In a large multicenter prospective cohort of adults with Clostridium difficile infection (CDI), risk factors for complications (cCDI) included older age, leukocytosis, leucopenia, azotemia, high C-reactive protein, low serum albumin, tachycardia, and tachypnea. Ribotype 027 was not associated with cCDI.
Keywords: Clostridium difficile, complications, mortality, risk factors, ribotype 027
Abstract
Background. Clostridium difficile infection (CDI) is the most common cause of nosocomial infectious diarrhea and may result in severe complications including death. We conducted a prospective study to identify risk factors for complications of CDI (cCDI).
Methods. Adult inpatients with confirmed CDI in 10 Canadian hospitals were enrolled and followed for 90 days. Potential risk factors were measured within 24 hours of diagnosis. Isolates were typed by polymerase chain reaction ribotyping. cCDI was defined as 1 or more of the following: colonic perforation, toxic megacolon, colectomy, admission to an intensive care unit for cCDI, or if CDI contributed to death within 30 days of enrollment. Risk factors for cCDI were investigated by logistic regression.
Results. A total of 1380 patients were enrolled. cCDI was observed in 8% of patients. The ribotype was identified in 922 patients, of whom 52% were infected with R027. Age ≥80 years, heart rate >90/minute, respiratory rate >20/minute, white cell count <4 × 109/L or ≥20 × 109/L, albumin <25 g/L, blood urea nitrogen >7 mmol/L, and C-reactive protein ≥150 mg/L were independently associated with cCDI. A higher frequency of cCDI was observed among R027-infected patients (10.9% vs 7.2%), but the association was not significant in adjusted analysis.
Conclusions. CDI complications were associated with older age, abnormal blood tests, and abnormal vital signs. These factors, which are readily available to clinicians at the time of diagnosis, could be used for outcome prediction and risk stratification to select patients who may need closer monitoring or more aggressive therapy.
(See the Editorial Commentary by Aronoff on pages 1789–91.)
Clostridium difficile infection (CDI) is the leading cause of healthcare-associated infectious diarrhea, represents 15%–25% of nosocomial diarrhea caused by antibiotics, and is associated with a high economic burden [1, 2]. In North America, the most dominant strain type isolated since 2002, NAP1/BI/027, has received much attention as it is associated with high incidence rates and increased risk of unfavorable outcomes [3]. However, other strains have also been associated with poor outcomes and increasing incidence, making CDI a persistent challenge in healthcare facilities worldwide [4, 5].
In large cohort studies, patients infected with C. difficile had a median risk of developing complications (toxic megacolon, colectomy, shock, perforation) of 11% and a median risk of all-cause 30-day mortality of 13% [4, 6–13]. Mortality and complications (cCDI) have been associated with older age, underlying medical comorbidities, renal failure, and high white blood cell count (WBC), while hypoalbuminemia and ribotype 027 (R027) have been correlated with mortality alone [14]. Most of these studies were limited by the use of small retrospective cohorts, high heterogeneity regarding their outcomes’ and variables’ definitions, as well as lack of multivariable analysis. Few studies have measured the association between strain type and unfavorable outcomes.
Oral vancomycin is the recommended treatment for severe CDI, and intravenous metronidazole with high-dose vancomycin is recommended for cCDI [15]. The ability to identify patients at high risk of cCDI early in the course of illness could improve clinical decision-making. Such patients might benefit from closer monitoring, more aggressive rehydration, selection of vancomycin as a first-line agent, adjunctive treatments, or earlier evaluation by surgeons. In this study, we aimed to identify independent risk factors for cCDI in a multicenter prospective cohort study of hospitalized adults with CDI, with an emphasis on strain type.
METHODS
Patients with confirmed CDI, hospitalized in 1 of 10 acute care hospitals in the provinces of Quebec and Ontario, Canada, were enrolled prospectively between June 2005 and October 2008. The participating hospitals are described in Supplementary Methods. The research ethics boards of all participating institutions approved the study.
Inclusion and Exclusion Criteria
Patients with CDI were eligible if they were aged ≥18 years, hospitalized at the time of diagnosis (or, if seen in the emergency or outpatient clinic, hospitalization was planned on the same day), and if the patient or proxy provided written consent to participate. Patients already experiencing 1 of the predefined outcomes at the time of enrollment (colonic perforation, toxic megacolon, septic shock, or indication for emergency colectomy) or receiving palliative care were excluded.
CDI Definition
CDI was defined as having at least 6 unformed stools over 36 hours or having a diagnosis of paralytic ileus, and either positive C. difficile toxin detection in a stool sample or pseudomembranous colitis demonstrated by endoscopy. Each participating center performed toxin detection according to local laboratory protocols (see Supplementary Methods). Rapid tests were used to minimize time between collection and results.
Clinical Predictors
Enrollment was performed as soon as possible after a positive C. difficile toxin test or endoscopy. At the time of enrollment, we recorded demographics, data on hospital admission, chronic comorbidities, immune status, and surgical procedures or gastrointestinal instrumentation. Information was collected from medical charts and patient interviews about antimicrobial therapy, gastric acid suppression (H2 receptor antagonists, proton pump inhibitors, and antacids), and antiperistaltic agents used within 2 months prior to diagnosis of CDI. Immunosuppression was defined as 1 or more of the following: leukemia, lymphoma, chemotherapy and/or radiation therapy, high-potency glucocorticoids (defined as any intravenous steroids, prednisone ≥20 mg or equivalent, or dexamethasone at any dosage for at least 2 weeks) within 6 months prior to CDI diagnosis, organ transplantation, other immunosuppressive drugs, or human immunodeficiency virus infection. We also collected the following clinical data on characteristics of CDI at enrollment: previous CDI episode, site of acquisition, clinical presentation (including vital signs, abdominal pain, and confusion), and treatment. All laboratory tests were performed at each participating center, except for C-reactive protein (CRP) measurements, which were carried out centrally at the Centre Hospitalier Universitaire de Sherbrooke (see Supplementary Methods). For vital signs and laboratory tests, the most abnormal value within 12 hours before or 24 hours post-enrollment was abstracted.
Clostridium Difficile Culture and Polymerase Chain Reaction Ribotyping
Details of procedures are described in Supplementary Methods. Clostridium difficile colonies from stool specimens were grown, and genomic DNA was extracted from a single colony. Amplified products of endpoint polymerase chain reactions (PCRs) were analyzed by automated chip-based microcapillary electrophoresis. Ribotype profiles were determined and analyzed with the GelCompar II software, version 5.1 (Applied Maths NV, Sint-Martens-Latem, Belgium). New ribotype groups were assigned when a group of strains did not match any of the reference strains in our library (electrophoresis profile with Pearson correlation <85% with reference profiles).
Follow-up and Outcomes
Follow-up evaluation was performed until day 90 after enrollment by reviewing medical charts and by directly contacting patients and/or attending physicians. Assessment of outcomes was determined by research assistants who were blinded to the status of the predictor variables. Based on published criteria, cCDI was defined when a patient met any of the following: admission to an intensive care unit (ICU) for complications associated with CDI; colonic perforation, toxic megacolon (defined by a transverse/ascending colon dilation of ≥6 cm on plain X ray and clinical criteria) [16], colectomy, or hemicolectomy; or CDI was the cause or contributed to death within 30 days after enrollment [17]. In a secondary analysis, a less specific definition for complicated CDI (cCDI-2) was used, which included any reason for admission to an ICU and 30-day all-cause mortality. This second analysis was performed to make our study comparable to several published studies and is provided in Supplementary Results. A patient could experience 1 or more of the listed outcomes.
Statistical Analyses
Statistical analyses were performed using IBM SPSS Statistics for Windows, version 22.0 (IBM Corp., Armonk, New York) and SAS 9.4 (SAS Institute Inc., Cary, North Carolina). Frequency of missing data was assessed during data screening. Few variables (mainly laboratory tests) had more than 10% of data missing. Imputation procedures were not considered necessary; missing data were considered as an additional category when relevant. Proportions were compared with 2-tailed χ2 or Fisher exact tests when appropriate. Logistic regression and Wald tests were used to identify predictors of cCDI and cCDI-2. Multivariable models were built up by selecting variables significant at P ≤ .1 in univariate analyses manually and by adding them one at a time in order of the lowest P value. Interactions were tested for relevant variables. Models were compared using the likelihood ratio test, and variables significant at P < .05 were kept in the final models.
RESULTS
A total of 1380 patients with CDI were enrolled in the 10 participating acute care centers; 65% (n = 894) in Ontario and 35% (n = 486) in Quebec. Patient characteristics are shown in Table 1. Overall, 88% (n = 1207) had at least 1 chronic underlying illness and 63% (n = 866) had 2 or more. The most frequent comorbidities were chronic heart disease (n = 565, 41%), diabetes (n = 364, 27%), chronic lung disease (n = 356, 26%), and cancer (n = 354, 26%). Within the 2 months preceding enrollment, 38% of patients had undergone surgery, 87% had received at least 1 antimicrobial agent, 66% had received at least 1 acid suppression medication, and and 5% had received at least 1 antidiarrheal. One-third of enrolled patients (n = 401) were immunocompromised, with the most frequent cause being receipt of glucocorticoid therapy (n = 221, 16%).
Table 1.
Characteristics of 1380 Patients With Clostridium difficile Infection
| Variable | Value or No. (%) |
|---|---|
| Age | Median 71 y (interquartile range, 58–80) |
| Sex | |
| Female | 665 (48.2) |
| Male | 715 (51.8) |
| Charlson comorbidity index | |
| 0–3 | 747 (54.1) |
| 4–6 | 399 (28.9) |
| ≥7 | 234 (17.0) |
| Antimicrobial exposurea | 1201 (87.0) |
| Fluoroquinolones | 707 (51.0) |
| Cephalosporins | 694 (50.0) |
| Carboxy/ureidopenicillins | 274 (20.0) |
| Macrolides/clindamycin | 251 (18.0) |
| Antistaphylococcal/aminopenicillins | 225 (16.3) |
| Acid suppression agents | 917 (66.4) |
| PPI | 577 (42.0) |
| H2-RA | 191 (13.8) |
| PPI + H2-RA | 148 (10.7) |
| CDI diagnosis method | |
| Conventional toxins A+B EIA | 1001 (72.5) |
| GDH + toxin A detection | 239 (17.3) |
| Cytotoxicity assay | 83 (6.0) |
| Rapid toxins A+B EIA | 52 (3.8) |
| Endoscopy | 5 (0.4) |
| Origin of CDI | |
| Hospital onset-HCFA | 1126 (81.6) |
| Community onset-HCFA | 111 (8.0) |
| Community acquired | 143 (10.4) |
| CDI treatment | |
| Metronidazole (PO or IV) | 1119 (81.1) |
| Vancomycin | 110 (8.0) |
| Metronidazole and vancomycin | 90 (6.5) |
| None | 61 (4.4) |
| CDI outcomes (n = 1367) | |
| 30-day all-cause mortality | 169 (12.2) |
| CDI-associated 30-day mortality | 54 (4.0) |
| cCDIb | 108 (7.9) |
| cCDI-2c | 212 (15.5) |
| At least 1 recurrenced | 322 (23.6) |
Abbreviations: cCDI, complications of CDI; CDI, Clostridium difficile infection; EIA, enzyme immunoassay; GDH, glutamate dehydrogenase; H2-RA, histamine type-2 receptor antagonists; HCFA, healthcare facility–associated; IV, intravenous; PO, per os; PPI, proton pump inhibitor.
a Each patient could have received more than 1 class of antimicrobials within 2 months of enrollment.
b cCDI is defined as 1 or more of the following: admission to an intensive care unit (ICU) for complications associated with CDI, colonic perforation, toxic megacolon, colectomy or hemicolectomy, or CDI being the cause or contributed to death within 30 days after enrollment.
c cCDI-2 is defined as 1 or more of the following: admission to an ICU for any reason, colonic perforation, toxic megacolon, colectomy or hemicolectomy, or 30-day all-cause mortality.
d Recurrence was defined by the presence of diarrhea and C. difficile toxin or compatible endoscopy, or prescription of an empiric CDI treatment, at least 48 hours after the completion of the last CDI treatment.
Almost all CDI cases were confirmed with toxin detection (>99%) and a very small number by endoscopy alone (n = 5). Conventional toxin A and B enzyme immunoassay (EIA) was used in all Ontario patients. In Quebec patients, EIA rapid cassette assays were most frequently used (60%) followed by conventional EIA (22%) and/or a combination of tests including cytotoxicity assay (6%). Most patients (95%) were enrolled within 24 hours of a positive toxin test, and the median time between the first symptoms and enrollment was 3 days (interquartile range [IQR], 2–5 days). The majority of CDI cases were classified as healthcare facility–associated (n = 1237, 90%) and corresponded to a first episode in 86% of cases (n = 1180). The median duration of hospitalization before CDI onset was 9 days (IQR, 3–21 days). Metronidazole was the most frequent initial treatment (n = 1119, 81%); however, the treatment was changed to oral vancomycin in 16% (n = 177) of patients, mainly due to clinical deterioration.
Strain ribotyping results are shown in Table 2. R027 was the most common strain, particularly in Quebec centers (72% vs 43% in Ontario; P < .001), and during years 2007 and 2008 (54% and 65%, respectively; P < .001). New ribotypes accounted for 25% of strains, followed by R01 and R014 (9% and 8%, respectively).
Table 2.
Ribotyping and Distribution of Frequent Ribotypes According to the Provinces and Year of Enrollment
| Ribotypes | Province |
Year of Enrollment |
|||||
|---|---|---|---|---|---|---|---|
| Cohort (n = 1380) | Ontario (n = 894) | Quebec (n = 486) | 2005 (n = 121) | 2006 (n = 410) | 2007 (n = 496) | 2008 (n = 353) | |
| Stool sample | 1054 (76.4) | 696 (77.8) | 368 (75.7) | 98 (81.0) | 331 (80.7) | 355 (71.6) | 270 (76.5) |
| Positive culture | 948 (89.9) | 623 (89.5) | 325 (88.3) | 92 (76.0) | 289 (70.5) | 326 (65.7) | 241 (68.3) |
| Ribotype obtained | 922 (66.8) | 614 (68.7) | 308 (63.4) | 90 (74.4) | 285 (69.5) | 314 (63.3) | 233 (66.0) |
| R027 | 483 (52.4) | 262 (42.7) | 221 (71.7) | 35 (28.9) | 126 (30.7) | 170 (34.3) | 152 (43.1) |
| R001 | 86 (9.3) | 82 (13.3) | 4 (1.3) | 9 (7.4) | 40 (9.8) | 30 (6.0) | 7 (2.0) |
| R014 | 75 (8.1) | 59 (9.6) | 16 (5.2) | 8 (6.6) | 27 (6.6) | 22 (4.4) | 18 (5.1) |
| Other ribotypesa | 48 (5.2) | 41 (6.7) | 9 (2.9) | 10 (8.3) | 18 (4.4) | 14 (2.8) | 6 (1.7) |
| New ribotypesb | 230 (24.9) | 172 (28.0) | 58 (18.8) | 28 (23.1) | 74 (18.0) | 78 (15.7) | 50 (14.2) |
Data is provided in no. (%).
a Other ribotypes were R002, R015, R037, and R078.
b New ribotypes corresponded to strains whose electrophoresis profile showed a Pearson correlation <85% with reference profiles.
Outcomes
Follow-up was completed for 1367 patients. Toxic megacolon occurred in 15 patients and an intestinal perforation in 2. Hemi- or total colectomy was performed in 16 patients. In 42 patients, ICU admission was needed for CDI management. Overall, 12% of patients died within 30 days (n = 169; median time to death, 10 days; IQR, 5–18 days), while 22% died within 90 days (n = 296). At 30 days, CDI was identified as the direct cause of death in 1% of cases (n = 15) and was associated with death in 3% (n = 39). Consequently, cCDI was observed in 8% of patients (n = 108). When considering all-cause ICU admission and mortality, cCDI-2 was observed in 16% (n = 212) of patients, of which 81% died within 30 days and 21% were admitted to an ICU.
Risk Factors for cCDI
Factors associated with cCDI in univariable analyses are shown in Table 3. The odds of developing cCDI were higher in patients with WBC ≥20 × 109/L, serum albumin <25 g/L, CRP ≥150 mg/L, blood urea nitrogen (BUN) ≥7 mmol/L, or serum creatinine ≥200 µmol/L. Age greater than 80 years; chronic heart, lung, or kidney disease; dementia; recent elective surgery; tachycardia; leukopenia (WBC <4 × 109/L); tachypnea; fever (>38°C); or hypothermia (<36°C) were also significantly associated with cCDI in univariable analysis but to a lesser extent. No statistically significant association was found between cCDI and gender, site of CDI acquisition, province, diabetes mellitus, cancer, immunosuppression, antimicrobial use, proton pump inhibitors, anti-peristaltic agents, confusion, and platelet count. In bivariable analysis, BUN was associated with cCDI (7–10 mmol/L: odds ratio [OR], 2.8; 95% confidence interval [CI], 1.4–5.6; ≥11 mmol/L: OR, 6.1; 95% CI, 3.2–11.8) but not serum creatinine (100–199 µmol/L: OR, 1.1; 95% CI, .6–1.9; ≥200 µmol/L: OR, 1.4; 95% CI, .7–2.6).
Table 3.
Variables Associated With Complicated Clostridium difficile Infection in Univariable Analysis
| Variable | No. Complicated Clostridium difficile Infections/Total (%) | Odds Ratio (95% Confidence Interval) |
P Value |
|---|---|---|---|
| Age (y) | |||
| 18–64 | 27/511 (5.3) | … | |
| 65–79 | 39/504 (7.7) | 1.50 (.94–2.50) | .12 |
| ≥80 | 42/342 (12.3) | 2.51 (1.52–4.16) | <.001 |
| Comorbidities | |||
| Dementia | |||
| No | 95/1264 (7.5) | … | |
| Yes | 13/93 (14.0) | 2.00 (1.07–3.73) | .03 |
| Heart disease | |||
| No | 46/803 (5.7) | … | |
| Yes | 62/554 (11.2) | 2.07 (1.39–3.09) | <.001 |
| Lung disease | |||
| No | 71/1011 (7.0) | … | |
| Yes | 37/346 (10.7) | 1.58 (1.04–2.41) | .03 |
| Kidney disease | |||
| No | 73/1061 (6.9) | … | |
| Yes | 35/296 (11.8) | 1.82 (1.19–2.78) | .01 |
| Surgery (≤2 mo) | |||
| None | 75/835 (9.0) | … | |
| Emergency | 14/169 (8.3) | 0.92 (.50–1.66) | .77 |
| Elective | 19/353 (5.4) | 0.58 (.34–.97) | .04 |
| Strain ribotype | |||
| Other | 31/430 (7.2) | … | |
| R027 | 52/475 (10.9) | 1.58 (.99–2.52) | .05 |
| Unavailable | 25/452 (5.5) | 0.75 (.44–1.29) | .31 |
| Heart rate | |||
| ≤90/min | 43/866 (5.0) | … | |
| >90/min | 65/484 (13.4) | 2.97 (1.98–4.44) | <.001 |
| Respiratory rate | |||
| ≤20/min | 43/866 (5.0) | … | |
| >20/min | 65/484 (13.4) | 2.97 (1.98–4.44) | <.001 |
| Fever (°C) | |||
| 36–38 | 86/1188 (7.2) | … | |
| <36 or >38 | 19/155 (12.3) | 1.79 (1.06–3.04) | .03 |
| White blood cell count (109/L) | |||
| <4 | 14/121 (11.6) | 2.57 (1.34–4.94) | .01 |
| 4–11.9 | 35/723 (4.8) | … | |
| 12–19.9 | 30/345 (8.7) | 1.87 (1.13–3.10) | .02 |
| ≥20 | 28/144 (19.4) | 4.75 (2.78–8.10) | <.001 |
| Serum albumin (g/L) | |||
| <25 | 52/364 (14.3) | 8.58 (3.06–24.10) | <.001 |
| 26–34.9 | 37/600 (6.2) | 3.38 (1.19–9.61) | .02 |
| ≥35 | 4/210 (1.9) | … | |
| Missing | 15/183 (8.2) | 4.60 (1.50–14.11) | .01 |
| C-reactive protein (mg/L) | |||
| <50 | 20/487 (4.1) | … | |
| 50–149.9 | 27/413 (6.5) | 1.63 (.90–2.96) | .11 |
| ≥150 | 32/122 (26.2) | 8.30 (4.55–15.17) | <.001 |
| Creatinine (µmol/L) | |||
| 0–99 | 44/868 (5.1) | … | |
| 100–199 | 35/301 (11.6) | 2.46 (1.55–3.92) | <.001 |
| ≥200 | 25/133 (18.8) | 4.33 (2.55–7.37) | <.001 |
| Dialysis | 3/34 (8.8) | 1.81 (.53–6.16) | .34 |
| Blood urea nitrogen (mmol/L) | |||
| <7 | 22/716 (3.1) | … | |
| 7–10.9 | 18/207 (8.7) | 3.00 (1.58–5.72) | .01 |
| ≥11 | 54/292 (18.5) | 7.16 (4.27–12.00) | <.001 |
| Dialysis | 3/35 (8.6) | 2.96 (.84–10.40) | .09 |
| Missing | 11/107 (10.3) | 3.62 (1.70–7.69) | .01 |
Univariable analysis was performed on 1357 patients.
The multivariable logistic regression model identified the following 7 independent correlates of cCDI (Table 4): age ≥80 years (OR, 2.2; 95% CI, 1.2–4.0), increased heart rate (>90/min; OR, 2.1; 95% CI, 1.4–3.3) and tachypnea (respiratory rate >20/min; OR, 1.7; 95% CI, 1.1–2.8), abnormal WBC (<4 × 109/L [OR, 2.6; 95% CI, 1.3–5.5] and/or ≥20 × 109/L [OR, 2.2; 95% CI, 1.2–4.0]), serum albumin <25 g/L (OR, 3.1; 95% CI, 1.0–9.3), BUN ≥11 mmol/L (OR, 4.9; 95% CI, 2.8–8.5), and CRP ≥150 mg/L (OR, 3.6; 95% CI, 1.8–7.2).
Table 4.
Independent Risk Factors for Complicated Clostridium difficile Infection on Multivariable Logistic Regression
| Variable | Adjusted Odds Ratio (95% Confidence Interval) |
P Value |
|---|---|---|
| Age (y) | ||
| 18–64 | … | |
| 65–79 | 1.13 (.64–1.99) | .67 |
| ≥80 | 2.20 (1.22–3.96) | .009 |
| Heart rate | ||
| ≤90/min | … | |
| >90/min | 2.13 (1.35–3.34) | .001 |
| Respiratory rate | ||
| ≤20/min | … | |
| >20/min | 1.75 (1.08–2.84) | .023 |
| White blood cell count (109/L) | ||
| <4 | 2.61 (1.25–5.45) | .011 |
| 4–11.9 | … | |
| 12–19.9 | 1.42 (.82–2.44) | .21 |
| ≥20 | 2.20 (1.20–4.03) | .011 |
| Serum albumin (g/L) | ||
| <25 | 3.11 (1.04–9.31) | .043 |
| 26–34.9 | 2.05 (.69–6.07) | .19 |
| ≥35 | … | |
| Missing | 2.33 (.69–7.78) | .17 |
| C-reactive protein (mg/L) | ||
| <50 | … | |
| 50–149.9 | 1.18 (.63–2.23) | .61 |
| ≥150 | 3.61 (1.81–7.20) | <.001 |
| Missing | 1.44 (.75–2.77) | .28 |
| Blood urea nitrogen (mmol/L) | ||
| <7 | … | |
| 7–10.9 | 2.61 (1.32–5.17) | .006 |
| ≥11 | 4.88 (2.81–8.48) | <.001 |
| Dialysis | 4.03 (1.02–15.90) | .046 |
| Missing | 3.70 (1.59–8.60) | .002 |
Multivariable analysis was performed on 1333 patients.
Although a higher frequency of cCDI was observed among R027 strains (10.9% vs 7.2%; P = .008), the association with R027 did not reach statistical significance after multivariable adjustment (OR, 1.6; 95% CI, .96–2.7). None of the other ribotypes were associated with cCDI.
Risk Factors for cCDI-2
In univariable analyses, in addition to all the risk factors associated with cCDI, we also found significant associations between chemotherapy, fluoroquinolones use in the prior 2 months, confusion, and cCDI-2 (Supplementary Table 1). No association between a specific ribotype and cCDI-2 was identified. The final multivariable logistic regression model contained 4 additional independent correlates of cCDI-2, for a total of 11. These variables were comorbidities (dementia and chronic heart disease), recent chemotherapy, and recent elective surgery, the latter as a protective factor (Supplementary Table 2).
DISCUSSION
In this large multicenter prospective cohort with CDI followed for 90 days after enrollment, 8% developed a cCDI. This is consistent with other studies using a similar definition for C. difficile complication [6, 9, 10]. The population was mainly hospitalized elderly patients with numerous comorbidities, frequent exposure to antimicrobials, acid suppression, and infected predominantly with the R027 strain. Independent predictors of cCDI were older age, increased heart and respiratory rates, leukocytosis, leukopenia, azotemia, high CRP, and hypoalbuminemia. In several studies, older age, increased WBC, and increased serum creatinine and/or BUN have been repeatedly reported as independent predictors of cCDI; other risk factors were only reported occasionally [6, 14, 18]. Systemic inflammatory response criteria (tachycardia, tachypnea, abnormal WBC) and high CRP likely reflect the severity of colonic inflammation. Elevated BUN may indicate severe diarrhea with subsequent dehydration resulting in inadequate renal perfusion and prerenal azotemia. Urea is the primary metabolite derived from dietary proteins and tissue protein turnover; hence, unlike creatinine, it is also affected by catabolic factors such as fever and sepsis [19]. While guidelines and previous investigations have considered elevation in serum creatinine (≥1.5 × baseline value) as a marker of severity [15], BUN appeared as a strong confounding factor for creatinine in our study, such that the association of creatinine with both outcomes was no longer significant after adjustment for BUN. Finally, hypoalbuminemia results from protein losses through inflamed colonic mucosa and from increased catabolism due to sepsis [20].
In addition to the recommended definition of cCDI [17], we tested a less specific but frequently used definition (cCDI-2, Supplementary Results). In addition to the other predictors found with the first model, cCDI-2 was independently associated with dementia, chronic heart disease, recent chemotherapy, and elective surgery (the latter being protective). The first 3 factors identify a subgroup of patients with chronic and severe preexisting medical conditions with a higher likelihood of mortality for which CDI may have played a minor role or none at all. On the other hand, a recent elective surgery may reflect better underlying health status. The use of a less specific definition of cCDI by previous researchers may have led to the identification of risk factors related to mortality in general, and thus may be less useful for patient management.
R027 accounted for half of the strains identified in our patients. We found a trend toward an association between R027 and cCDI, nearly reaching the level of significance (P = .053). When cCDI-2 was used as the outcome, no association was demonstrated. Thus, the selection of a less specific outcome definition might, in part, explain why several studies were unable to demonstrate a significant association between the epidemic strain and the risk of an unfavorable outcome [21–23]. Furthermore, despite the advanced PCR ribotyping that was used [24, 25], more discriminatory techniques are suggested to clarify the relation of R027 subtypes with the outcomes [26–28]. Genetic changes have been observed in the pathogenicity locus of C. difficile, but observational studies are inconsistent concerning their impact on CDI outcomes [21, 23, 29–31]. These studies varied substantially in CDI incidence rates across centers, definitions of outcomes, and typing techniques [14]. Associations with cCDI have been found for R018 and R056 [4, 32], but the limited number of these strains among our patients precluded us to verify these findings.
This study had some limitations. Stool samples were unavailable for subsequent testing in almost 25% of included patients and 10% of the specimens were negative in culture. Consequently, we managed to type bacterial strains in only 67% of patients. Some patients might have had false-positive tests for toxin, since the reported specificities of EIAs range from 95% to 99% and the testing was performed before commercial PCR assays became available [33]. Missing values for clinically important variables were inevitable despite the prospective enrollment, particularly when additional blood samples needed to be drawn. For some categorical variables with missing data, we performed dummy variable adjustment instead of dropping them to avoid reducing the sample size and generating differentially distorted associations [34] and to reach an acceptable event per variable ratio [35]. Establishing causal links between death, admission to an ICU, and CDI is challenging [36–38], particularly in patients with multiple comorbidities. However, since this assessment was made prospectively, we believe it was reliable. Finally, patients in this cohort were enrolled from 2005 to 2008; the currently circulating strains might differ from the ones collected during that period.
The study also had several strengths. This multicenter prospective cohort was assembled specifically to identify predictors of cCDI at the time of diagnosis based on methodologically accepted criteria [39]. Diagnosis was based on toxin identification plus clinical symptoms compatible with CDI. We preselected variables that are routinely available at the time of diagnosis, and these variables were measured as close as possible to the time of diagnosis. Outcomes were evaluated by assistants blinded to the data collected at enrollment.
CONCLUSIONS
Through a large multicenter prospective cohort, we identified age ≥80 years, increased respiratory (>20/min) and heart rate (>90/min), WBC <4 and ≥20 × 109 /L, elevated CRP (≥150 mg/L), hypoalbuminemia (serum albumin <25 g/L), and elevated BUN (≥7 mmol/L) as independent risk factors for CDI complications. These predictors, which were readily available at the time of diagnosis, could serve to develop and validate a score that can be used to identify patients who could benefit from more aggressive treatment and closer monitoring. Ultimately, the usefulness of prediction scores will need to be evaluated by examining the fate of patients treated with either metronidazole or vancomycin whose “at-risk” status differs whether a given score or the current Infectious Diseases Society of America criteria are used.
Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.
Notes
Acknowledgments. We thank all the project research assistants for their efforts.
Author contributions. L. V., A. M., A.-C. L., A. E. S., W. L. G., M. P. M., S. M. C., and J. P. conceived and designed the study. L. V., A. M., A.-C. L., A. E. S., W. L. G., M. P. M., K. K., and J. P. obtained data. C. N. A. C., J. P., L. V., J. R. G., and L.-C. F. performed the analyses and interpretation. C. N. A. C. drafted the article. All authors reviewed and approved the final version.
Financial support. This work was supported by the Canadian Institutes of Health Research (MOP-74514). C. N. A. C. is a recipient of a scholarship for doctoral training from the Fonds de recherche du Québec-Santé. L. V. is a member of the Fonds de recherche du Québec-Santé–funded Centre de Recherche du Centre Hospitalier Universitaire de Sherbrooke.
Potential conflicts of interests. A. M. was a member of an advisory board for Optimer Pharmaceuticals (now Merck & Co., Inc.). She has received grants for phase 3 clinical trial from Merck & Co. Inc. and from GlaxoSmithKline Inc. for an observational study. A. E. S. received an honorarium for consultancy on advisory boards for Merck Canada Inc. and Cubist Pharmaceuticals Inc. L. V. was a consultant to Pfizer and Optimer Pharmaceuticals; has received research grants from Pfizer, Optimer Pharmaceuticals, Merck, and Sanofi Pasteur; and has received honorary for lectures from Optimer Pharmaceuticals. All other authors report no potential conflicts.
All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.
References
- 1.Kwon JH, Olsen MA, Dubberke ER. The morbidity, mortality, and costs associated with Clostridium difficile infection. Infect Dis Clin North Am 2015; 29:123–34. [DOI] [PubMed] [Google Scholar]
- 2.Nanwa N, Kendzerska T, Krahn M et al. . The economic impact of Clostridium difficile infection: a systematic review. Am J Gastroenterol 2015; 110:511–9. [DOI] [PubMed] [Google Scholar]
- 3.Gerding DN, Lessa FC. The epidemiology of Clostridium difficile infection inside and outside health care institutions. Infect Dis Clin North Am 2015; 29:37–50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Bauer MP, Notermans DW, Van Benthem BH et al. . Clostridium difficile infection in Europe: a hospital-based survey. Lancet 2011; 377:63–73. [DOI] [PubMed] [Google Scholar]
- 5.Lessa FC, Gould CV, Mcdonald LC. Current status of Clostridium difficile infection epidemiology. Clin Infect Dis 2012; 55(suppl 2):S65–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Henrich TJ, Krakower D, Bitton A, Yokoe DS. Clinical risk factors for severe Clostridium difficile-associated disease. Emerg Infect Dis 2009; 15:415–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Pepin J, Valiquette L, Alary ME et al. . Clostridium difficile-associated diarrhea in a region of Quebec from 1991 to 2003: a changing pattern of disease severity. CMAJ 2004; 171:466–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Cloud J, Noddin L, Pressman A, Hu M, Kelly C. Clostridium difficile strain NAP-1 is not associated with severe disease in a nonepidemic setting. Clin Gastroenterol Hepatol 2009; 7:868–73. [DOI] [PubMed] [Google Scholar]
- 9.Khanna S, Aronson SL, Kammer PP, Baddour LM, Pardi DS. Gastric acid suppression and outcomes in Clostridium difficile infection: a population-based study. Mayo Clin Proc 2012; 87:636–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Morrison RH, Hall NS, Said M et al. . Risk factors associated with complications and mortality in patients with Clostridium difficile infection. Clin Infect Dis 2011; 53:1173–8. [DOI] [PubMed] [Google Scholar]
- 11.Rao K, Micic D, Natarajan M et al. . Clostridium difficile ribotype 027: relationship to age, detectability of toxins A or B in stool with rapid testing, severe infection, and mortality. Clin Infect Dis 2015; 61:233–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Stewart DB, Hollenbeak CS. Clostridium difficile colitis: factors associated with outcome and assessment of mortality at a national level. J Gastrointest Surg 2011; 15:1548–55. [DOI] [PubMed] [Google Scholar]
- 13.Gasperino J, Garala M, Cohen HW, Kvetan V, Currie B. Investigation of critical care unit utilization and mortality in patients infected with Clostridium difficile. J Crit Care 2010; 25:282–6. [DOI] [PubMed] [Google Scholar]
- 14.Abou Chakra CN, Pepin J, Sirard S, Valiquette L. Risk factors for recurrence, complications and mortality in Clostridium difficile infection: a systematic review. PLoS One 2014; 9:e98400. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Cohen SH, Gerding DN, Johnson S et al. . Clinical practice guidelines for Clostridium difficile infection in adults: 2010 update by the Society for Healthcare Epidemiology of America (SHEA) and the Infectious Diseases Society of America (IDSA). Infect Control Hosp Epidemiol 2010; 31:431–55. [DOI] [PubMed] [Google Scholar]
- 16.Gan SI, Beck PL. A new look at toxic megacolon: an update and review of incidence, etiology, pathogenesis, and management. Am J Gastroenterol 2003; 98:2363–71. [DOI] [PubMed] [Google Scholar]
- 17.Mcdonald LC, Coignard B, Dubberke E, Song X, Horan T, Kutty PK. Recommendations for surveillance of Clostridium difficile-associated disease. Infect Control Hosp Epidemiol 2007; 28:140–5. [DOI] [PubMed] [Google Scholar]
- 18.Byrn JC, Maun DC, Gingold DS, Baril DT, Ozao JJ, Divino CM. Predictors of mortality after colectomy for fulminant Clostridium difficile colitis. Arch Surg 2008; 143:150–4. [DOI] [PubMed] [Google Scholar]
- 19.Hosten AO. BUN and creatinine. In: Walker HK, Hall WD, Hurst JW, eds. Clinical Methods: The History, Physical, and Laboratory Examinations. Boston: Butterworths, 1990. [PubMed] [Google Scholar]
- 20.Gatta A, Verardo A, Bolognesi M. Hypoalbuminemia. Intern Emerg Med 2012; 7(suppl 3):S193–9. [DOI] [PubMed] [Google Scholar]
- 21.Walk ST, Micic D, Jain R et al. . Clostridium difficile ribotype does not predict severe infection. Clin Infect Dis 2012; 55:1661–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Rao K, Walk ST, Micic D et al. . Procalcitonin levels associate with severity of Clostridium difficile infection. PLoS One 2013; 8:e58265. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.See I, Mu Y, Cohen J et al. . NAP1 strain type predicts outcomes from Clostridium difficile infection. Clin Infect Dis 2014; 58:1394–400. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Schumann P, Pukall R. The discriminatory power of ribotyping as automatable technique for differentiation of bacteria. Syst Appl Microbiol 2013; 36:369–75. [DOI] [PubMed] [Google Scholar]
- 25.Kurka H, Ehrenreich A, Ludwig W et al. . Sequence similarity of Clostridium difficile strains by analysis of conserved genes and genome content is reflected by their ribotype affiliation. PLoS One 2014; 9:e86535. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Huber CA, Foster NF, Riley TV, Paterson DL. Challenges for standardization of Clostridium difficile typing methods. J Clin Microbiol 2013; 51:2810–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Knetsch CW, Lawley TD, Hensgens MP, Corver J, Wilcox MW, Kuijper EJ. Current application and future perspectives of molecular typing methods to study Clostridium difficile infections. Euro Surveill 2013; 18:20381. [DOI] [PubMed] [Google Scholar]
- 28.Eyre DW, Cule ML, Wilson DJ et al. . Diverse sources of C. difficile infection identified on whole-genome sequencing. N Engl J Med 2013; 369:1195–205. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Martin H, Willey B, Low DE et al. . Characterization of Clostridium difficile strains isolated from patients in Ontario, Canada, from 2004 to 2006. J Clin Microbiol 2008; 46:2999–3004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Hubert B, Loo VG, Bourgault AM et al. . A portrait of the geographic dissemination of the Clostridium difficile North American pulsed-field type 1 strain and the epidemiology of C. difficile-associated disease in Quebec. Clin Infect Dis 2007; 44:238–44. [DOI] [PubMed] [Google Scholar]
- 31.Sirard S, Valiquette L, Fortier LC. Lack of association between clinical outcome of Clostridium difficile infections, strain type, and virulence-associated phenotypes. J Clin Microbiol 2011; 49:4040–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Soes LM, Brock I, Persson S, Simonsen J, Pribil Olsen KE, Kemp M. Clinical features of Clostridium difficile infection and molecular characterization of the isolated strains in a cohort of Danish hospitalized patients. Eur J Clin Microbiol Infect Dis 2012; 31:185–92. [DOI] [PubMed] [Google Scholar]
- 33.Novak-Weekley SM, Marlowe EM, Miller JM et al. . Clostridium difficile testing in the clinical laboratory by use of multiple testing algorithms. J Clin Microbiol 2010; 48:889–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Little RJA, Rubin DB. Statistical analysis with missing data. 2nd ed Hoboken: Wiley Interscience, 2002. [Google Scholar]
- 35.Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol 1996; 49:1373–9. [DOI] [PubMed] [Google Scholar]
- 36.Gilca R, Frenette C, Theriault N, Fortin E, Villeneuve J. Attributing cause of death for patients with Clostridium difficile infection. Emerg Infect Dis 2012; 18:1707–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Hota SS, Achonu C, Crowcroft NS, Harvey BJ, Lauwers A, Gardam MA. Determining mortality rates attributable to Clostridium difficile infection. Emerg Infect Dis 2012; 18:305–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Mlangeni DA, Harris MD, Franklin L, Hunt P, Karas JA, Enoch DA. Death certificates provide a poor estimation of attributable mortality due to Clostridium difficile when compared to a death review panel using defined criteria. J Hosp Infect 2011; 77:370–1. [DOI] [PubMed] [Google Scholar]
- 39.Laupacis A, Sekar N, Stiell IG. Clinical prediction rules. A review and suggested modifications of methodological standards. JAMA 1997; 277:488–94. [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.