Clostridium difficile colitis (CDC) is associated with an increased short-term mortality risk in hospitalised ulcerative colitis (UC) patients. We sought to determine whether CDC also impacts long-term risks of adverse health events in this population.
To determine whether CDC also impacts long-term risks of adverse health events in this population.
A population-based retrospective cohort study was conducted of UC patients hospitalised in Ontario, Canada between 2002 and 2008. Patients with and without CDC were compared on the rates of adverse health events. The primary outcomes were the 5-year adjusted risks of colectomy and death.
Among 181 patients with CDC and 1835 patients without CDC, the 5-year cumulative colectomy rates were 44% and 33% (P = 0.0052) and the 5-year cumulative mortality rates were 27% and 14% (P < 0.0001) respectively. CDC was associated with a higher adjusted 5-year risk of mortality [adjusted hazard ratio (aHR) 2.40, 95% CI 1.37–4.20], but not of colectomy (aHR 1.18, 95% CI 0.90–1.54). CDC impacted mortality risk both during index hospitalisation (adjusted odds ratio 8.90, 95% CI 2.80–28.3) as well as over 5 years following hospital discharge among patients who recovered from their acute illness (aHR 2.41, 95% CI 1.37–4.22). Colectomy risk was not influenced by CDC in this cohort.
Clostridium difficile colitis is associated with increased short-term and long-term mortality risks among hospitalised ulcerative colitis patients. As colectomy risk is not similarly impacted by Clostridium difficile colitis, factors predictive of death among C. difficile-infected ulcerative colitis patients require elucidation.
Clostridium difficile colitis (CDC) is the leading cause of nosocomial infectious diarrhoea and is associated with substantial morbidity and mortality. The past decade has witnessed a rise in the incidence and severity of CDC in hospitals across North America and Europe, largely attributable to the emergence of highly virulent strain of this organism.[1, 2]
Patients with ulcerative colitis (UC) are at increased risk of acquiring CDC.[3-5] The incidence of CDC among UC patients has also been rising in recent years, with CDC now being implicated in up to 5% of UC hospital admissions.[3-5] Nationwide studies have further reported that CDC is associated with a 3.8–5.5-fold higher acute mortality risk among hospitalised UC patients.[3, 6] The impact of CDC on acute colectomy risk in UC patients has been less consistent across studies.[3, 6-8]
In addition to increasing short-term health risks, CDC may pose sustained health risks to UC patients. Chronic C. difficile colonisation following an acute infectious episode may increase the risk of future CDC episodes in many patients. Moreover, the impact of CDC on UC disease behaviour remains unknown. One study reported that CDC was associated with a greater 1-year colectomy risk in these patients. However, no study has assessed the longer term impact of CDC on health outcomes in this population. As UC afflicts more than 500 000 individuals across North America, this could have significant implications for hospital care policy and health resource utilisation in these patients.[10, 11] Therefore, we sought to investigate the impact of CDC on 5-year health outcomes among hospitalised UC patients.
Materials and Methods
Data for this study was acquired primarily from the Ontario version of the Canadian Institutes of Health Information Discharge Abstract Database (CIHI-DAD), which contains demographic and health information pertaining to all patient hospitalisations throughout Ontario. Additional information was obtained from The Ontario Health Insurance Plan (OHIP) database, The Ontario Registered Persons Database (RPDB), The Ontario Cancer Registry (OCR) and Ontario Census Databases. Patient information was linked across hospitalisations and physician encounters and between databases at the Institute for Clinical Evaluative Sciences (ICES), permitting detailed patient profiles and longitudinal health outcome assessments (http://www.ices.on.ca). This study was approved by the ICES Privacy Office.
Patients and study design
A retrospective population-based cohort study was conducted of UC patients hospitalised at acute care hospitals in Ontario, Canada, between 31 March 2002 and 31 March 2008. The study start date coincided with the rising incidence of highly virulent C. difficile infectious outbreaks in North American hospitals and also with the onset of diagnostic reporting using ICD-10 codes in CIHI-DAD.[1, 2] Patients were eligible for study if they were given a ‘Most Responsible Diagnosis (MRD)’ of UC (K51.x in ICD-10) or else a MRD of CDC (A04.7 in ICD-10) with a co-morbid or secondary diagnosis of UC on a hospital discharge abstract corresponding to any hospitalisation within the study period. The MRD in CIHI-DAD corresponds to the acute illness that most significantly affected a patient's treatment course or accounted for the greatest length of hospital stay (LOHS). A co-morbid diagnosis is one that contributed significantly to a patient's in-hospital course, whereas a secondary diagnosis refers to a preadmission co-morbidity that did not complicate a patient's hospital course. A subgroup of patients who were discharged from hospital without undergoing colectomy was also evaluated, as it remains unknown whether CDC alters long-term disease behaviour specifically among UC patients who recover from their acute illness.
Patients were excluded from the study for the following indications: (i) Age younger than 18 years; (ii) Invalid identifying information in CIHI-DAD; (iii) Permanent residence outside of Ontario; (iv) Initial UC-related hospitalisation within the study period for elective indication (based on a specific designation for admission type in CIHI-DAD); (v) Hospitalisation for CDC within the previous 5 years; (vi) History of partial or total colectomy (using a 5-year look-back window), and (vii) Diagnosis of colorectal cancer from 5 years prior to index hospitalisation to the end of the study period.
To increase the positive predictive value (PPV) of a UC diagnosis, a published algorithm for identifying UC patients within health administrative data from the province of Manitoba, Canada, was applied to generate the final study cohort. This algorithm required that patients had at least five UC-related physician contacts if they were registered with OHIP for 2 or more years or at least three UC-related physician contacts if they were registered with OHIP for less than 2 years (UC-related hospitalisation counted as a single contact). Physician contacts were identified in the CIHI-DAD and OHIP databases. The sensitivity and specificity of this definition in the original study were 74.4–87.7% and 91.3–93.7% respectively. Notably, this definition has not been specifically validated for a hospitalised cohort or within Ontario health administrative data.
Patients who were given a MRD or a co-morbid diagnosis of CDC at any hospitalisation during the study period were evaluated from the time of their first hospitalisation with this diagnosis (‘UC-CDC’ group). These patients were compared with those who were not given a diagnosis of CDC in any hospitalisation during the study period (‘UC-noCDC’ group) on the rates of adverse health events over 5 years following index hospitalisation, with a final follow-up date of 31 March 2010.
The primary outcomes were the adjusted 5-year risks of colectomy and death. Other outcomes assessed included: (i) The 5-year risks of colectomy, death, non-elective UC-related hospital readmission, and non-elective all-cause hospital readmission, among colectomy-free discharged patients; and (ii) The risks of colectomy and death and the acute care length of hospital stay (LOHS) during index hospitalisation.
The associations between CDC and adverse outcomes were adjusted for the potentially confounding effects of age, gender, weighted Charlson co-morbidity score, prior UC-related hospitalisation (using a 5-year look-back window), UC duration (using an 8-year look-back window), hospital teaching status, annual hospital UC admission volumes and year of hospitalisation. LOHS was also adjusted for the effects of in-hospital colectomy and in-hospital death. Prior UC-related hospitalisation and UC duration were used to partially adjust for baseline UC severity. Information on all variables was obtained solely from CIHI-DAD, except for the variable ‘UC duration’ for which both CIHI-DAD and OHIP databases were surveyed.
Bivariate comparisons of in-hospital outcomes were conducted using the Chi-squared test (or Fisher's exact test) for categorical variables and the Wilcoxon rank-sum test for interval variables. The Kaplan–Meier product limit method with log-rank test was used to compare time-to-event outcomes. To evaluate the adjusted effects of CDC on 5-year (longitudinal) and in-hospital (binary) outcomes, while simultaneously accounting for clustering of patients within hospitals, Cox proportional hazards models with robust variance estimates and generalised estimating equation (GEE) methods (to estimate logistic regression models) were used respectively. Negative binomial regression analysis was used to model acute care LOHS. A stepwise modelling approach was used when evaluating death outcome to ensure a minimum 1:10 ratio of covariates to outcome events.[14, 15] As such, hospitalisation year was excluded from the analysis of 5-year mortality risk (132 deaths), whereas no additional covariates (aside from CDC status) were included in the analysis of in-hospital mortality risk (13 deaths). Statistical significance was based on a 2-sided type I error rate of 5%. All analyses were performed using sas 9.1 software (SAS Institute Inc., Cary, NC, USA).
Patients were censored at the time of colectomy in the analysis of 5-year mortality risk so that only patients with intact colons who were at risk of experiencing UC-related adverse events were analysed. In addition, patients who underwent colectomy on the date of rehospitalisation were censored on that date in the analyses of non-elective hospital readmission. Otherwise, patients were censored at the last study follow-up date. Loss to follow-up in this study was assumed to be negligible based on an annual Ontario emigration rate of 0.18% (http://www40.statcan.gc.ca/l01/cst01/demo33b-eng.htm).
Of 5743 UC patients who were hospitalised for UC or CDC during the study period, 2016 met entry criteria. Of excluded patients, 2278 did not satisfy the study case definition for UC, 478 were under 18 years of age, 580 were admitted for elective large bowel surgery at index hospitalisation; 155 had undergone prior partial or total colectomy; 130 had been previously hospitalised for CDC within the preceding 5 years; 100 were identified as having prevalent colorectal cancer, 4 were not Ontario residents and 2 had invalid identifying information.
In the study cohort, 181 patients (9.0%) were diagnosed with CDC during hospitalisation. Baseline characteristics of study patients are presented in Table 1. Eleven patients had missing information on one or more variables in this study. The cumulative 5-year rates of colectomy and death in the overall cohort were 34% and 15% respectively. In addition, 10% of patients underwent colectomy and 0.64% of patients died during index hospitalisation. Among patients who were discharged from hospital without undergoing colectomy, 46% were rehospitalised for a UC-related indication, and 70% were rehospitalised overall up to 5 years following hospital discharge.
Table 1. Baseline characteristics of study patients
Among hospitalised UC patients with and without CDC, the cumulative 5-year colectomy rates were 44% and 33% (P = 0.0073) and the cumulative 5-year mortality rates were 27% and 14% (P < 0.0001) respectively (Figures 1 and 2).
Following covariate adjustment, CDC was associated with a higher 5-year mortality risk [adjusted hazard ratio (aHR) 2.40, 95% CI 1.37–4.20], but was not associated with a higher 5-year colectomy risk (aHR 1.18, 95% CI 0.90–1.54) (Table 2). Male gender (aHR 1.27, 95% CI 1.09–1.48), prior hospitalisation for UC (aHR 1.36, 95% CI 1.09–1.70), UC duration greater than 2 years (aHR 1.31, 95% CI 1.11–1.55) and admission to a teaching hospital (aHR 1.66, 95% CI 1.27–2.17) were independently associated with colectomy risk, whereas age (aHR 14.0, 95% CI 4.28–45.7 for age 40–70 vs. age <40; aHR 91.9, 95% CI 27.5–308 for age >70 vs. age <40) and co-morbidity burden (aHR 1.84, 95% CI 1.10–3.06 for Charlson score 1 vs. 0; aHR 4.28, 95% CI 2.68–6.82 for Charlson score ≥2 vs. 0) were strongly associated with mortality risk in this cohort (Table S1).
Table 2. Adjusted in-hospital and 5-year risks of colectomy and death with CDC among hospitalised UC patients
Clostridium difficile-infected patients had a similar in-hospital rate of colectomy (12% vs. 9.8%, P = 0.30), but a higher in-hospital rate of death (3.3% vs. 0.38%, P < 0.0001) as compared with uninfected patients. There was no difference in the median time-to-colectomy between patients with CDC (2.5 days, IQR 1.0–6.0 days) and those without CDC (2.0 days, IQR 1.0–6.0 days) (P = 0.75). No in-hospital deaths occurred among patients who underwent colectomy in this cohort. The acute care LOHS was 11 days [interquartile range (IQR) 6–19 days] in patients with CDC and 6 days (IQR 4–11 days) in patients without CDC (P < 0.0001).
Following covariate adjustment, CDC was associated with close to a 9-fold higher risk of in-hospital death (aOR 8.90, 95% CI 2.80–28.3), but was not associated with an increased risk of in-hospital colectomy (aOR 0.91, 95% CI 0.53–1.55) (Table 2). CDC also increased the adjusted acute care LOHS by 55% (95% CI 37–76%). Co-morbid illnesses and in-hospital complications of patients who died during index hospitalisation are presented in Supplemental Table S2.
Five-year outcomes in patients discharged from hospital without undergoing colectomy
Among UC patients who were discharged from hospital following partial or complete recovery from their acute illness, patients who had been diagnosed with CDC had higher cumulative 5-year rates of colectomy (46% vs. 36%, P = 0.0054), death (12% vs. 14%, P < 0.0001), non-elective UC-specific hospital readmission (55% vs. 45%, P = 0.012) and non-elective all-cause hospital readmission (81% vs. 69%, P < 0.0001) as compared with patients who had not contracted CDC.
Following covariate adjustment, CDC was associated with higher 5-year risks of mortality (aHR 2.41, 95% CI 1.37–4.22) and all-cause hospital readmission (aHR 1.36, 95% CI 1.10–1.68), but not of colectomy (aHR 1.36, 95% CI 0.98–1.90) or UC-specific hospital readmission (aHR 1.22, 95% CI 0.89–1.66) (Table 3).
Table 3. Adjusted 5-year risks of adverse health events with CDC among UC patients discharged from hospital without undergoing colectomy
In this study of hospitalised UC patients, CDC was an independent risk factor for both 5-year mortality and 5-year rehospitalisation. CDC was also strongly predictive of in-hospital death and longer acute care LOHS, although the former association could not be adjusted for covariates. Conversely, CDC was not independently associated with 5-year or in-hospital risks of colectomy or 5-year risk of UC-specific hospital readmission. Colectomy risk was mainly influenced by disease-specific and hospital-specific factors in this study, which is compatible with other studies demonstrating that UC severity and colonic disease extent are predictive of colectomy risk in UC patients.[16-19]
Collectively, these findings implicate an episode of CDC as a potential risk factor for long-term mortality and rehospitalisation among hospitalised UC patients, possibly as a result of recurrent CDC episodes or through altering long-term UC behaviour. Colonised C. difficile spores and vegetative organisms in some patients could theoretically give rise to abnormal and/or exaggerated immune responses, as has been hypothesised for other host and pathogenic microbes in IBD.[20-26] In addition, the physiologic stress associated with CDC may exacerbate other illnesses, such as cardiovascular disease and renal insufficiency, which could further influence long-term prognosis in these patients.
Clostridium difficile colitis may also give rise to a more aggressive form of acute colitis, which may partly explain the higher in-hospital mortality risk observed in these patients. Use of immunosuppressive therapy, in addition to antibiotics, to treat CDC in the setting of UC may have also increased the duration and/or severity of colitis in many individuals. In addition, some physicians may be inclined to persist longer with medical therapy in C. difficile-infected patients, which could lead to serious complications in patients who are not responding to therapy. It is possible that some patients who died would have benefited from colectomy earlier in the course of their acute illness. CDC could also have exacerbated life-threatening co-morbidities in some patients.
On the other hand, it is possible that unadjusted differences in co-morbidity burden, baseline UC severity or immunosuppressive medication use between infected and uninfected patients influenced the observed associations in this study. Although the Charlson index has been widely validated as a predictor of mortality among hospitalised patients, it may not capture the full extent of co-morbidity burden among hospitalised UC patients.[13, 28-30] Baseline disease severity was adjusted through variables that had reasonable face validity, including prior hospitalisation for UC and UC duration (based on the notion that UC is most aggressive in the first 2 years following diagnosis[31, 32]), as a validated UC disease activity index could not be calculated. Still, admission of some patients with less severe colitis to expedite surgery may have partly selected for patients who were less likely to suffer death, but certain to undergo colectomy among uninfected patients. In addition, no information was available regarding in-hospital medications in this study. Additional studies with more complete adjustment for confounding are thus required to confirm the findings of this study.
Furthermore, C. difficile exposure status may have been misclassified in a proportion of patients, as a result of both misreporting on hospital discharge abstracts as well as inadequate stool testing for C. difficile. No studies have evaluated the diagnostic accuracy of CDC among UC patients within hospital administrative databases. Exposure misclassification may have largely been a random event, which would have attenuated measured associations without invalidating significant ones. However, some patients with fulminant colitis on presentation may have died or undergone colectomy prior to C. difficile testing, which could have further biased the measured associations low. Conversely, C. difficile testing may have been performed more often in patients with severe or treatment-refractory disease, which could have then biased the measured associations high. Future studies with more rigorous case ascertainment methods are thus required to corroborate the findings of this study. Importantly, a high level of coding accuracy has been demonstrated for demographic data and codes for procedures in the Ontario version of the CIHI-DAD.
This is the first study to demonstrate an association between CDC and long-term mortality risk among hospitalised UC patients. Population-based studies have reported a 3.8–5.5-fold higher in-hospital mortality risk and a significantly longer acute care LOHS with CDC in this setting.[3, 6] Although longer LOHS may be a risk factor for C. difficile acquisition, studies have shown that this infection is predominantly acquired prior to hospital admission among IBD patients.[4, 5] With respect to colectomy risk, a US nationwide study reported a decreased risk, a UK nationwide study demonstrated a 1.7-fold higher risk, and two single centre studies reported no difference in risk of in-hospital colectomy with CDC among hospitalised UC patients, although the latter studies may have been underpowered.[7, 8] Notably, one study reported a 2.4-fold higher colectomy risk at 1 year with CDC in this setting. In addition to differences in study methodology, varying strategies for hospital-based care of UC patients in different settings may account for some of these differences.
In summary, this study has demonstrated that CDC is associated with an increased risk of short-term and long-term adverse outcomes, particularly mortality, among hospitalised UC patients. Future prospective studies with rigorous case ascertainment methods and confounder adjustment, which evaluate the impact of CDC among both hospitalised and ambulatory IBD patients, will be important to confirm and expand upon the findings of this study. In particular, aetiologic factors underlying the observed associations in this study require elucidation. Importantly, the results of this study should not be generalised to ambulatory UC patients, who carry a variety of different C. difficile strains as compared with nosocomial patients and who may thus face considerably different risks with this infection. Efforts should nevertheless be made by all health care practitioners to reduce the risk of C. difficile transmission among UC patients.
All authors made significant contributions to the study design, data collection and data analysis and interpretation. S. K. M. drafted the manuscript while all co-authors made critical revisions to the manuscript. S. K. M. assumes full responsibility for data integrity and submission of the final manuscript. G. C. N. and A. H. S. were equally contributing senior authors for this study. Declaration of personal interests: None. Declaration of funding interests: Data acquisition for this study was funded by the The Canadian Institutes of Health Research, The Canadian Association of Gastroenterology and Axcan Pharma (now Aptalis Pharma) through a Fellowship Award with operating allowance. None of these bodies played any role in the study design, data analysis or interpretation, drafting of the manuscript or the decision to submit the article for publication.