Telephone: 414-805-0820; FAX: 414-805-0988;
Article first published online: 24 JUL 2012
Copyright © 2012 American Association for the Study of Liver Diseases
Volume 18, Issue 8, pages 972–978, August 2012
How to Cite
Ali, M., Ananthakrishnan, A. N., Ahmad, S., Kumar, N., Kumar, G. and Saeian, K. (2012), Clostridium difficile infection in hospitalized liver transplant patients: A nationwide analysis. Liver Transpl, 18: 972–978. doi: 10.1002/lt.23449
This project was presented as a research abstract at Digestive Diseases Week 2011.
This study was supported in part by the Clinical and Translational Science Institute of Southeast Wisconsin (National Institutes of Health grant UL1RR031973).
- Issue published online: 24 JUL 2012
- Article first published online: 24 JUL 2012
- Accepted manuscript online: 14 APR 2012 06:56AM EST
- Manuscript Accepted: 28 MAR 2012
- Manuscript Received: 12 JAN 2012
The incidence of Clostridium difficile infection (CDI) is increasing among hospitalized patients. Liver transplantation (LT) patients are at higher risk for acquiring CDI. Small, single-center studies (but no nationwide analyses) have assessed this association. We used the Nationwide Inpatient Sample of the Healthcare Cost and Utilization Project (2004-2008) for this retrospective, cross-sectional study. Patients with any discharge diagnosis of LT composed the study population, and they were identified with International Classification of Diseases, Ninth Revision, Clinical Modification codes. Those with a discharge diagnosis of CDI were considered cases. Our primary outcomes were the prevalence of CDI and the effects of CDI on inpatient mortality. Our secondary outcomes included the length of stay and hospitalization charges. A regression analysis was used to derive odds ratios (ORs) adjusted for potential confounders. There were 193,174 discharges with a diagnosis of LT from 2004 to 2008. The prevalence of CDI was 2.7% in the LT population and 0.9% in the non-LT population (P < 0.001). Most of the LT patients were 50 to 64 years old. LT patients had higher odds of developing CDI [OR = 2.88, 95% confidence interval (CI) = 2.68-3.10]. Increasing age and increasing comorbidity (including inflammatory bowel disease and nasogastric tube placement) were also independent CDI risk factors. CDI was associated with a higher mortality rate: 5.5% for LT patients with CDI versus 3.2% for LT-only patients (adjusted OR = 1.70, 95% CI = 1.29-2.25). In conclusion, the prevalence of CDI is higher for LT patients versus non-LT patients (2.7% versus 0.9%). CDI is an independent risk factor for mortality in the LT population. Liver Transpl, 2012. © 2012 AASLD.
Clostridium difficile infection (CDI) is an important cause of morbidity and mortality in hospitalized patients.1, 2 The past decades have seen an increasing incidence of CDI in North America.3, 4 C. difficile is a gram-positive, spore-forming anaerobe that causes toxin-mediated invasive infections.5, 6 Its clinical presentation varies from asymptomatic carriage to life-threatening colitis requiring surgical resection.6 Established risk factors for CDI include older age, increased comorbidity, and recent antibiotic or health care exposure, and immunosuppression, organ transplantation, and gastric acid suppression have recently been implicated.7, 8 There is increasing recognition of newer at-risk subgroups, including patients with inflammatory bowel disease (IBD), pregnant women, and patients with community-acquired CDI.
Liver transplantation (LT) patients are predisposed to CDI because of decreases in patient defense mechanisms resulting from a combination of debilitating disease, operative stress, immunosuppressants, and an imbalance in gut flora (due to antibiotic use).9-11 Various retrospective, single-center studies have placed the incidence of CDI in LT patients at 3% to 8%.12-14 Some studies have shown LT patients as a high-risk population for CDI.10 Albright et al.12 retrospectively studied charts from 467 LT procedures and found the prevalence to be approximately 8%, and they established that the early postoperative period (<28 days after transplantation) was associated with a higher risk of developing CDI. Other risk factors established in the study were a high Model for End-Stage Liver Disease score, an inability to bypass the intensive care unit after the operation, and postoperative vascular, biliary, or incisional complications. Hashimoto et al.13 retrospectively studied 242 living donor LT patients and identified an age > 55 years, male sex, and a serum creatinine level > 1.5 mg/dl as risk factors for CDI; their reported prevalence was 5%. Stelzmueller et al.14 studied solid organ transplant patients with CDI in a single-center study and found the prevalence of CDI to be 3.45% among LT patients. This was the highest prevalence in their cohort of single-organ recipients; only multivisceral transplant patients had a higher prevalence (20%).
There is a paucity of data on the nationwide prevalence of CDI in LT patients and its effects on inpatient mortality and health care resources. Our study aims were (1) to determine the prevalence of CDI in hospitalized LT patients, (2) to determine whether LT is independently associated with CDI in hospitalized patients, (3) to examine the effects of CDI on mortality in hospitalized LT patients, and (4) to compare the prevalence of CDI in the LT population with its prevalence in another solid organ transplant population (ie, the renal transplant population).
PATIENTS AND METHODS
Data Source and Study Design
The Nationwide Inpatient Sample (NIS) of the Healthcare Cost and Utilization Project (2004-2008) was used for this analysis.15 NIS is the largest all-payer inpatient discharge database. The patients are a 20% stratified sample of all the discharges occurring in a given year from approximately 1000 hospitals in 32 to 37 states (depending on the year of the study), and the database contains information on approximately 8 million discharges per year. A primary discharge diagnosis and as many as 14 secondary discharge diagnoses are listed for the patients. There can also be as many as 15 procedure codes associated with the discharges. We used a retrospective, cross-sectional study design.
International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes were used to identify patients from the database (Table 1). The study population consisted of patients with any LT discharge diagnosis code, whereas LT patients with a concomitant CDI discharge diagnosis code were considered cases. The use of ICD-9-CM codes for identifying patients with CDI has been validated in other studies, and good accuracy has been demonstrated.16, 17 Patients with no transplant codes were used as nontransplant controls. These codes have been validated in the literature for such analyses.16, 18-21 The subgroup of patients with an LT procedure code was also studied so that we could look at CDI during admissions related to the primary surgery. To examine whether the risk of CDI was specific to LT patients or was common to all patients with transplant-related immunosuppression, we selected a control population with a discharge diagnosis of renal transplantation.
|LT||V42.7, 50.51, 50.59, 996.82|
|Renal transplantation||V42.0, 55.61, 55.69, 996.81|
|Nasogastric tube placement||96.07|
|Acute liver rejection||996.82|
Definition of Variables
NIS contains demographic information, including age, sex, race, and primary and secondary insurance, for all hospitalizations. Patient comorbidity was adjusted with the Deyo modification of the Charlson index, which is a tool that has been validated for use in administrative databases.22, 23 We defined a new variable, any infection, for adjustment purposes as a surrogate for antibiotic use. For this variable, we included all ICD-9-CM codes used for a wide variety of infections (other than CDI and viral illnesses) that can potentially be treated with antibiotics. Similar coding was used by Martin et al.24 for a study of sepsis. The ICD-9-CM codes used for this variable are listed in the supporting information. Disposition from the hospital was also included in the analysis. Known risk factors for CDI (ie, nasogastric tube placement, IBD, and critical illness as indicated by mechanical ventilation) were used for the adjusted analysis.25-28 We could not determine antibiotic or proton pump inhibitor use from the database because of a lack of information on medication administration within NIS.
We analyzed the prevalence of CDI in hospitalized LT patients and the in-hospital mortality of admitted LT patients as well as secondary outcomes, which included the length of stay and hospitalization charges associated with LT plus CDI and the hospital discharge disposition. The discharge dispositions available from the database were home discharge, home discharge with care, facility discharge, and hospital discharge.
Categorical variables were compared with the χ2 test, and the t test was used for continuous variables. Logistic regression was used to determine the association of LT with CDI and to determine the effects of CDI on mortality in this cohort. The analysis was adjusted for variables such as age, sex, race, insurance, and comorbidities (including general comorbidity with the Charlson index and any infections). The analyzed outcomes included the mortality rate, the length of stay, the hospitalization charges, and the discharge disposition. For comparison, we performed a similar analysis for hospitalized renal transplant patients discharged in the same years. We also performed a subgroup multivariate analysis that included only LT-related and renal transplant–related admissions in the data. All analyses were performed with STATA 10 (Stata Corp., College Station, TX) and with the appropriate survey estimation command and strata weights provided in each NIS file. The study was approved by the institutional review board of the Medical College of Wisconsin.
Prevalence of CDI in LT Patients
There were 193,174 discharges for LT between 2004 and 2008; 5159 of these discharges (2.7%) were associated with CDI. In the nontransplant population, the prevalence of CDI was 0.9% (P < 0.001; Fig. 1). The prevalence of CDI in patients during the index admission for LT was 2.9%, which was not significantly different from the prevalence of CDI in patients during any LT-related admission. Among the LT patients, 7.6% also had a kidney transplant. The prevalence of CDI in patients receiving both liver and kidney transplants was 2.3%. The prevalence of CDI was 3.2% for LT patients with a concomitant code for acute rejection and 2.5% for patients with admissions not related to acute rejection (P < 0.001).
Characteristics of the Patients
In an unadjusted analysis, LT patients with CDI and LT patients without CDI had similar age distributions, with most patients being 50 to 64 years old. They were similar in terms of sex, race, and insurance providers. More LT patients with CDI were admitted to teaching hospitals (Table 2).
|Characteristic||LT only (n = 193,174)||LT Plus CDI (n = 5,159)||P Value|
|Home with care||13.2||20.6|
|Charlson index (%)||0.38|
|Hospital type (%)||0.01†|
|Hospital location (%)||<0.01†|
|Mechanical ventilation (%)|
|Length of stay (days)||7.7||17.8||<0.01†|
|Length of stay for survivors (days)||7.2||16.1||<0.01†|
|Hospital charges (×$1000)|
Factors Associated With CDI
According to an adjusted analysis, the odds of having CDI were almost 3 times higher for LT patients versus non-LT patients [odds ratio (OR) = 2.88, 95% confidence interval (CI) = 2.68-3.10]. This analysis was adjusted for any infections in the patients. Other risk factors for CDI were found to be increasing age and comorbidities, including mechanical ventilation, nasogastric tube placement, IBD, and any infections. Any race other than white was associated with a lower risk in the unadjusted analysis (Table 3).
The mortality rate of hospitalized patients who underwent LT and developed CDI was 5.5%, which was higher than the rate of 3.2% for non-LT patients (P < 0.001). There was no difference in the mortality rates for CDI patients during index admissions for LT surgery versus admissions not associated with LT surgery (6.7% versus 5.2%, P = 0.43). Teaching hospitals were associated with higher mortality rates for both LT patients and LT patients with CDI. CDI was an independent predictor of mortality in LT patients with an adjusted OR of 1.70 (95% CI = 1.29-2.25; Table 4). LT patients with CDI were also more likely to be on mechanical ventilation. The length of stay increased from a mean of 7.7 days (95% CI = 7.1-8.3) for LT patients to a mean of 17.8 days (95% CI = 15.6-20.0) for LT patients with CDI (mean difference = 9.9 days, 95% CI = 9.2-10.7). LT with CDI was also associated with almost twice the hospitalization charges (mean difference = $69,131, 95% CI = $60,652-$77,610). Restricting our analysis to patients who survived hospitalization, we found similar increases in the length of stay and hospitalization charges (Table 2). In comparison with LT patients, LT patients with CDI were more likely to be discharged to a facility or require home care.
Comparison With Renal Transplant Patients
We performed a comparative analysis of renal transplant patients to determine whether the association of CDI with LT is exclusive to LT or is associated with a transplant state. The prevalence of CDI in the renal transplant population was 2.1%, which was significantly lower than prevalence in the LT population (P < 0.001). A renal transplant state was also independently associated with the risk of developing CDI (OR = 2.81, 95% CI = 2.70-2.93). The mortality rate of renal transplant patients with CDI was 5.0%, which was not significantly lower than rate for the LT population with CDI (5.5%, P = 0.15). In our subgroup multivariate analysis including only LT patients and kidney transplant patients, LT remained independently associated with CDI (OR = 1.28, 95% CI = 1.11-1.48).
The incidence of C. difficile is increasing in hospitalized patients. Recent single-center studies have shown an increased predisposition in LT patients,7, 12, 14 but these studies have been limited by their sample sizes and single-center perspective. Using a nationally representative database, we demonstrated the following: (1) CDI occurred more frequently in LT patients versus nontransplant controls as well as renal transplant patients, and (2) CDI was an independent risk factor for mortality in LT patients and was associated with significant increases in the length of stay and hospitalization charges.
We found the prevalence of CDI in LT patients to be 2.7%. Various studies have placed the prevalence anywhere from 3% to 8%.12-14, 29 Stelzmueller et al.14 retrospectively studied solid organ transplants at a single center and found the prevalence of CDI in LT patients to be 3.45%. A multiyear, retrospective study by Albright et al.12 from the Mayo Clinic found the prevalence of CDI in LT patients to be 8%; they also found that patients in the early posttransplant period (<28 days) had the highest risk of developing CDI. Hashimoto et al.13 reported a prevalence of 5% and a median time to CDI onset of 19 days. The fact that our study showed a prevalence near the lower end of previous study estimates may be explained by the fact that studies from single centers may be more likely to reflect referral populations with greater comorbidity and a higher risk for CDI. Although our study lacks data on the timing of infection, the subgroup analysis, which showed a slightly higher prevalence for index admissions (2.9%), corroborates reports from other studies: patients in the early postoperative period have the highest risk for developing CDI. The earlier onset of CDI may be related to the more potent immunosuppression used in the immediate postoperative period30 and the debilitated state of patients before transplantation.10 The admissions related to acute rejection had a significantly higher prevalence of CDI than the admissions not associated with acute rejection, and this was likely due to the increased immunosuppression required to treat acute rejection.
Increasing age was found to be a risk factor for CDI in our study; this is in line with the literature, in which advanced age is a well-established risk factor for CDI.25, 31, 32 Although race was not a significant factor in the unadjusted analysis, white race was found to be a risk factor for CDI in LT patients in the final model. There are no correlating data in the literature indicating that white race is a risk factor for CDI. Females had a slightly higher risk of developing CDI among LT patients. Studies are mixed in terms of the sex-based risk for CDI: some have shown that males are more likely to develop CDI, whereas others have shown that females are.13, 33, 34 The significance of these contrasting findings is unclear. More LT patients with CDI were admitted to teaching hospitals versus nonteaching hospitals, and this is possibly due to the fact that these patients were referred to transplant centers or teaching hospitals because of their inherent complexity.
According to an adjusted analysis, LT was independently associated with CDI. This was found after adjustments for various variables such as age, race, and general comorbidity (Charlson index) as well as known risk factors for CDI that could be taken from the database (including nasogastric tube placement, mechanical ventilation, and IBD). Proton pump inhibitor use and antibiotic use were not used in the analysis because of a lack of information on medications. The any infection variable was also used in this adjustment as a proxy for antibiotic use. A similar variable was used by Martin et al.,24 but their list was less comprehensive than the set of infections included in our study. The use of this variable makes the association between LT and CDI only stronger: we hypothesize that if this variable had not been used, the association would likely have been weaker because it would not have been adjusted for these multiple infections.
CDI was found to be an independent risk factor for mortality in hospitalized LT patients, and this emphasizes the importance of a high index of suspicion for an early diagnosis and the appropriate initiation of treatment. In addition to being a risk factor for mortality in this population, CDI also was a burden on health care because it increased the length of stay and hospital costs almost 2-fold. It also altered the discharge disposition for patients. Fewer CDI patients were able to be discharged home, and more patients required discharges to a facility or home with care. This highlights the burden on the health care system posed by this complex cohort of patients. One reason could be that this cohort of patients already has multiple comorbidities, and the added need for isolation and antibiotic treatment alters the disposition to a care facility or home with care.
Most LT patients have underlying cirrhosis before transplantation, and studies have shown that patients with cirrhosis have a higher risk of developing CDI. Bajaj et al.18 analyzed a nationwide database as well as tertiary care center data and found that CDI was associated with a higher mortality rate, a longer length of stay, and higher hospital charges.18 This underlying or pretransplant comorbidity likely contributes to the risk of CDI in the LT population.
We compared the risk of developing CDI in LT patients and renal transplant patients and found that renal transplantation was also a risk factor for developing CDI, although the prevalence of CDI was higher in LT patients versus renal transplant patients. In Stelzmueller et al.'s study,14 the prevalence of CDI was also higher among LT patients versus renal transplant patients. The exact reason for this finding is unclear because the liver is considered less immunogenic than the kidney and requires less immunosuppression after transplantation. This finding could result from differences in patient demographics or the differential use of antibiotics due to other comorbidities in the 2 populations. Our subgroup multivariate analysis with LT patients and kidney transplant patients only indicated that LT patients have a higher risk of CDI than kidney transplant patients. The reasons for this could be multifactorial and include residual effects of pretransplant cirrhosis, different immunosuppression regimens after LT versus kidney transplantation, and different usage of antibiotics (predisposing to CDI) after LT versus kidney transplantation. These findings emphasize that the comorbid state and immunosuppressants play roles in the risk for CDI.10, 14, 33
Our study had several limitations. There were no patient identifiers, so patients could not be tracked over time, and we could not account for multiple admissions by the same patient. We could not look at the timing of the infection because of a lack of such information, and we also could not examine the impact of recurrent CDI, which may have been seen in up to a third of the patients. Because of the nature of the database and the resulting lack of laboratory data, the Model for End-Stage Liver Disease score and proton pump inhibitor and antibiotic use could not be determined. Because of the administrative nature of the database, there may be coding errors leading to missed or erroneous diagnosis coding. However, we would expect such errors to be nondifferential between the various groups, and we would not anticipate any significant alterations in our results. As in many retrospective observational studies, effect sizes reflected associations and could not definitively lead to conclusions on causation. Nevertheless, our findings were robust to adjustments for multiple comorbidities that likely influenced effect sizes. However, the effects of unmeasured confounders are recognized, and our findings merit cautious interpretation.
In conclusion, with a nationwide analysis of LT patients, we identified a 2.7% prevalence of CLD in LT patients. LT was found to be independently associated with CDI in hospitalized patients, and CDI was an independent predictor of mortality in hospitalized LT patients. We suggest vigilance in prevention and a low threshold for the assessment of CLD in hospitalized LT patients, particularly in the early postoperative period. We suggest that patients with acute rejection, increasing age, white race, mechanical ventilation, IBD, and any infections be more vigilantly screened for CDI and possibly be considered for more active probiotic prophylaxis during antibiotic therapy.
- 3Clostridium difficile hospitalizations in Louisiana: a 10 year review. J La State Med Soc 2011; 163: 192-195., .
- 9Gastrointestinal infections after hematopoietic stem cell or solid organ transplant. In: Bowden RA, Ljungman P, Paya CV, eds. Transplant Infections. Philadelphia, PA: Lippincott Williams & Wilkins; 2003: 198-221., ., Chapter 15:
- 10Infections caused by Clostridium difficile in kidney or liver graft recipients. Ann Transplant 2005; 10: 70-74., , , , , .
- 15Healthcare Cost and Utilization Project. Overview of the Nationwide Inpatient Sample. http://www.hcup-us.ahrq. gov/nisoverview.jsp. Accessed November 2011 and March 2012.
Additional Supporting Information may be found in the online version of this article.
Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.