First recognition of Clostridium difficile as a cause of pseudomembranous colitis occurred in 1978.1 The burden of disease due to C difficile has been increasing in the past decade with a doubling of incidence between 1991 and 2003.2 3 The cost of this disease has been estimated to exceed $1.1 billion annually, accounting for significant morbidity and mortality in hospitalised patients.4

Inflammatory bowel diseases (IBDs)—ulcerative colitis (UC), Crohn’s disease (CD)—are chronic relapsing disorders that often lead to hospitalisation. Patients with IBD can present with abdominal pain and diarrhoea requiring admission to hospital, a presentation similar to C difficile infection. Early reports about the role of C difficile in patients with IBD have been inconsistent. Whereas some authors identified C difficile infrequently in patients with IBD,5 6 other studies report finding C difficile as an incriminating cause in 5–19% of the patients admitted with an IBD flare.7^–10 Data have been published recently about an increase in C difficile associated disease (CDAD) in patients with IBD, especially UC.11 12 However, these data are from single tertiary academic medical centres where patients receiving care for their IBD may be more ill and have greater comorbidity. There are no national estimates across all hospitals about the burden of CDAD in patients with IBD or the excess of morbidity and mortality associated with it.

We performed this study with the following objectives: (a) to identify the morbidity and mortality associated with CDAD in hospitalised patients with IBD; (b) to compare this with patients admitted to hospital with a diagnosis of IBD not associated with C difficile infection; (c) to compare outcomes after hospitalisation with C difficile infection in patients with and without underlying IBD; and (d) to identify secular trends in the burden of CDAD from a nationwide sample of hospitalised patients between 1998 and 2004.

METHODS

Data source

The Healthcare Cost and Utilisation Project (HCUP) Nationwide Inpatient Sample (NIS) is the largest all-payer inpatient care database in the USA, and is maintained by the Agency for Healthcare Research and Quality. It covers about a 20% stratified sample of US community hospitals, including all non-federal short-term general and subspecialty hospitals, public hospitals and academic medical centres, based on ownership, hospital bed size, teaching status, US region and rural/urban location. For our primary analysis we used data from the HCUP NIS 2003, which contains data from 37 states and 994 hospitals, accounting for over 38 million discharges.

Definition of variables

Demographic information was obtained on age, race, sex, patient residence and insurance. Patients with age at admission <18 years (n = 1454) or >90 years (n = 520) were excluded from the unweighted survey sample. For each hospitalisation, the NIS contains one primary discharge diagnosis and up to 14 secondary diagnoses. Comorbid conditions were counted as present if they were listed among the secondary diagnoses for the hospitalisation. The burden of comorbid illness was assessed based on the Deyo modification of the Charlson comorbidity index.13 14 The Charlson index is a well-validated measure of comorbidity burden and has been modified to produce reliable estimates using ICD-9-CM codes from administrative datasets.

Definition of study groups

We defined three main study groups. Our primary case group (C difficile–IBD group) included patients who had a primary diagnosis of C difficile colitis (ICD-9-CM code 008.45) and a secondary diagnosis of either CD (ICD-9 CM codes 555.0, 555.1, 555.2, 555.9) or UC (ICD-9-CM codes 556.0, 556.1, 556.2, 556.3, 556.4, 556.5, 556.6, 556.8, 556.9). Patients were also included in this group if they had a primary diagnosis of CD or UC and a secondary diagnosis of C difficile colitis, since both these conditions have similar clinical presentations. This was done in order to capture all patients hospitalised for this combination of conditions. However, we performed a subanalysis by running the regression models while including only those patients with a primary diagnosis of C difficile infection and a secondary diagnosis of CD or UC as our main study group.

Patients admitted to hospital with a primary diagnosis of CD or UC without a diagnosis of C difficile colitis formed one of our comparison groups (IBD group). Our second comparison group included discharges with a primary diagnosis of C difficile colitis but no listed diagnosis of IBD (C difficile group). For the purpose of our primary analysis, we excluded discharges with only a secondary diagnosis of either C difficile or UC/CD because these patients might have been admitted to hospital for unrelated conditions that could influence their estimates of mortality or length of stay. However, we performed a second subanalysis including all patients with any listed diagnosis of C difficile or IBD as our control groups. ICD-9-CM discharge codes have been used to estimate reliably the burden of disease due to IBD,15 and have also been shown to correlate well with the results of C difficile toxin assays ordered for medical care during hospitalisation.16

Definition of outcomes

Our primary outcome was in-hospital mortality. Our secondary outcomes included length of hospital stay, total hospital charges, frequency of lower gastrointestinal (GI) endoscopy (sigmoidoscopy: ICD-9-CM codes 45.24, 48.23; colonoscopy: ICD-9-CM codes 45.23, 45.25, 45.22, 48.24), computed tomography of the abdomen and pelvis (ICD-9-CM code 88.01), parenteral nutrition (ICD-9-CM code 99.15), surgical procedures including small bowel resection (ICD-9-CM codes 45.64, 45.61, 45.62, 45.63, 45.33), small bowel anastomosis (ICD-9-CM codes 45.90, 45.91, 45.5, 45.50, 45.51), partial or total colectomy (ICD-9-CM codes 45.7×, 45.8) and large bowel anastomosis (ICD-9-CM codes 45.92–45.95).

Estimation of secular trends

For analysis of secular trends in the prevalence of CDAD in patients with IBD, we queried the NIS databases for the years 1998–2004. Cases with any listed diagnosis of CD or UC formed the denominator. The hospitalisations with coexisting diagnosis of C difficile and either CD or UC were included in the numerator to calculate the number of cases of C difficile each year per thousand admissions of UC or CD.

Statistical analysis

Statistical analysis was performed with Stata 8.0 (Stata Corp, College Station, TX, USA) using the strata weights provided in the NIS core file and the appropriate survey estimation commands. Analysis of variance and the χ² test for proportions were used to compare continuous and categorical variables between the three groups, respectively. Univariate logistic regression was used to identify predictors of in-hospital mortality and other dichotomous outcomes. Characteristics that were significantly associated with the outcome (p<0.10) were included in the final multivariate model. Univariate and multivariate linear regression were used for analysing length of stay and total hospital charges. The predictors were considered significant in the multivariate model at p<0.05. The χ² test for proportions was used to calculate the year-to-year differences in the proportion of CDAD complicating hospitalisation for UC or CD during the period 1998–2004.

The study was approved by the institutional review board of the Medical College of Wisconsin.

RESULTS

A total of 44 400 hospitalised patients had a primary diagnosis of C difficile, 77 366 a primary diagnosis of CD or UC, and 2804 a diagnosis of a combination of C difficile colitis and IBD.

Characteristics of the study population

Table 1 presents the characteristics of the three groups. Patients in the C difficile group tended to be older (median age 73 years) than patients in the IBD (42 years) or C difficile–IBD group (54 years) (p<0.05). Patients in the C difficile–IBD group had comorbidities intermediate between the other two groups. The C difficile group had the largest proportion of patients with a Charlson index of 3 or more (7.3%), compared with 2.4% and 0.8% in the C difficile–IBD and IBD groups, respectively.

Mortality was significantly higher in patients in the C difficile (3.7%) and C difficile–IBD group (4.2%) than in the IBD group (0.5%) (p<0.05) (table 2). Compared with patients admitted to hospital with IBD, patients in the C difficile–IBD group had a longer hospital stay, higher hospitalisation costs, and higher rates of lower GI endoscopy, but lower rates of bowel surgery. Compared with the C difficile only group, the patients with coexisting IBD tended to have more procedures, including higher rates of bowel surgery and lower GI endoscopy.

Multivariate analysis of in-hospital mortality

On univariate analysis, age, insurance and the Charlson index were significantly associated with mortality, in addition to C difficile and IBD status, and were included in the final model. Sex and race did not influence mortality and were not included in the final model. On multivariate analysis, patients older than 65 years were 10 times more likely to die than those in the age group 18–35 years (adjusted odds ratio (aOR) 10.0, 95% CI 4.8 to 20.9) (table 3). Patients with a Charslon index of ⩾ 3 had a threefold increased risk of death during the hospitalisation (aOR = 3.5, 95% CI 2.4 to 4.9) compared with those with no comorbidities. Patients in the C difficile–IBD group had a greater mortality than patients with IBD alone (aOR = 4.7, 95% CI 2.9 to 7.9) or C difficile alone (aOR = 2.2, 95% CI 1.4 to 3.4) even after adjusting for age and comorbidity.

Multivariate analysis of secondary outcomes

Patients in the C difficile–IBD group had a longer hospital stay and hospitalisation costs than patients with IBD alone (table 4) or C difficile alone (table 5). Patients in the C difficile–IBD group were more likely to undergo lower GI endoscopy (aOR = 1.9) or parenteral nutrition (aOR = 1.9) than patients with IBD, but there was no significant difference in blood transfusions or abdominal CT scans (table 4). The C difficile–IBD group was only half as likely to undergo bowel surgery as patients admitted to hospital for IBD alone (aOR = 0.6), but was much more likely to undergo surgery than those in the C difficile only group (aOR = 6.6). Restricting analysis to surgical procedures on the colon did not significantly change the odds ratios for either estimate.

Subanalyses

We performed a subanalysis restricting our case group to only those patients in the C difficile–IBD group with a primary discharge diagnosis of C difficile colitis and a secondary diagnosis of CD or UC. Patients in the C difficile–IBD group still had a higher odds of death than patients admitted to hospital with IBD alone (aOR = 5.7, 95% CI 2.9 to 11.3), which did not differ from the estimate for patients with a primary discharge diagnosis of IBD and a secondary diagnosis of C difficile colitis (aOR = 4.1). Compared with the IBD group, the C difficile–IBD group had longer hospital stays (2.0 days, 95% CI 0.8 to 3.2) but the revised odds ratio for lower GI endoscopy was no longer significant (aOR = 1.1, 95% CI 0.8 to 1.5). Also, patients in the C difficile–IBD group still had a higher mortality (aOR = 2.6, 95% CI 1.4 to 5.1), were more likely to undergo lower GI endoscopy (aOR = 2.5, 95% CI 2.1 to 2.9) or bowel surgery (aOR = 4.3, 95% CI 2.2 to 8.1) than patients with C difficile alone.

Our second subanalysis included patients with any listed diagnosis of C difficile as our control group in order to be more representative of most of the hospitalised patients who acquire C difficile after prolonged hospital stays for other conditions. In this analysis, the odds ratios for in-hospital mortality for the C difficile–IBD group was no longer statistically different from the C difficile group (aOR = 0.7, 95% CI 0.5 to 1.2), but the C difficile–IBD group still had higher rates of surgery (aOR = 4.4) or lower GI endoscopy (aOR = 7.1). Expanding the IBD group to include patients with any listed diagnosis of IBD yielded a higher mortality (aOR = 3.3) and use of endoscopy (aOR = 4.5) in the C difficile–IBD group, while the odds ratio for bowel surgery which had previously favoured the IBD alone group was now neutral (aOR = 1.1).

Comparison of UC and CD

Table 6 presents the differences in outcomes associated with C difficile disease in patients with UC (n = 1843) and CD (n = 961). There were significantly higher mortality, endoscopy and surgery rates in patients with UC than in those with CD, but no significant difference in length of stay or median hospital charge.

Secular trends

Figure 1 shows the secular trends in the estimated rate of hospitalisations for IBD that were associated with C difficile infection. From 1998 to 2004 there was a significant increase in the proportion of hospitalisations complicated by C difficile infection for both UC (24/1000 vs 39/1000, p<0.05) and CD (8/1000 vs 12/1000, p<0.05).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Secular trends in the proportion of estimated hospital admissions for inflammatory bowel disease (IBD)—ulcerative colitis (UC), Crohn’s disease (CD)—complicated by Clostridium difficile infection between 1998 and 2004.

DISCUSSION

C difficile is emerging as an important pathogen in hospitalised patients.2 3 While there are recent reports of increased susceptibility among patients with IBD in acquiring C difficile infection, our study shows that IBD may also be a risk factor for worse outcomes associated with C difficile infection compared with those without underlying IBD. Patients admitted to hospital with IBD who had a coexisting diagnosis of C difficile infection had higher mortality than patients admitted with IBD alone. Hospitalised patients with UC had more severe disease from C difficile than those with CD.

Comorbid illness17^–19 and severity of underlying illness20 have both been recognised as risk factors for C difficile infection.20 In our analysis we found that patients admitted to hospital with IBD who had associated C difficile infection were older and had more comorbid illnesses than patients hospitalised with a primary diagnosis of IBD alone. However, even after controlling for age and burden of comorbid illness, patients in the C difficile–IBD group had a higher mortality than those hospitalised for IBD alone. They also had longer hospital stays and a greater hospitalisation costs. Although it is possible that a longer hospital stay is a risk factor for, and not a consequence of, C difficile infection, two recent studies demonstrated that the average time to positive C difficile toxin assay in patients admitted to hospital with IBD was less than 48 hours suggesting that most infections were acquired before the hospitalisation.11 12 This suggests that the excess hospital stay we identified in our study is more likely to be a consequence of C difficile colitis complicating IBD. Patients with C difficile–IBD had a mean hospital charge in excess of $11 000 compared with those hospitalised with IBD alone or with C difficile alone, representing a significant healthcare burden. However, because of the nature of our study design, we were not able to demonstrate causality in this association, and suggest that further prospective research is required in this area to calculate the burden of excess hospitalisation that is attributable to C difficile infection in patients with IBD.

Patients in the C difficile–IBD group had a higher adjusted mortality, longer hospital stays and higher rates of lower gastrointestinal surgery and endoscopy than those with C difficile alone without underlying IBD. This suggests that underlying IBD might be a risk factor for not only acquisition but also worse outcomes after development of CDAD. To our knowledge, this finding has not been demonstrated previously.

Patients with UC appear to have an increased susceptibility to C difficile compared with those with CD, with a greater proportion of C difficile cases per 1000 admissions of patients with UC compared with CD (fig 1). Patients with UC also had higher mortality and higher rates of surgery and endoscopy than patients with CD. The exact reasons for this difference are as yet unclear. The universal colonic involvement in UC compared with less frequent involvement in CD may contribute to a higher susceptibility to C difficile infection, though small bowel infection with C difficile has also been reported in patients with IBD after surgery.21 We did not have information about concurrent drugs and immunosuppressant drug use to examine if this was a reason for the stronger association of C difficile with UC.

Our analysis of secular trends suggests a rising incidence of CDAD in patients with IBD, especially UC. However, the possibility of diagnostic bias must be considered in the interpretation of these data as it is possible that with rising awareness of CDAD, there might be a larger number of patients admitted to hospital with IBD who are being tested for the infection. Preliminary analysis of the summary data from HCUP NIS 2005 shows a continuing trend of increasing rate of C difficile infection in patients with UC (42/1000 discharges) or CD (15/1000 discharges), lending further importance to the need for recognition of the adverse outcomes associated with this infection in patients with IBD.

We examined the robustness of our definition of cases and controls with separate subanalyses, exploring different definitions for cases and controls. These analyses yielded results similar in direction to our original estimates for most of our end points, except that the mortality difference between the C difficile and the C difficile–IBD groups was no longer statistically significant. We offer the following as an explanation for this discrepancy. The group of patients with IBD who acquire C difficile (secondary diagnosis of C difficile) during hospitalisation have a number of other severe illnesses as shown by the high rate of comorbidity in this revised control group (22% with a Charlson index of ⩾3). In our opinion, these illnesses may influence their mortality and thus it is uncertain if the mortality estimate in this group is attributable significantly to C difficile alone. However, outcomes that may more directly be related to C difficile, such as the need for endoscopy or bowel surgery, continued to remain higher in the C difficile–IBD group.

There are several strengths to our study. To our knowledge, this is the first study examining the excess of morbidity and mortality associated with CDAD in patients with IBD from a nationwide representative sample. Using a national sample like the NIS avoids possible biases that may be seen in single-centre studies. We believe that by carefully limiting our primary control groups to patients for whom the principal diagnosis for the hospitalisation was identified on discharge to be C difficile or IBD, and excluding patients who might have been hospitalised for another primary diagnosis helps provide accurate estimates of the excess mortality and morbidity associated with C difficile infection in patients with IBD.

Our study also has several limitations. The major limitation is the inability to adjust for the severity of underlying IBD. Possibly, patients with severe manifestations of IBD are at a greater risk of both acquiring C difficile infection and of worse outcomes after development of disease. However, this information was not available to us for our analysis, and to our knowledge, there are no validated measures of determining IBD severity from large administrative databases or retrospective analysis of hospital discharge codes. We were also unable to determine the frequency of performing C difficile toxin assays or the timing of the positive assay with respect to the hospital course. It is also possible that some patients in the IBD alone group might have had mild C difficile disease but were not tested for the toxin, a finding which may weaken the magnitude of some or our associations. Studies using discharge coding data are susceptible to coding errors. However, we do not expect a systematic bias towards any particular group due to this. We were unable to differentiate those with asymptomatic carriage of C difficile from those with clinical disease. In addition, we were unable to acquire information on the use of antibiotics and other concurrent drugs, including proton pump inhibitors and immunosuppressive drugs. However, while these factors are well recognised to increase the risk of acquiring C difficile, their roles in determining patient outcomes associated with CDAD, which was the focus of our study, remain as yet undefined.3 22 We were also unable to assess the impact of different treatment regimens on C difficile infection. Finally, since the unit of study in the NIS is a hospital discharge and not individual patients, we were unable to differentiate between first infections and relapses or recurrences.

The clinical and policy implications of our study are many. C difficile infection has a significant impact on mortality and healthcare costs in patients with IBD nationwide. It is important for clinicians treating patients with IBD in all hospitals to have a high degree of awareness about this complication in these patients. The significant healthcare burden associated with C difficile in the IBD population necessitates a prudent use of antibiotics in this group to decrease the incidence of this complication. Early recognition and institution of appropriate treatment is essential to improve outcomes from this disease.

CONCLUSION

C difficile is an emerging pathogen with increasing importance in patients with IBD who may be particularly susceptible to this pathogen. Compared with patients admitted to hospital with IBD alone, those with CDAD have higher mortality and longer hospital stays. Patients with IBD have worse outcomes associated with C difficile infection than those without underlying IBD. Patients with UC have both higher rates of C difficile infection and worse outcomes than those with CD. There is an urgent need for research into the mechanisms of increased susceptibility as well as optimal treatment regimens for C difficile infection in patients with IBD.