the hospital were 2.0 days (IQR 1.4-3.3 days) and 4 days (IQR 3-8 days), respectively. A total of 4 (1.0%) died while admitted to the hospital. A total of 151 (37.3%) children had 197 bleeding events in 305 days while admitted to the ICU (Table I). Of those with bleeds, most had only 1 event (79.6%) (Figure 1; available at www.jpeds.com). The worst bleed was a CRB in 37 children (major in 30 and nonmajor in 7) for an incidence of 9.1%, and minor in 114 children (28.1%). Of the 197 bleeding events,
Table I. Characteristics of children analyzed* Cohort (n = 405) Demographics Age, mo Male Diagnostic category Cardiovascular Endocrinologic Gastrointestinal Genetic Infectious Injury/poisoning Neurologic Oncologic Orthopedic Respiratory Others Predicted risk of mortality Surgical admission Laboratory tests‡ Lowest hemoglobin Lowest platelet count Highest international normalized ratio Highest activated partial thromboplastin time Medications Antithrombotic Antifibrinolytic Aspirin Steroid Nonsteroidal anti-inflammatory drug Stress ulcer prophylaxis Vasopressor Transfusions Red blood cell Plasma Platelets Other interventions Worst level of respiratory support None Noninvasive Invasive Extracorporeal organ support
35 (7-130) 229 (56.5) 66 24 11 18 15 18 42 19 14 138 40 0.008 117
(16.3) (5.9) (2.7) (4.4) (3.7) (4.4) (10.4) (4.7) (3.5) (34.1) (9.9) (0.002-0.015) (28.9)
10.8 242 1.2 26.7
(9.2-12.4) (157-345) (1.0-1.3) (24.0-34.6)
44 31 33 161 126 207 53
(10.9) (7.7) (8.2) (39.8) (31.1) (51.1) (13.1)
81 (20.0) 27 (6.7) 16 (4.0)
Cases (n = 37) 56.4 (3-102) 25 (67.6) 8 0 1 3 3 0 5 0 2 9 6 0.042 11 9.7 181 1.3 32.3 2 2 1 14 10 24 13
(21.6) (0) (2.7) (8.1) (8.1) (0) (13.5) (0) (5.4) (24.3) (16.7) (0.008-0.039) (29.7) (7.7-11.6) (108-257) (1.1-1.6) (24.9-40.7) (5.4) (5.4) (2.7) (37.8) (27.0) (64.8) (35.1)
Controls (n = 148) 64.9 (6-135.5) 81 (54.7) 27 7 4 8 6 7 12 7 2 56 12 0.019 40 11.1 259 1.1 26.9 18 5 12 49 36 60 13
P value† .53 .15 .06
(18.4) (4.7) (2.7) (5.4) (4.1) (4.7) (8.1) (4.7) (1.4) (37.8) (8.1) (0.002-0.014) (27.0)
.07 .74
(9.4-12.5) (163-351) (0.8-1.3) (18.0-36.8)
.005 .04 .07 .15
(12.2) (3.4) (8.1) (33.1) (24.3) (40.5) (8.8)
.23 .56 .25 .58 .74 .01 <.001
19 (51.4) 5 (13.5) 6 (16.2)
15 (10.1) 4 (2.7) 3 (2.0)
<.001 .02 .003
1 3 33 2
46 40 62 4
<.001 113 94 198 16
(27.9) (23.2) (48.9) (4.0)
(2.7) (8.1) (89.2) (5.4)
(31.1) (27.0) (41.9) (2.7)
.42
*Data is expressed as median (IQR) or count (percentage). †P values refer to comparisons between cases and controls. ‡Multiple imputations were used to analyze the laboratory tests.
Epidemiology of Bleeding in Critically Ill Children FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
3
THE JOURNAL OF PEDIATRICS • www.jpeds.com Table II. Sites of bleeding* Clinically relevant (n = 47) Respiratory including thoracic cavity Surgical site Vascular catheter or insertion site Gastrointestinal Other skin bleeding not catheter-related Genitourinary Central nervous system Pericardium Ear/eye/nose/mouth
24 12 4 3 1 1 1 1 0
(51.1) (25.5) (8.5) (6.4) (2.1) (2.1) (2.1) (2.1) (0)
Minor (n = 150) 0 104 11 10 10 8 0 0 7
(0) (69.3) (7.3) (6.7) (6.7) (5.3) (0) (0) (4.7)
*Data are expressed as counts (percentages).
47 (23.9%) were CRBs (36 major and 11 nonmajor bleeds) and 150 were minor. Among children with any bleed, the median number of days in the ICU that a child had a bleed was 2 days (IQR 1-2 days). The median number of days with CRB was 1 day (IQR 1-2 days), whereas the median time to the first CRB was 2 days (IQR: 1-3 days). The most common site for a CRB was the respiratory system (51.1%) and that for minor bleeds was the surgical site (69.3%; Table II). Risk Factors for Clinically Relevant Bleed The 37 children with a CRB were matched to 148 children without a CRB. Cases had lower hemoglobin and platelet count prior to the CRB than their controls (Table I). Although only 125, 123, 65, and 65 matched children had values for hemoglobin, platelet count, international normalized ratio, and activated partial thromboplastin time, respectively, the comparisons were similar with and without multiple imputation. The proportion of cases on stress ulcer prophylaxis, vasopressor, transfused with red blood cells, plasma or platelets, or invasive ventilation prior to the CRB were higher than that of controls. There was no significant difference in the propor-
Volume ■■ tion of children with a minor bleed prior to the CRB between cases and controls. In the multivariable analysis, invasive ventilation (OR vs no respiratory support: 61.35; 95% CI 6.27600.24) and stress ulcer prophylaxis (OR 2.70; 95% CI 1.086.74) were associated with increased risk of CRB, whereas surgical admission (OR 0.29; 95% CI 0.10-0.84) and aspirin (OR 0.04; 95% CI 0.002-0.58) were associated with decreased risk (Table III). Complications of Clinically Relevant Bleed Children with a CRB had longer times to discharge from the ICU (HR 0.21; 95% CI 0.13-0.33) and from the hospital (HR 0.49; 95% CI 0.33-0.73) after adjusting for age and predicted risk of mortality (Figure 2, A and B). The median length of stay in the ICU was 4.9 days (IQR 2.1-9.6 days) for cases and 2.1 days (IQR 1.5-4.0 days) for controls, whereas the median length of stay in the hospital was 8 days (IQR 6-25 days) for cases and 4 days (IQR 2-9 days) for controls. Mortality was not significantly different between cases and controls (HR 4.76; 95% CI 0.40-56.58). After adjusting for use of vasopressor prior to match day and remaining days in the ICU, cases were on vasopressors longer than controls by 0.4 days (95% CI 0.20.6 days). Cases also received more red blood cells (mean difference: 5 mL/kg; 95% CI 1-9 mL/kg), but not plasma (mean difference 0.2 mL/kg; 95% CI −0.4 to 0.7 mL/kg) or platelets (mean difference 0.6 mL/kg; 95% CI −1.3 to 2.5 mL/kg) than controls, after adjusting for transfusion of blood products prior to match day and remaining days in the ICU.
Discussion Using the definitions proposed by the ISTH, we showed that CRBs seem to be common in this patient population with an incidence of 9.1%. Invasive ventilation and stress ulcer prophylaxis were associated with increased risk of a CRB, whereas
Table III. Association of selected patient characteristics and therapies with CRB Unadjusted OR Age per 12-mo increase Predicted risk of mortality per 1% increase Surgical admission Lowest platelet count per 10 000 increase Highest international normalized ratio per 1.0 increase Highest activated partial thromboplastin time per 1 s increase Antithrombotic Antifibrinolytic Aspirin Steroid Nonsteroidal anti-inflammatory drug Stress ulcer prophylaxis Vasopressor Plasma transfusion Platelet transfusion Worst level of respiratory support Noninvasive vs none Invasive vs none Extracorporeal organ support Minor bleed
95% CI
0.99 1.06 1.15 0.97 3.16 1.02 0.39 1.67 0.29 1.23 1.15 2.66 6.70 5.00 8.00
0.99-1.00 1.00-1.13 0.51-2.56 0.93-1.00 0.93-10.74 0.99-1.06 0.08-1.83 0.30-9.39 0.03-2.42 0.58-2.63 0.51-2.57 1.26-5.61 2.49-18.06 1.34-18.62 2.00-31.99
5.52 25.91 2.00 0.53
0.60-51.22 3.40-197.73 0.37-10.92 0.21-1.35
4
Adjusted OR
95% CI
0.29
0.10-0.84
0.04
0.002-0.58
2.70
1.08-6.74
4.95 61.35
0.50-49.12 6.27-600.24
White et al FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
■■ 2017
ORIGINAL ARTICLES
Figure 2. Kaplan-Meier curves of time to discharge from A, the ICU and B, the hospital.
surgical admission and aspirin were associated with decreased risk. Complications associated with CRBs were prolonged stay in both the ICU and the hospital, prolonged use of vasopressor, and increased transfusion of red blood cells. Our findings suggest that, like adults, bleeding complicates critical illness in children. Bleeding is perceived to occur commonly in certain children in the ICU.13 Yet, few studies have investigated bleeding in this population. The incidence of a CRB in our study is consistent with that in critically ill adults and in selected children admitted to the ICU with illnesses or therapies that increase the risk of bleeding.1-9 However, in general, bleeding seemed less common in critically ill children than in adults. In contrast to adults in whom as many as 90% have at least 1 bleed while admitted to the ICU, only 37.3% of children in our cohort had a bleed.2 Among those who had a bleed, the median number of bleeds in our cohort was only 1 compared with 3 in adults. CRBs also tended to be shorter in duration in children with a median of 1 day vs 4 days in adults. The differences in the characteristics of the bleed probably reflect the differences in underlying illnesses and therapies between the 2 age groups, and physiological changes in the hemostatic system that occur with age.1,14 We found invasive ventilation to be associated with increased risk of a CRB. It is unclear whether this risk factor is modifiable. Although bleeding from the lungs may be due to suctioning or ventilator-induced lung injury, it is also possible that the bleeding is due to the underlying lung disease. The associations of stress ulcer prophylaxis and aspirin with a CRB likely represent confounding by indication. Stress ulcer prophylaxis is commonly provided to invasively ventilated children in our ICU. Children who are at increased risk of bleeding, on the contrary, are unlikely to be given aspirin. Surgical admission seemed to be protective of a CRB. However, this is likely the result of excluding children who already had a CRB
upon admission to the ICU. Surgical site was the most common site of minor bleed, which was not an exclusion criterion. Unlike in adults, we did not find platelet count or activated partial thromboplastin to be associated with a CRB.1 The effects of these derangements in coagulation on the risk of a CRB may have been negated by the opposite effects of transfusion of platelets and plasma. Similar to adults, we demonstrated that a CRB was associated with complications.2 The prolonged durations of stay in both the ICU and the hospital translate to increased resource utilization and cost with a CRB. We also found increased vasopressor use and transfusion of red blood cells in children with CRBs. Matching by time, adjusting for these therapies prior to match day and only counting the days on vasopressor and volume of blood products transfused from match day minimized confounding and suggests that the association is real. However, although the associations were statistically significant, it is unclear whether they are physiologically significant or simply reflect our local practice.15 Multicenter studies are needed to confirm our results. We were unable to find a difference in mortality with a CRB, as was shown in adults, because of the low mortality rate in our cohort. In children on extracorporeal organ support with higher risks of mortality, bleeding was associated with a nearly 2-fold increased risk of death.6 Our findings have implications for research. The novel bleeding assessment tool we developed can be used for future studies, particularly on anticoagulation. Excellent agreement can be achieved with the tool. However, intensive training prior to the study and regular re-training, called calibration exercises, will be needed to maintain acceptable agreement throughout the study.16 Although we initially achieved a k of 1.0 during phase 1 of our study, k decreased to 0.47-0.57 during phase 2. The incidence of bleeding that we report can inform the design of studies with bleeding as an outcome. In particular, the
Epidemiology of Bleeding in Critically Ill Children FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
5
THE JOURNAL OF PEDIATRICS • www.jpeds.com baseline incidence of bleeding is essential in calculating sample sizes. Certain patient populations may also need to be excluded from studies that can potentially increase the risk of bleeding based on their baseline risks. Certain limitations should be noted. It is possible that we may have overestimated the incidence of CRBs because of the definitions we used. In some cases, bleeds were automatically categorized as a CRB based on site. For example, any bleed from the respiratory system was considered a CRB, which was not the case in adults.1,2 The critical nature of each site should be re-assessed with respect to the resulting physiologic derangement if a bleed were to occur. In other cases, a bleed that would otherwise have been minor would be considered a CRB because blood was transfused. It was not always clear from the medical records whether the transfusion was in response to the bleed. Therapy provided would need to be separated from physiologic derangements to provide a more accurate assessment of the severity of a bleed.16 We may have underestimated the incidence of minor bleeds. Unlike CRBs, minor bleeds were less likely to be documented. Documentation of bleeds were not consistent with multiple medical providers recording contrasting details that could have affected the agreement between investigators. Prospective studies in which details of the bleed can be confirmed will likely result in improved agreement. We performed the study in a single center. Our findings should be confirmed in other centers. Our sample size was not powered to detect specific associations between risk factors and CRBs. It is possible that we missed other important risk factors because of our sample size.6 Our sample size also resulted in lack of precision in the magnitude of the statistically significant associations that we detected. Lastly, we had to rely on multiple imputation because of missing data, particularly on the laboratory tests. Consistent measurement of these tests may further characterize the association between derangements in coagulation and CRBs. The novel bleeding assessment tool that we developed can be used in future studies in children with bleeding as an outcome. In addition, revising the definition proposed by ISTH to focus primarily on the physiologic derangements resulting from a bleed may provide a more accurate assessment of the severity of a bleed. ■ Submitted for publication Oct 17, 2016; last revision received Dec 14, 2016; accepted Jan 10, 2017 Reprint requests: Edward Vincent S. Faustino, MD, MHS, Yale School of Medicine, 333 Cedar Street, New Haven, CT 06520. E-mail: [email protected]
Volume ■■
References 1. Lauzier F, Arnold DM, Rabbat C, Heels-Ansdell D, Zarychanski R, Dodek P, et al. Risk factors and impact of major bleeding in critically ill patients receiving heparin thromboprophylaxis. Intensive Care Med 2013;39:2135-43. 2. Arnold D, Donahoe L, Clarke F, Tkaczyk A, Heels-Ansdell D, Zytaruk N, et al. Bleeding during critical illness: a prospective cohort study using a new measurement tool. Clin Invest Med 2007;30:E93-102. 3. Kuhle S, Eulmesekian P, Kavanagh B, Massicotte P, Vegh P, Mitchell LG. A clinically significant incidence of bleeding in critically ill children receiving therapeutic doses of unfractionated heparin: a prospective cohort study. Haematologica 2007;92:244-7. 4. Schroeder AR, Axelrod DM, Silverman NH, Rubesova E, Merkel E, Roth SJ. A continuous heparin infusion does not prevent catheter-related thrombosis in infants after cardiac surgery. Pediatr Crit Care Med 2010;11:48995. 5. Nadel S, Goldstein B, Williams MD, Dalton H, Peters M, Macias WL, et al. Drotrecogin alfa (activated) in children with severe sepsis: a multicentre phase III randomised controlled trial. Lancet 2007;369:836-43. 6. Dalton HJ, Garcia-Filion P, Holubkov R, Moler FW, Shanley T, Heidemann S, et al. Association of bleeding and thrombosis with outcome in extracorporeal life support. Pediatr Crit Care Med 2015;16:167-74. 7. Sahin S, Ayar G, Yazici MU, Koksal T, Akman AO, Gunduz R, et al. Stress induced gastrointestinal bleeding in a pediatric intensive care unit: which risk factors should necessitate prophilaxis? Minerva Pediatr 2014;68:1926. 8. Longmuir SQ, McConnell L, Oral R, Dumitrescu A, Kamath S, Erkonen G. Retinal hemorrhages in intubated pediatric intensive care patients. J AAPOS 2014;18:129-33. 9. Deerojanawong J, Peongsujarit D, Vivatvakin B, Prapphal N. Incidence and risk factors of upper gastrointestinal bleeding in mechanically ventilated children. Pediatr Crit Care Med 2009;10:91-5. 10. Mitchell LG, Goldenberg NA, Male C, Kenet G, Monagle P, NowakGottl U, et al. Definition of clinical efficacy and safety outcomes for clinical trials in deep venous thrombosis and pulmonary embolism in children. J Thromb Haemost 2011;9:1856-8. 11. Vriesendorp TM, van Santen S, DeVries JH, de Jonge E, Rosendaal FR, Schultz MJ, et al. Predisposing factors for hypoglycemia in the intensive care unit. Crit Care Med 2006;34:96-101. 12. Slater A, Shann F, Pearson G. PIM2: a revised version of the Paediatric Index of Mortality. Intensive Care Med 2003;29:278-85. 13. Kenet G, Strauss T, Kaplinsky C, Paret G. Hemostasis and thrombosis in critically ill children. Semin Thromb Hemost 2008;34:451-8. 14. Monagle P, Chan AK, Goldenberg NA, Ichord RN, Journeycake JM, NowakGottl U, et al. Antithrombotic therapy in neonates and children: antithrombotic Therapy and Prevention of Thrombosis, 9th ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines. Chest 2012;141:e737S-801S. 15. Bateman ST, Lacroix J, Boven K, Forbes P, Barton R, Thomas NJ, et al. Anemia, blood loss, and blood transfusions in North American children in the intensive care unit. Am J Respir Crit Care Med 2008;178:26-33. 16. Arnold DM, Lauzier F, Rabbat C, Zytaruk N, Barlow Cash B, Clarke F, et al. Adjudication of bleeding outcomes in an international thromboprophylaxis trial in critical illness. Thromb Res 2013;131:2049.
6
White et al FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
■■ 2017
ORIGINAL ARTICLES
Bleeding in the PICU Instructions for Assessors: Complete this form daily while the patient was admitted to the PICU. USE A SEPARATE FORM FOR EACH UNIQUE SITE OF BLEEDING. * Required
1. Patient Study ID *
2. Study Collaborator *
3. Day of Admission in the PICU * Day of admission to the PICU = Day 1. Please do NOT include days admitted to other units in the hospital.
4. Bleeding Site * Choose the site of the bleeding event on this day. If the patient did not have a bleeding event on this day, select "No Bleeding Event." Mark only one oval.
Stop filling out this form.
No Bleeding Event
Vascular Catheter or Insertion Site
Skip to question 5.
Other Skin Bleeding (NOT CVC-Related) Skip to question 8.
Gastrointestinal
Respiratory (Including the Thoracic Cavity) Surgical Site Genitourinary
Skip to question 7.
Skip to question 9.
Skip to question 10. Skip to question 11. Skip to question 12.
Intra/Retroperitoneal Central Nervous System
Skip to question 13.
Eye, Ear, Nose, or Mouth
Skip to question 14.
Pericardium Joint
Skip to question 17.
Skip to question 15.
Other (Describe)
Skip to question 16.
Bleeding Site - Vascular or Catheter Insertion Site 5. Vascular or Catheter Insertion Site * Mark only one oval.
Peripheral Vein - Arm Peripheral Vein - Leg Femoral Vein Internal Jugular Vein Subclavian Vein
Other:
Appendix. Bleeding assessment tool. (Continues)
Epidemiology of Bleeding in Critically Ill Children FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
6.e1
THE JOURNAL OF PEDIATRICS • www.jpeds.com
Volume ■■
6. Specify Side * Mark only one oval.
Left Right
Skip to question 17.
Bleeding Site - Other Skin Bleeding (NOT CVC-Related) 7. Other Skin Bleeding Site(s) - NOT CVC-Related * Please select all that apply. Check all that apply.
Bruising Petechiae Non-Surgical Wound
Other: Skip to question 17.
Bleeding Site - Gastrointestinal 8. Gastrointestinal Bleeding Site * Please select all that apply. Check all that apply. Hematemesis Melena Hematochezia
Other: Skip to question 17.
Bleeding Site - Respiratory (Including the Thoracic Cavity) 9. Respiratory/Thoracic Cavity Bleeding Site * Please select all that apply. Check all that apply. ETT/Tracheostomy Aspirate Hemoptysis Hemothorax Other: Skip to question 17.
Appendix. Continues.
6.e2
White et al FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
■■ 2017
ORIGINAL ARTICLES
Bleeding Site - Surgical Site 10. Surgical Bleeding Site * Please select all that apply. Check all that apply. Incision Drain/Tube
Other: Skip to question 17.
Bleeding Site - Genitourinary 11. Genitourinary Bleeding Site * Please select all that apply. Check all that apply. Gross Hematuria Bleeding Around a Urinary Catheter Menstrual Bleed (that required medical consultation or intervention) Other: Skip to question 17.
Bleeding Site - Intra/Retroperitoneal 12. Intra/Retroperitoneal Bleed * Please describe the location of the bleed.
Skip to question 17.
Bleeding Site - Central Nervous System 13. Central Nervous System Bleed * Please select all that apply. Check all that apply.
Epidural Bleed Subdural Bleed Subarachnoid Bleed
Intraparenchymal Bleed Intraventricular Bleed Intraspinal Bleed (any site within the spinal column)
Other: Skip to question 17.
Appendix. Continues.
Epidemiology of Bleeding in Critically Ill Children FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
6.e3
THE JOURNAL OF PEDIATRICS • www.jpeds.com
Volume ■■
Bleeding Site - Eye, Ear, Nose, or Mouth 14. Eye, Ear, Nose, or Mouth Bleed * Please select all that apply. Check all that apply. Subconjunctival Bleed
Retinal Bleed Hemotympanum Epistaxis Oropharyngeal Bleed Other: Skip to question 17.
Bleeding Site - Joint 15. Bleeding in the Joint * Please specify which joint(s) is/are involved. Check all that apply. Shoulder Elbow Wrist Finger
Hip Knee Ankle Toe Other: Skip to question 17.
Bleeding Site - Other (Describe) 16. Other Bleeding Site * Please describe the location of the bleed (if possible).
Skip to question 17.
Appendix. Continues.
6.e4
White et al FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
■■ 2017
ORIGINAL ARTICLES
Bleeding Severity Please select AT LEAST one item from ONLY ONE of the bleeding severity categories below. Please select the most severe category as it applies to the bleeding event documented on this current day of admission. Remember to use a different form for another bleeding event on this same day but at a different site.
17. Major Bleeding Please check all that apply. Check all that apply.
Fatal Bleeding Clinically Overt Bleeding with Decrease in Hgb of >= 2g/dL in a 24-hr Period Bleed that is Intra/Retroperitoneal, Pulmonary, or Involves the CNS Bleeding Requiring Surgical Intervention in the Operating Room 18. Clinically Relevant Non-Major Bleeding
Please check all that apply. Check all that apply. Overt Bleeding Requiring Blood Product Administration Bleeding that Requires Medical or Surgical Intervention NOT in the Operating Room (and
Does NOT Include Menstrual Bleeding) 19. Minor Bleeding
Please check all that apply. Check all that apply. Overt or Macroscopic Evidence of Bleeding that Does NOT Fulfill Above Criteria (for Major Bleeding or Clinically Relevant Non-Major Bleeding)
Menstrual Bleeding Resulting in Medical Consult and/or Intervention Skip to question 20.
Timing 20. Type of Bleed * Mark only one oval. New Bleed Ongoing Bleed
Recurrent Bleed at the Same Site 21. Bleeding Start Date Please type which day of admission the bleed started.
22. Bleeding End Date
Please type which day of admission the bleed ended.
23. Started Prior to Today's Study Day? *
Mark only one oval. Yes No
Appendix. Continued.
Epidemiology of Bleeding in Critically Ill Children FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017
6.e5
THE JOURNAL OF PEDIATRICS • www.jpeds.com
Volume ■■
Figure 1. Frequency of bleeding based on severity.
6.e6
White et al FLA 5.4.0 DTD ■ YMPD8948_proof ■ February 6, 2017