Venous thromboembolism, consisting of deep venous thrombosis and/or pulmonary embolism, represents a rare, but serious disease diagnosed in children. Paediatric hospital-associated venous thromboembolism is on the rise and children with congenital or acquired heart disease are at an increased risk for morbidity and mortality from venous thromboembolism.Reference Raffini, Huang, Witmer and Feudtner 1 In the past few years, novel retrospectively derived risk scores have emerged for hospital-associated venous thromboembolism in both critically ill and non-critically ill children; however, given their unique comorbidities, children with cardiac disease undergoing cardiothoracic surgery or cardiac catheter-based therapeutic intervention have not been addressed by these models. These previous studies have indicated that hospital-associated venous thromboembolism risk substantively differs between critically ill and non-critically ill children, but have classically included cardiac patients with all critically ill patients, remaining insufficiently powered for the determination of distinct risk factors for hospital-associated venous thromboembolism in this unique patient population.
Children with congenital or acquired heart disease who undergo cardiothoracic surgery or therapeutic cardiac catheterisation are particularly at risk for thromboembolic events and attendant morbidity and mortality.Reference Gruenwald, Manlhiot and Crawford-Lean 2 These palliative or corrective interventions may further perturb blood flow, increase activation and disruption of the endothelium and platelets, and promote hypercoagulability via perioperative or postoperative anticoagulant consumption and inflammatory response. Temporary or permanent placement of foreign or autologous material in the heart and vasculature may cause turbulent blood flow, resulting in disruption in vascular flow, promoting thrombosis. Further, patients with congenital cardiac disease can have immature coagulation systems that alter their ability to inhibit and promote clot formation, leading to increased risk for thromboembolism.Reference Monagle, Barnes and Ignjatovic 3 Thrombotic events in this patient population have been associated with significant morbidity and mortality, including increased length of hospital stay, and increased odds of early re-operation, early catheter re-intervention, cardiac arrest, and in-hospital mortality.Reference Manlhiot, Menjak and Brandão 4
In this study, we sought to identify risk factors for hospital-associated venous thromboembolism among critically ill children following cardiothoracic surgery or therapeutic cardiac catheterisation, via a single-institutional case–control study.
Methods
Subjects
This case–control study utilised the electronic health record-derived data warehouse at Johns Hopkins All Children’s Hospital to derive a data set for all admissions to Johns Hopkins All Children’s Hospital’s cardiovascular intensive care unit, with a current size of 22 beds, from 1 January, 2006, through 10 April, 2013. The study was approved by the Johns Hopkins Medicine Institutional Review Board at Johns Hopkins All Children’s Hospital, with a waiver granted for informed consent.
Hospital-associated venous thromboembolism cases were identified based on ICD-9 coding at the time of discharge via the electronic health record-derived data warehouse. Inclusion criteria consisted of the following: admission to the Johns Hopkins All Children’s Hospital’s cardiovascular intensive care unit during the study period; venous thromboembolism diagnosis based on ICD-9 coding at the time of discharge or on coding from an encounter within 30 days of discharge; and radiological confirmation – that is, diagnostic validation – of venous thromboembolism diagnosis by physician (C.M.A., N.A.G.) review of diagnostic imaging reports. Exclusion criteria included the following: signs or symptoms of venous thromboembolism noted in admission history and physical examination; or radiological evidence of venous thromboembolism within 24 hours of admission to hospital, without a history of hospital admission in the preceding 30 days. ICD-9 codes for venous thromboembolism included the following: 325, phlebitis and thrombophlebitis of intracranial venous sinuses; 437.6, non-pyogenic thrombosis of intracranial venous sinus; 452, portal vein thrombosis; 453.2, thrombosis of the inferior caval vein; 453.4, venous thromboembolism of deep vessels of the lower extremity; 453.8, embolism or thrombosis of other specified veins; 453.9, embolism and thrombosis of unspecified site; and 415.1, pulmonary embolism and infarction. Notably, we did not include cardiac shunt thrombosis in the definition of venous thromboembolism for this study, but did include other thrombi involving the right atrium, right ventricle, or superior vena cava, coded under 453.8 and 453.9. Diagnostic imaging modalities included compression ultrasound scanning with Doppler for extremity deep venous thrombosis, CT with pulmonary angiography for pulmonary embolism, echocardiography or cardiac catheterisation for intracardiac thrombosis, and CT or magnetic-resonance venography for caval, splanchnic, or cerebral sinovenous thrombosis, in accordance with published guidelines.Reference Chalmers, Ganesen and Liesner 5 For each hospital-associated venous thromboembolism case, two contemporaneous controls – as available – who were without venous thromboembolism, but met all other eligibility criteria above, were randomly selected from the electronic health record-derived data warehouse during the study period. The imaging studies of all controls were reviewed to exclude the possibility of evidence of thrombosis or emboli that would warrant potential classification as a venous thromboembolism case. Cases and controls were then limited to those in whom cardiothoracic surgery or therapeutic cardiac catheterisation was performed during the index hospitalisation. For a flow diagram of selection criteria, please see Results and Figure 1.
Flow diagram of study participants.
The study period was selected, in part, based on a consistent team of cardiovascular intensive care unit attending physicians, two consistent paediatric cardiothoracic surgeons, and a single consistent interventional cardiologist having managed the institution’s cardiac disease patient population throughout the study period. The institutional practice for the care of critically ill children before or following cardiothoracic surgery did not involve routine use of prophylactic anticoagulation; when used, the regimen consisted of a low-dose continuous IV infusion of unfractionated heparin at 10 U/kg/hour, without routine monitoring. In accordance with international consensus-based guidelines, an unfractionated heparin bolus of 75–100 U/kg was given as primary prophylaxis against femoral vascular thrombosis at the onset of cardiac catheterisation procedures, with activated clotting time monitoring every 30 minutes and repeat heparin bolus for activated clotting time <200. Antiplatelet therapy with aspirin at a dose of 3–5 mg/kg was routinely given once daily as primary thromboprophylaxis after the placement of intracardiac shunts, bioprosthetic valves, homograft right-ventricle-to-pulmonary-artery conduits, transcatheter atrial septal defect or ventricular septal defect closure, and placement of intravascular stents. In addition, warfarin was used for primary thromboembolic prophylaxis in children with mechanical heart valves, with a target international normalised ratio of 2.0–3.0.Reference Monagle, Chan and Goldenberg 6
Data collection
For both hospital-associated venous thromboembolism cases and controls, clinical data on demographics and putative risk factors for venous thromboembolism were extracted from the electronic health record-derived data warehouse. These data included the following: age; gender; venous thromboembolism-free length of hospital stay; previous history of venous thromboembolism; prior hospitalisation within 30 days; dehydration; PRISM-III scoreReference Pollack, Patel and Ruttimann 7 ; STAT mortality categoryReference Jacobs, Jacobs and Maruszewski 8 of the index cardiac surgical operation; obesity; use of a central venous catheter; CHD and cardiomyopathy status; therapeutic cardiac catheterisation; number and type of cardiothoracic surgeries; cardiac bypass during surgery; use of deep hypothermic circulatory arrest, mechanical ventilation, extracorporeal membrane oxygenation, and ventricular assist device; premature birth; and a history of rheumatic heart disease, major infection major non-cardiothoracic surgery, and chronic inflammatory disease. Data on cardiothoracic surgeries, therapeutic cardiac catheterisation, and other putative risk factors were collected and analysed in instances in which they had occurred or were present before the time of venous thromboembolism presentation.
Dehydration was identified based on its notation in the admission history. Obesity was also defined based on its notation in the admission history and/or physical exam, because of a high degree of missing data on patient height by which to calculate an accurate body mass index percentile, before 2012. Cardiothoracic surgery consisted of the following procedures: Norwood, Glenn, Fontan, Blalock–Taussig shunt, central shunt, heart transplant, truncus repair, transposition of the great arteries repair, tetralogy of Fallot repair, coarctation repair, ventricular and/or atrial septal defect repair, total anomalous pulmonary venous connection repair, and valve repair. Therapeutic cardiac catheterisation procedures included device closure of intracardiac shunts, coil embolisation of venous or arterial collateral vessels, balloon angioplasty or stent implantation in stenotic blood vessels, aortic or pulmonary balloon valvuloplasty, and stent implantation in occluded aorta-to-pulmonary-artery shunts. Major infection included meningitis, abscess, necrotising enterocolitis, pneumonia, osteomyelitis, bacteraemia, fungaemia, tracheitis, and pyelonephritis. The determination of tracheitis was based on clinical diagnosis documented in the medical record, and required that antibiotic treatment had been administered for this indication. Chronic inflammatory diseases included diabetes mellitus, inflammatory bowel disease, namely Crohn’s disease and ulcerative colitis, systemic lupus erythematosus, juvenile rheumatoid arthritis, graft versus host disease, and autoimmune encephalitis. Central venous catheters were further categorised as long-term – for example, Broviac, Mediport – versus short-term – for example, peripherally inserted central catheter (PICC), temporary subclavian, or femoral. In order to capture post-discharge venous thromboembolism episodes, patients in whom a radiological diagnosis of venous thromboembolism was made within 24 hours of admission were evaluated for a preceding admission within 30 days; for those in whom this was the case, data on the aforementioned venous thromboembolism risk factors were extracted from the electronic health record-derived data warehouse for the preceding admission. In order for a risk factor to be included in univariate and multivariate analysis, the risk factor had to have been present before the diagnosis or presentation of the venous thromboembolism.
Statistical analyses
Dichotomous or categorical variables were compared between cases and controls using χ2 analysis or Fisher’s exact test, as appropriate. Continuous variables were compared between groups using Mann–Whitney U testing. Putative risk factors were evaluated via univariate and multivariate logistic regression analyses, with 95% confidence intervals calculated for odds ratios using Firth’s penalised likelihood approach to overcome the problem of low prevalence of some of the risk factors. Although p-values of <0.05 were used to reflect statistical significance for all inferential statistics, a p-value threshold of ⩽0.1 was used in univariate analysis for inclusion of putative risk factors in a multivariate, adjusted model. Cardiomyopathy was not retained in the analysis because of a very low frequency of cases. Regarding children who underwent both cardiothoracic surgery and therapeutic cardiac catheterisation, the combination of the two procedures was coded as a unique variable, in order to separately determine whether this modulated hospital-associated venous thromboembolism risk. Positive likelihood ratios for hospital-associated venous thromboembolism in association with risk factor combinations were calculated using the formula: LR(+)=sensitivity/(1 − specificity).
Results
A flow diagram of included and excluded patients is provided in Figure 1. The final study population consisted of 57 hospital-associated venous thromboembolism cases and 76 contemporaneous controls who had undergone cardiothoracic surgery and/or therapeutic cardiac catheterisation, and met all other study eligibility criteria. As displayed in Table 1, venous thromboembolisms primarily consisted of lower- (42%) and upper- (40%) extremity deep venous thrombosis, with/without concomitant pulmonary embolism. At the time of venous thromboembolism diagnosis, 54% of venous thromboembolism cases were symptomatic, with patients demonstrating signs or symptoms concerning for venous thromboembolism that directly led to investigation via diagnostic imaging. The predominant diagnostic imaging modality was compression ultrasound scanning with Doppler (84% of venous thromboembolism cases; Table 2). Prophylactic anticoagulation, primarily with unfractionated heparin, was used in 6.8% of the study population, and was more prevalent among patients who subsequently developed venous thromboembolism – that is, in cases – than in controls.
Demographic and clinical characteristics of study participants.
CVC=central venous catheter; SD=standard deviation; VTE=venous thromboembolism
Because of rounding, not all percentages sum to exactly 100%
* For hospital-associated venous thromboembolism cases, included conditions are abscess (n=1), bacteraemia (n=12), fungaemia (n=2), surgical site infection (n=1), tracheitis (n=14), and missing (n=1); for controls, included conditions are bacteraemia (n=2), surgical site infection (n=1) and tracheitis (n=2)
** For indications, see Methods
Venous thromboembolism locations, symptomaticity, and diagnostic imaging modalities among cases (n=57).
DVT=deep venous thrombosis; PE=pulmonary embolism
Univariate logistic regression analyses (Table 3) revealed that age ⩽1 year, venous thromboembolism-free length of hospital stay, STAT mortality category of 5, central venous catheter, cardiomyopathy, therapeutic cardiac catheterisation with or without cardiothoracic surgery during the index hospitalisation, mechanical ventilation, cardiopulmonary resuscitation, major infection, and absence of major non-cardiothoracic surgery were each found to be statistically significant, unadjusted risk factors for hospital-associated venous thromboembolism. Univariate analyses also suggested that the use of unfractionated heparin was associated with hospital-associated venous thromboembolism (odds ratio=6.56, 95% confidence interval=1.26–34.1, p=0.025); however, because it could not be reliably distinguished via a retrospective study design as to whether administration of unfractionated heparin was used prophylactically versus in response to signs and symptoms of concern for possible hospital-associated venous thromboembolism, this variable was judged as unreliable, and, hence, not included in the modelling.
Results of univariate and multivariate analyses for putative hospital-associated venous thromboembolism risk factors.
CI=confidence interval; HA-VTE=hospital-associated venous thromboembolism; n/a=not applicable; OR=odds ratio; ref=reference category
Results are shown for those variables for which one or more univariate analyses met an a priori p-value threshold of ⩽0.1 (see also Methods)
* Parameter estimate imprecise after adjustment, because of the small number of patients who underwent cardiothoracic surgery without therapeutic cardiac catheterisation during the index hospitalisationStatistically significant independent risk factors are shown in bold.
After adjustment using multiple logistic regression, major infection (odds ratio=5.77, 95% confidence interval=1.06–31.4), age ⩽1 year (odds ratio=6.75, 95% confidence interval=1.13–160) and central venous catheter (odds ratio=7.36, 95% confidence interval=1.13–47.8) remained as statistically significant, independent risk factors for hospital-associated venous thromboembolism. Sensitivity analyses revealed that these remained independent risk factors for hospital-associated venous thromboembolism both with and without inclusion of unfractionated heparin administration in the model. The presence of all three risk factors was associated with a positive likelihood ratio of 15.3 (see Methods). On the basis of this measure, the risk for hospital-associated venous thromboembolism in critically ill children following cardiothoracic surgery or therapeutic cardiac catheterisation was found to be markedly increased – from 2%, a conservative estimate of pre-test probability using our observed prevalence of 2% in the cardiovascular intensive care unit study population at large, to 24%, post-test probability – among patients in whom all three risk factors were present.
Discussion
This study provides novel findings on hospital-associated venous thromboembolism occurrence and risk factors in critically ill children following cardiothoracic surgery or therapeutic cardiac catheterisation. We found major infection, infancy, and central venous catheter to be independent risk factors for hospital-associated venous thromboembolism, which, in combination, define a high-risk group. In a recent prospective study by Manlhiot et al. focussing specifically on patients undergoing cardiac surgery,Reference Manlhiot, Menjak and Brandão 4 risk factors associated with increased odds of venous thromboembolism included age <31 days, oxygen saturation <85%, history of thrombosis, heart transplantation, use of deep hypothermic circulatory arrest, postoperative use of ventricular assist device/extracorporeal membrane oxygenation, and duration of central venous catheter. Central venous catheter has been a consistently identified risk factor when considering that study, previous studies of hospital-associated venous thromboembolism risk in other paediatric sub-populations,Reference Atchison, Arlikar and Amankwah 9 – Reference Branchford, Mourani, Bajaj, Manco-Johnson, Wang and Goldenberg 12 a meta-analysis of hospital-associated venous thromboembolism risk factors among children in general,Reference Mahajerin, Branchford and Amankwah 13 and the present research. Although population-based evidence has demonstrated a peak in childhood venous thromboembolism risk during infancy, young age seldom has been implicated as a hospital-associated venous thromboembolism risk factor per se, except in the study by Manlhiot and the present work. Differences between our findings on hospital-associated venous thromboembolism risk factors in children with heart disease and those of Manlhiot, however, could be due, in part, to differences in study design and patient population. Unlike the Manlhiot study, the present analysis was retrospective and included not only children who had undergone cardiothoracic surgery but also those who had undergone therapeutic cardiac catheterisation.
Infection is a well-described, prevalent comorbid condition among cohorts and registries of paediatric venous thromboembolism,Reference Goldenberg, Knapp-Clevenger and Manco-Johnson 14 , Reference Monagle, Adams and Mahoney 15 and has been implicated as a hospital-associated venous thromboembolism risk factor in critically ill and non-critically ill children without cardiac disease.Reference Arlikar, Atchison and Amankwah 10 , Reference Amankwah, Atchison and Arlikar 11 Multiple mechanisms for infection as a risk factor for venous thromboembolism have been described/postulated, principally relating to the cross-talk between inflammatory and coagulation cascades. Clinically significant bacterial infection can result in acquired deficiencies of the native anticoagulants protein C, protein S, and antithrombin, and may also give rise to elevated levels of procoagulant factors such as fibrinogen and factor VIIIReference Goldenberg, Knapp-Clevenger and Manco-Johnson 14 or to the presence antiphospholipid antibodies as part of both the acute-phase response and subacute or chronic inflammation.Reference Sharathkumar, Goldenberg and Chan 16 In the present study, types of major infection observed in hospital-associated venous thromboembolism patients (see Table 1, footnote) included abscess (n=1), pneumonia (n=2), bacteraemia (n=16), fungaemia (n=2), surgical site infection (n=3), and tracheitis (n=19); however, we observed a rather low rate of necrotising enterocolitis in our study population, perhaps reflecting the postoperative feeding protocol and serving as an area of inquiry through ongoing quality improvement efforts at our institution.
Although neither cardiopulmonary bypass nor extracorporeal membrane oxygenation was found to be an independent risk factor for venous thromboembolism in the study population, in these settings, it is believed that the inflammatory response and activation of both platelets and the plasma-based coagulation cascade confer a prothrombotic state,Reference Lyle and Goldenberg 17 and it may be hypothesised that the impact of clinically significant infection on venous thromboembolism risk in this setting may be additive or even synergistic. This risk may be further compounded by immobility caused by critical illness and the high prevalence of indwelling central venous catheters.
Although not statistically significant, it is interesting that those patients in our study who developed hospital-associated venous thromboembolism tended to have had a higher STAT mortality category at the time of hospital admission/surgery. In addition, although also not statistically significant, hospital-associated venous thromboembolism cases also tended to have a higher prevalence of preceding mechanical ventilation. Interestingly, the latter observation is consistent with the senior author’s earlier findings on hospital-associated venous thromboembolism risk factors in critically ill children without congenital cardiac disease, from a different institution.Reference Branchford, Mourani, Bajaj, Manco-Johnson, Wang and Goldenberg 12
The strengths of our study include the following aspects: the team of cardiothoracic surgeons, interventional cardiologists, and intensivists was consistent throughout the study period, minimising the potential for treatment- and technique-related biases; diagnostic criteria did not change during the study period and the approach of case-identification and validation mirrored that of recent publications in paediatric hospital-associated venous thromboembolism risk modellingReference Atchison, Arlikar and Amankwah 9 – Reference Branchford, Mourani, Bajaj, Manco-Johnson, Wang and Goldenberg 12 ; cases were matched with controls with regard to date of admission, helping to mitigate the potential for time-period bias; and proper temporal relationship was enforced in the classification of putative risk factors – that is, putative risk factors were determined to be present before venous thromboembolism diagnosis.
Several potential limitations of our study warrant consideration, however. The single-institutional nature of the study, with attendant limitations in sample size, led to wide confidence intervals for some putative risk factors, emphasising the importance of substantiating the present findings via larger, multi-institutional studies. Another important limitation was the study’s retrospective design. Accordingly, our approach to venous thromboembolism case validation (see Methods above) from within an electronic health record-derived data set may not have been as reliable as that for venous thromboembolism within a prospective venous thromboembolism registry or cohort study. In addition, given the retrospective study design, our analyses were in some instances limited by missing data. For example, procedure length was not collected or easily accessible via our data warehouse, and should be assessed in future studies. In addition, PRISM-III score was able to be retrospectively calculated in 63.7% of our study population. Although we did not include vaso-inotropic score in our analysis, we did include both STAT mortality category and PRISM-III score as validated covariates for clinical severity, and also addressed the use of mechanical ventilation, extracorporeal membrane oxygenation, and ventricular assist device in our analyses, providing additional measures of severity of clinical condition. Nevertheless, vaso-inotropic score should be considered for inclusion in similar studies carried out in the future.
Use of heparin prophylaxis appeared to be positively associated with hospital-associated venous thromboembolism risk in univariate analyses, but was not included in multivariate analyses because of concern that this variable may have been confounded: see Results. Although sensitivity analyses suggested that the presence or absence of unfractionated heparin prophylaxis in the model did not substantively impact our findings, the unadjusted finding that hospital-associated venous thromboembolism risk was higher among patients receiving heparin prophylaxis suggests that such prophylaxis was insufficient in preventing hospital-associated venous thromboembolism. This observation highlights the potent thrombogenicity of cardiothoracic surgery, other invasive procedures, and even underlying disease in these patients, as described above. It also highlights the need for future risk-stratified studies on the safety and effectiveness of anticoagulation strategies as primary prophylaxis against hospital-associated venous thromboembolism in critically ill children undergoing cardiothoracic surgery or therapeutic cardiac catheterisation.
It must also be emphasised that shunt thromboses were not included among the hospital-associated venous thromboembolism events of interest in this study, such that the findings of our study cannot be extrapolated to patients with shunt thrombosis. Given their unique pathophysiology and their exclusion from this study, shunt thrombosis has been the subject of separate investigation for elucidation of risk factors and future research should be devoted, specifically, to risk models of shunt thrombosis. Further, it should be noted that asymptomatic venous thromboembolism are likely under-represented via our study approach, which, nevertheless, reflects a preference in study design to develop risk models for clinically significant venous thromboembolism events in order to ultimately inform future prevention trials. In addition, patients who underwent diagnostic cardiac catheterisation only, without cardiothoracic surgery or therapeutic cardiac catheterisation, were excluded from our study, such that our findings cannot be applied to such children; however, the focus on patients undergoing cardiothoracic surgery or therapeutic cardiac catheterisation served to reduce heterogeneity of the study population and thereby strengthen our analysis. Finally, because of sample size constraints, despite being a rather large single-institutional study, there was insufficient power to further separate the risk factor analysis into sub-populations of cardiothoracic surgery and therapeutic cardiac catheterisation. The possibility that differences in hospital-associated venous thromboembolism risk factor profiles may exist between the two sub-populations of critically ill children following cardiac intervention should be explored in future research, ideally via multicentre prospective collaborative cohort studies.
Notwithstanding these limitations, our study provides novel data regarding risk factors and a high-risk group for the development of hospital-associated venous thromboembolism in critically ill children following cardiothoracic surgery or therapeutic cardiac catheterisation. Because of the paucity of information on this vulnerable population and the increased risk for post-thrombotic complications, our study is of particular importance. Larger studies are needed to substantiate our findings and identify additional and novel risk factors, providing the foundation for future investigations of optimal risk-stratified approaches for hospital-associated venous thromboembolism prophylaxis in this population.
Acknowledgement
Dr Atchison’s work in this research was supported in part through a 2013 AOA Carolyn L. Kuckein Student Research Fellowship.
Conflicts of Interest
None.
