Inflammatory Bowel Disease and Thrombosis: A National Inpatient Sample Study


 Introduction Thrombosis is more common in inflammatory bowel disease (IBD) patients than the general population, but disease-specific correlates of thrombosis remain unclear.
 Methods We performed a retrospective analysis of discharge data from the National Inpatient Sample between 2009 and 2014, using International Disease Classification codes to identify IBD and non-IBD patients with or without thrombosis. We used NIS-provided discharge-level weights to reflect prevalence estimates. Categoric variables were analyzed by Rao-Scott Chi-square test, continuous variables by weighted simple linear regression, and covariates associated with thrombosis by weighted multivariable logistic regression.
 Results Thrombosis prevalence in IBD was significantly greater than in non-IBD, 7.52 versus 4.54%, p < 0.0001. IBD patients with thrombosis were older and more likely to be Caucasian than IBD without thrombosis, each p < 0.001. Thrombosis occurred most commonly in the mesenteric vein. Thrombotic risk factors in IBD include surgery, ports, malignancy, dehydration, malnutrition, and steroids at 53.7, 13.2, 13.1, 12.4, 8.9, and 8.2%, respectively. Those with thrombosis had greater severity of illness, 1.42 versus 0.96; length of stay, 7.7 versus 5.5 days; and mortality, 3.8 versus 1.5%; all p < 0.0001. Adjusting for age and comorbidity, odds ratios for predictors of thrombosis included ports, steroids, malnutrition, and malignancy at 1.73, 1.61, 1.34, and 1.13, respectively, while Asian race, 0.61, was protective, each p < 0.001.
 Conclusion Thrombosis prevalence is 1.7-fold greater in IBD than non-IBD patients. Adjusting for age and comorbidity, the odds ratio for thrombosis in IBD was 73% higher with ports, 61% higher with steroids, 34% with malnutrition, and 13% with malignancy. Whether long-term anticoagulation would benefit the latter is unknown.

increase in the platelet count that promotes thrombosis; at the same time, decreased expression of tissue plasminogen activator and increased levels of plasminogen activator inhibitor-1 contribute to decreased fibrinolysis. 7,8 Under these conditions, development of venous thromboembolism (VTE) is one of the most common extrasystemic manifestations of IBD, with a risk 2 to 3 times greater than in the general population. Moderate and severe disease flares, hospitalization, and surgery appear to increase thrombosis risk further. [9][10][11] Current guidelines suggest the use of pharmacologic VTE prophylaxis in IBD patients admitted to the hospital for any reason, especially if the admission is for a disease flare, even if accompanied by nonsevere GI bleeding, as the latter is not a contraindication to anticoagulation. 12 However, surveys of both American and Canadian gastroenterologists reveal a lack of institutional standards, with a wide variation in the application of these guidelines, many of which are based on low-quality evidence. 13,14 Morbidity and mortality remain elevated for these patients, and it is unclear what additional risk factors may place patients at risk for thrombosis. We analyzed the National Inpatient Sample to further characterize the prevalence and risk factors for thrombosis in hospitalized patients with IBD.

Methods Study Design and Data Source
This was a retrospective analysis of discharge data from the National Inpatient Sample (NIS) between January 1, 2009 and December 31, 2014. The NIS is an all-payer database that approximates a 20% stratified sample of discharges from U.S. community hospitals participating in the Healthcare Cost and Utilization Project. 15 It contains unweighted data from more than 7 million yearly hospital stays, and once weighted, estimates approximately 35 million stays. It includes deidentified clinical and nonclinical elements such as primary and secondary diagnoses, patient demographics, payment source, length of stay, and severity and comorbidity measures. Data during this time period were classified by the International Classification of Diseases, Ninth Revision (ICD-9) diagnostic codes.

Populations
Patient discharges, both with and without IBD and with and without thrombosis, were identified using ICD-9 diagnostic codes 555.0 (regional enteritis) and 556.0 (ulcerative colitis

Potential Risk Factors
The frequency of potential risk factors for thrombosis in patients with IBD was evaluated and identified using diagnosis codes: dehydration (276. 5 Deyo's modification of the Charlson's comorbidity index was used to assess the severity of illness between groups; by assigning weights to 16 diseases based on the strength of their association with mortality, the index controls for confounding.

Statistical Analysis
Thrombosis prevalence was estimated among admissions with and without IBD. The two groups were compared with respect to patient characteristics (age, race, and gender), insurance type, length of stay, inpatient mortality, and risk factors for thrombosis. These variables were further compared with and without thrombosis within each IBD group. Univariate analyses for between-group comparisons used Rao-Scott Chi-square test for categorical variables (e.g., gender and risk factors) and weighted simple linear regression for continuous variables (e.g., age). Weighted multivariable logistic regression was performed to identify factors associated with thrombosis among those with IBD, controlling for age and comorbidity. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated. Multivariable logistic regression was performed to identify factors associated with thrombosis among those with IBD. The multivariable model included covariates selected based on clinical and statistical significance. All analyses used discharge-level weights provided by NIS to reflect national estimates. All analyses were conducted in SAS (SAS version 9.4, SAS corporation), and pvalues < 0.05 were considered statistically significant. (7.51%, n ¼ 28,155) was 1.7 times higher than the prevalence of thrombosis in patients without IBD (4.53%, n ¼ 2,029,915), p < 0.0001. IBD patients had an increased prevalence of thrombotic risk factors including general surgery, abdominal surgery, dehydration, ports/central venous catheters, mal-nutrition, and long-term steroid use, all p < 0.0001, as compared with non-IBD patients (►Table 1). Markers of overall health, including the Charlson's score (9.99 vs. 1.25, p < 0.0001) and length of stay (5.63 vs. 4.58 days) were significantly worse, while inpatient mortality (1.66 vs. 1.90%, p < 0.0001) was significantly lower in IBD patients compared with patients without IBD (►Table 1). Among patients with IBD, those with thrombosis were significantly older (57.48 vs. 50.44 years, p < 0.0001), more likely to be Caucasian (81.54 vs. 79.68%, p < 0.0001), and less likely to be Asian (0.62 vs. 1.06%, p < 0.0001), than those without thrombosis. There was no significant gender effect on thrombosis prevalence in IBD patients (►Table 2).

Admission Characteristics
Among venous thromboses, mesenteric vein thrombosis was the most common type. Patients with IBD were nearly twice as likely to have mesenteric vein thrombosis (7.45 vs. 3.94%, p < 0.0001) or portal vein thrombosis (2.53 vs. 1.42%, p < 0.0001), as patients without IBD. By contrast, pulmonary embolism (6.46 vs. 10.50%) and deep venous thrombosis (6.21 vs. 7.96%) were significantly lower in IBD than in those without IBD, each p < 0.0001. Among arterial thromboses, Among all thrombosis patients, comorbid conditions including cirrhosis and liver disease were significantly more common in IBD patients, while hypertension, hyperlipidemia, cardiac disease, diabetes, renal failure, obesity, hepatitis C, and HIV were significantly less common, as compared with non-IBD patients; all p < 0.0001. Smoking was not different between the groups, while markers of overall health, including the Charlson's score (1.42 vs. 2.17, p < 0.0001) and inpatient mortality (3.76 vs. 4.47%, p < 0.0001) were significantly less severe and length of stay (7.75 vs. 6.72 days, p < 0.0001) longer in IBD patients with thrombosis (►Table 2).

Discussion
This cross-sectional study of discharges in the NIS cohort demonstrates that the prevalence of thrombosis is 1.7-fold greater in those with IBD as compared with those without IBD, consistent with previous studies. [1][2][3][9][10][11] The most common site was intra-abdominal, that is, mesenteric vein thrombosis, as previously reported. 16 IBD patients with thrombosis were younger, and, accordingly also had lower comorbidity score and lower in-hospital mortality. Thrombosis-specific risk factors, including steroid use, malnutrition, dehydration, port use, abdominal surgery, and general surgery, were all significantly more common in IBD than non-IBD thrombosis patients, consistent with previous studies. 10,[16][17][18][19][20][21] Malnutrition was likely associated with other risks for VTE, including requirement for port/central venous catheter, hospitalization, and/or surgery. 19 Further, women with IBD, similar to those without IBD, were significantly more likely than men to develop thrombosis, as previously reported, consistent with thrombosis risk in pregnancy and in the postpartum period. 22,23 Asian ethnicity was found to be protective, consistent with general risk of thromboembolism in this group. 24,25 By contrast, typical risk factors for VTE, including hypertension, hyperlipidemia, cardiac disease, diabetes, renal failure, and obesity, were significantly lower in IBD with thrombosis, likely, in part, related to their lower age.
Whether any risk groups for thrombosis among those with IBD would benefit from anticoagulation prophylaxis remains unknown. Thromboprophylaxis is typically avoided in patients with IBD due to concerns regarding GI tract bleeding. 26 Despite this, anticoagulation prophylaxis has been shown to be safe in IBD patients, with no greater bleeding than in non-IBD patients. 18,26,27 Further, it has been recommended that anticoagulation should be extended beyond acute thrombosis and/or hospitalization, 18,28,29 especially in those with frequent flares and chronic steroid use. 17,18,30,31 Our findings indicate that thrombosis risk exists in other subsets of IBD patients, that is, those using ports and undergoing general and  abdominal surgery. In fact, it has been shown that IBD patients receiving thromboprophylaxis within 24 hours of admission are half as likely to develop VTE. 18 However, more research is needed to identify those who might benefit from long-term anticoagulation. The observation that thrombosis site differs between those with IBD and those without IBD is of interest. Specifically, those with IBD were nearly twice as likely to have mesenteric or portal vein thrombosis, but significantly less likely to have pulmonary embolism, deep venous thrombosis, cerebral arterial embolism, or cerebral arterial thrombosis than in those without IBD. This suggests the possibility that the pathophysiology of IBD may involve prothrombotic signaling or pathways that promote local thrombosis in the GI tract. Further mechanistic studies of IBD thrombosis might lead to a better understanding of thrombosis, in general, and potentially identify new targets for thrombosis prevention.
Why mortality was significantly lower in individuals with IBD, compared with those without IBD is unknown. It is possible, but not proven that lower IBD mortality could be related to the lower prevalence of African American ethnicity, a group with poorer access to care and poorer health outcomes; or the lower IBD mortality could be related to a lower prevalence of malignancy and associated shortened survival. However, these are not proven, and more research is needed to determine the causes for reduced mortality in IBD.

Limitations
There are several limitations to this study. First, as the NIS is an inpatient database, there is potential bias toward a sicker population, as healthier patients who did not require admission are not included, introducing selection bias. Second, the NIS represents only 20% of the total inpatient population, so discharge-level-weights were used to determine a representative sample. Third, this sample is dependent on discharge diagnoses codes, which are limited by coding accuracy and potential misclassification bias, and further prevent tracking of the individual patient as the data are of discharge level rather than patient level. For the same reason, it is not possible to determine whether thrombosis at the patient level was provoked, symptomatic, or whether thromboprophylaxis was given. However, previous studies examining hospital discharge data have indicated sufficient accuracy for use in research studies. 32,33 Fourth, as this was a retrospective study, it is subject to bias and cannot be used to determine causality. Fifth, the NIS database does not contain laboratory values or drug treatment information, and thus it cannot be used to assess the relation between thrombosis risk and severity of IBD, duration of steroids or port use, or use of thromboprophylaxis, nor is it possible to adjust results for IBD duration or disease activity. A significant strength of the NIS is its size, this sample offers a large patient pool to assess an uncommon disease like IBD.

Conclusion
In conclusion, this study confirms that venous thrombosis is significantly increased among IBD patients. Recognition of those at high risk may help to identify potential patients for thromboprophylaxis safety and efficacy studies. Future trials will be critical to develop evidence for the optimal management of these patients.
Authors' Contributions J.B.C. and M.V.R. contributed to the study design, data acquisition, interpretation of the data and writing of the manuscript. D.M.C. contributed to data acquisition, performance of the data analysis, and critical review of the manuscript. J.G.Y. contributed to the study design, data acquisition, performance of the data analysis, and critical review of the manuscript. The dataset and analysis of the selected years from the NIS can be obtained by contacting the corresponding author.