Abstract

Introduction

Venous thromboembolism (VTE) is increasing in children, especially in the tertiary care setting. Hospital-associated VTE (HA-VTE) is a potentially preventable cause of major morbidity and mortality. However, the incidence of HA-VTE VTE is low in children Risk stratification tools may aid in identification of hospitalized high risk pediatric patients who may benefit from VTE prophylaxis.

Methods

We conducted a case-control study of pediatric patients with HA-VTE (21 years or younger at the time of diagnosis) admitted to the Johns Hopkins Hospital from 2008-2010. Cases were identified using ICD-9 codes for DVT and PE and verified by reviewing hospital records and radiologic imaging reports. HA-VTE was defined as: 1) VTE was diagnosed ≥48 hours after hospital admission without signs/symptoms of VTE on admission, or 2) VTE was diagnosed within 90 days of hospital discharge. Two contemporaneous controls matched for age, sex and admission unit were selected for each case. Records of cases and controls were reviewed for presence of a priori identified putative VTE risk factors at admission. Univariate and conditional multivariable logistic regression analyses with backward elimination were used to develop risk-prediction models. Based on results of univariate analysis, we sought to evaluate two multivariable models, one without length of stay (LOS) with relevance to assessment at admission, and one in which LOS was included with relevance to re-assessment after several days of hospitalization. All variables selected for the multivariable model were tested for interaction with a significance threshold level of p<0.2. Except for this, all hypothesis testing was two tailed and a p value of <0.05 was considered significant. Receiver operator curves (ROC) were constructed using risk factors on multivariate analysis.

Results

Table 1 lists the results putative risk factors by univariate analysis with a) significantly higher odds of VTE and b) higher odds of VTE but not statistically significant. In multivariable logistic regression analysis, central venous catheter (CVC), VTE predisposition and immobility or LOS >5 days were independently associated with HA-VTE. The combination of CVC and VTE predisposition with either immobility or LOS was predictive of HA-VTE (area under the curve for ROC of 76.6% and 80.6%, Table 2).

Table 1

Univariate analyses: Unadjusted odds of Hospital-Associated VTE. Variables meeting criteria for inclusion in multiple logistic regression model in boldface type.

Putative Predictor Variable Odds Ratio 95% Wald Confidence Limits P-value 
VTE Predisposition (Known thrombophilia or prior history of VTE) 4.1 1.5 11.0 <0.01 
Central Venous Catheter 7.6 3.3 17.4 <.0001 
Immobility 2.5 1.3 4.9 <0.01 
Intubation 2.3 1.1 5.2 <0.05 
Baseline/chronic Immobility 7.8 2.1 29.6 <0.005 
Bacteremia 2.5 1.1 5.8 <0.05 
Length of Hospital stay > 5 days 5.7 2.8 11.6 <0.001 
Every 1 additional hospital length of stay 1.0 1.0 1.0 <0.01 
Cardiac Pathology* 2.4 1.1 4.9 <0.05 
Any PICU/NICU stay 1.6 0.8 3.1 0.20 
Obese (weight >95th centile or BMI >30) 1.4 0.7 3.0 0.36 
Inflammatory disorder 1.5 0.3 7.1 0.59 
Systemic Infection 1.7 0.8 3.7 0.15 
Nephropathology 1.6 0.6 4.1 0.38 
Malignancy 1.5 0.6 4.0 0.39 
Surgery in last 90 days 1.5 0.8 2.9 0.25 
Putative Predictor Variable Odds Ratio 95% Wald Confidence Limits P-value 
VTE Predisposition (Known thrombophilia or prior history of VTE) 4.1 1.5 11.0 <0.01 
Central Venous Catheter 7.6 3.3 17.4 <.0001 
Immobility 2.5 1.3 4.9 <0.01 
Intubation 2.3 1.1 5.2 <0.05 
Baseline/chronic Immobility 7.8 2.1 29.6 <0.005 
Bacteremia 2.5 1.1 5.8 <0.05 
Length of Hospital stay > 5 days 5.7 2.8 11.6 <0.001 
Every 1 additional hospital length of stay 1.0 1.0 1.0 <0.01 
Cardiac Pathology* 2.4 1.1 4.9 <0.05 
Any PICU/NICU stay 1.6 0.8 3.1 0.20 
Obese (weight >95th centile or BMI >30) 1.4 0.7 3.0 0.36 
Inflammatory disorder 1.5 0.3 7.1 0.59 
Systemic Infection 1.7 0.8 3.7 0.15 
Nephropathology 1.6 0.6 4.1 0.38 
Malignancy 1.5 0.6 4.0 0.39 
Surgery in last 90 days 1.5 0.8 2.9 0.25 
*

Includes congenital heart disease (PDA, TOF, ASD, VSD), or poor cardiac function

Table 2

Odds of HA-VTE, multivariable logistic regression models derived by stepwise elimination.

MODEL 1 (on hospital admission) (AUC 76.6%)* Odds Ratios 95% Wald Confidence Limits P-value 
VTE Predisposition 4.4 1.5 13.2 <0.01 
CVC 7.8 3.1 19.6 <.0001 
Immobility 2.9 1.3 6.4 <0.01 
MODEL 2 (during hospital stay) (AUC 80.6%)*     
VTE Predisposition 3.4 1.1 10.4 <0.05 
CVC 6.0 2.4 15.4 <0.001 
LOS > 5 days 5.5 2.3 13.2 <0.001 
MODEL 1 (on hospital admission) (AUC 76.6%)* Odds Ratios 95% Wald Confidence Limits P-value 
VTE Predisposition 4.4 1.5 13.2 <0.01 
CVC 7.8 3.1 19.6 <.0001 
Immobility 2.9 1.3 6.4 <0.01 
MODEL 2 (during hospital stay) (AUC 80.6%)*     
VTE Predisposition 3.4 1.1 10.4 <0.05 
CVC 6.0 2.4 15.4 <0.001 
LOS > 5 days 5.5 2.3 13.2 <0.001 
*

AUC—Area Under the Curve for Receiver Operator Curve

Conclusion

We found independently associated risk factors with that may potentially be used in a predictive model of HA-VTE in children. Further prospective validation studies of these and other risk factors may serve as the basis of future risk-stratified randomized control trials of primary prevention of pediatric HA-VTE.

Disclosures:

Streiff:Bristol Myers Squibb: Research Funding; Sanofi: Consultancy, Honoraria; Eisai, Daiichi-Sankyo, Boehringer-Ingelheim, Janssen HealthCare: Consultancy. Strouse:NIH: Research Funding; Doris Duke Charitable Foundation: Research Funding; Masimo Corporation: Membership on an entity's Board of Directors or advisory committees, Research Funding. Takemoto:Novonordisk: Research Funding.

Author notes

*

Asterisk with author names denotes non-ASH members.