Reducing the risk of severe complications among patients with Clostridium difficile infection

1Sunnybrook Health Sciences Centre; 2Division of Infectious Diseases & Clinical Epidemiology, Department of Medicine, University of Toronto, Toronto, Ontario Correspondence: Dr Nick Daneman, Division of Infectious Diseases & Clinical Epidemiology, G-Wing Room 106, Sunnybrook Health Sciences Centre, University of Toronto, 2075 Bayview Avenue, Toronto, Ontario M4N 3M5. Telephone 416-480-6100 ext 2791, fax 416-480-5808, e-mail nick.daneman@sunnybrook.ca Received for publication August 19, 2010. Accepted January 11, 2011 Patients seek health care to improve their well-being; however, each year, more than 200,000 Canadians acquire an infection during a hospital stay, and approximately 8000 die from these illnesses (1). Clostridium difficile colitis has long been among the most burdensome of hospital-acquired infections and, increasingly so, given the emergence of the virulent NAP1/027 strain (2). Over the past decade, the incidence of C difficile infections has doubled (3), and the attributable mortality rate has quadrupled (4). Previous studies (5-12) suggested that the likelihood of C difficile treatment failure and complications are greater for patients of advanced age, with underlying illness, fever, cognitive impairment, leukocytosis, hypoalbuminemia, renal failure, bowel obstruction or ileus, imaging evidence of colitis, colonoscopic evidence of pseudomembranes or intensive care unit location. These predictors are useful in identifying patients who may require more intensive monitoring, more aggressive treatments and earlier surgical referral. However, these nonmodifiable factors do not offer direct avenues to improve the outcomes of patients infected with C difficile. Therefore, the current study examined modifiable processes of care associated with an altered risk of complications among patients with C difficile infection. METHODS Study cohort A retrospective cohort study (with prospective case ascertainment) was conducted at Sunnybrook Health Sciences Centre (SHSC) – a large, 700-bed academic health sciences centre in Toronto, Ontario. All consecutive cases of C difficile infection were prospectively identified by the Department of Infection Prevention and Control between January 1, 2007, and December 31, 2008. As per standard provincial guidelines, the case definition required laboratory confirmation of a positive toxin assay, together with diarrhea or visualization of pseudomembranes on sigmoidoscopy, colonoscopy or histopathology (13). Diarrhea was defined as two or more loose/watery bowel movements in a 24 h period that was unusual or different for the patient, and with no other recognized etiology (13). During the period examined in the present study, stool testing at SHSC was performed using enzyme immunoassay (EIA) for C difficile toxins A and B (TECHLAB, Inverness Medical, United Kingdom). Patients were excluded from analysis only in the rare event that they were enrolled in an investigational study of a novel C difficile therapy (n=3), or if their medical chart was unavailable for abstraction after three attempts (n=5). oRiginal aRticle


Primary outcome measure
The primary outcome measure was a composite of severe complications including severe hypokalemia (potassium level lower than 2.5 mM), toxic megacolon (colonic distension of greater than 7 cm or cecal distension of greater than 12 cm), bowel perforation, lower gastrointestinal bleeding requiring blood transfusion, intensive care unit transfer or death before completion of treatment for an episode of C difficile infection (4).In secondary sensitivity analyses, the outcome definition was varied to exclude the least severe events (eg, hypokalemia) or least preventable complications (eg, outcome events occurring before positive diagnostic test results).Outcome events were adjudicated by retrospective chart review conducted by two of the study authors.

Nonmodifiable predictors (baseline characteristics and clinical presentation)
Extensive data regarding baseline patient characteristics were collected.Demographic data included age, sex and place of residence.The source of infection was determined prospectively and defined as nosocomial if the onset occurred more than 72 h after hospital admission, or was related to a previous admission to a health care facility within the previous eight weeks (13).Comorbidities of interest included the following: cardiac disease, lung disease, liver disease, renal disease, neurological disease, inflammatory bowel disease, malignancy, diabetes mellitus, HIV/AIDS, preceding surgery during the current admission and gastrointestinal tube feeding.The acuity of the clinical presentation during the first 48 h of illness was assessed by measurement of maximum heart rate, minimum systolic blood pressure (mmHg), maximum number of stools per day, presence of blood in stool, abdominal pain, confusion, maximum white blood cell count (×10 9 cells/L), maximum creatinine level (mM) and minimum albumin level (g/L).

Modifiable predictors related to processes of care
Emphasis was placed on modifiable care processes relevant to the diagnosis and treatment of patients with C difficile infection.Important diagnostic measures included the duration of symptoms before diagnosis (days), laboratory stool assay turnaround time (same day or longer), negative stool EIA results preceding a positive stool EIA result (as a measure of impact of imperfect test sensitivity), and the use of abdominal x-ray testing or abdominal computed tomography scans.Treatment measures included the following: delays in treatment beyond 24 h from the time of the positive C difficile test result, any use of metronidazole, any use of oral/rectal vancomycin, inclusion of vancomycin in the initial therapeutic regimen, adjunctive use of probiotics (Saccharomyces boulardii or Lactobacillus preparations), use of potentially exacerbating antibiotic treatments after C difficile diagnosis and the use of antimotility agents (diphenoxylate or loperamide).

Statistical analysis
Univariate analysis was undertaken to compare each modifiable and nonmodifiable predictor variable among C difficile episodes with and without severe acute complications.The c 2 test was used to assess differences for binary predictors, while Wilcoxon rank-sum tests were used to examine differences for continuous variables.Multivariable logistic regression was the primary analysis used to assess the incremental impact of modifiable and nonmodifiable predictors on the risk of severe complications during an episode of C difficile infection.Assuming a baseline complication rate of 22%, a power of 85% (alpha = 0.05) was estimated to detect a 15% increased risk of complications associated with each predictor (4,14).Generalized estimating equations were used to account for the fact that some patients experienced multiple C difficile episodes (15).Variable reduction was accomplished by backward selection after including significant predictors on univariate analysis (ie, P<0.1) as well as prespecified inclusion of initial vancomycin therapy.Laboratory tests (albumin), imaging tests (x-ray and computed tomography scans) and endoscopy were not performed in the majority of patients and, therefore, could not be included in multivariable models.To assess the robustness of the results, secondary multivariable analysis was also repeated with variations in the composite outcome (as described above).All analyses were performed using SAS version 9.1 (SAS Institute, USA).The study was approved by the research ethics board of the SHSC.

Risk and timing of severe complications
Severe complications occurred in 97 of 365 (27%) episodes of C difficile infection.Table 1 summarizes the rates of each individual complication.These complications occurred rapidly (median three days postdiagnosis, interquartile range one to nine days), and 15 (16%) preceded the positive stool assay (Figure 1).

individual nonmodifiable predictors of C difficile complications
Different baseline characteristics were demonstrated between patients who experienced uncomplicated and complicated episodes of C difficile   2).Severe complications were more common among older patients, and those with underlying cardiac or neurological disease.Clinical presentation also differed between uncomplicated and complicated cases (Table 3).A greater risk of severe complications was associated with tachycardia, hypotension, confusion, recurrent episodes of infection, leukocytosis, elevated creatinine levels and hypoalbuminemia.Results were similar when recurrent episodes of C difficile infection were excluded from the analysis (data not shown).

Modifiable predictors of C difficile complications
Diagnosis of C difficile infection was delayed beyond four days of diarrhea symptoms in 20% of patients; however, there was no difference in diarrhea duration between uncomplicated and complicated episodes (Table 4).Laboratory results were not available on the same calendar day as specimen collection for 93 (25%) episodes, with no differences between uncomplicated and complicated episodes.An initial stool assay was negative before the positive result in 59 (16%) episodes, which was associated with a delay in diagnosis (7±8 days versus 3±5 days; P<0.0001), but not with an increased risk of complications.Abdominal x-rays and computed tomography scans were more frequently obtained for patients with complicated compared with uncomplicated disease; however, the proportion of abnormal imaging findings was similar in both groups.Initiation of effective antibiotic therapy (metronidazole or vancomycin) was delayed more than one calendar day from the positive stool assay result for 20% of episodes; however, the likelihood of delayed therapy was similar between complicated and uncomplicated   cases (Table 4).Almost all C difficile infections were treated with metronidazole (92%), and only a minority of cases were treated with vancomycin at some point in their course (28%).There was a trend toward greater use of vancomycin in the initial treatment regimen for uncomplicated versus complicated episodes (8.3% versus 4.1%; P=0.18).The use of probiotics and antimotility agents was similar between complicated and uncomplicated episodes, but the continued use of exacerbating antibiotics after diagnosis was far more common among patients who experienced complicated episodes of infection (78% versus 54%; P<0.0001).
Multivariable analysis: independent predictors of severe C difficile complications Multivariable analysis yielded six independent predictors of severe C difficile complications, including four nonmodifiable factors and two potentially modifiable factors ( ).These findings were unchanged in sensitivity analyses, which excluded the least severe outcome (hypokalemia) and the least preventable outcomes (those developing before the date of positive stool specimen) (data not shown).

DiSCuSSiON
Our study highlights the gravity of C difficile infection among hospitalized patients, as well as prospects to improve their outcomes.Severe complications were common among patients with C difficile infections, occurring in more than one in four episodes, and corroborated the findings of a recent national surveillance study (4).These complications occurred rapidly, often even before confirmatory test results, thereby emphasizing the importance of infection prevention and control measures to prevent new cases of infection (16).However, recent literature suggests that high-performing institutions are distinguished not only by their prevention of iatrogenic illnesses, but also by their capacity to detect and manage them (17).Our study revealed some ways that detection and treatment of C difficile episodes can be more effectively optimized.Occasionally, diagnosis of C difficile infection is delayed, with 20% of patients experiencing more than four days of diarrhea before diagnosis, due in part, to the imperfect sensitivity of the EIA and initial negative test results before a positive specimen.However, neither initial negative assay results nor delayed diagnosis were associated with an increased risk of complications, suggesting that there is little yield in improved diagnostic strategies.
In contrast, our data indicated that there is substantial room for improving C difficile outcomes by optimizing treatment.The use of vancomycin as initial therapy was associated with a 76% reduction in the risk of severe complications.Our findings confirm the results of a recent randomized controlled trial that demonstrated the superior efficacy of vancomycin over metronidazole (18), and suggests that vancomycin should be considered as first-line therapy in patients with moderate-to-severe C difficile infections.There is a theoretical risk of increasing colonization rates of vancomycinresistant Enterococci by promoting vancomycin to first-line therapy for C difficile; however, this has not been confirmed in recent clinical studies (19).Reserving vancomycin use for patients who do not respond to metronidazole therapy may also be inappropriate, given the rapid tempo of illness and early onset of complications.
Perhaps even more important than the selection of the best anti-C difficile agent is the discontinuation and avoidance of other exacerbating antibiotic agents.Ongoing or newly started antibiotics hinder restoration of bowel flora and, in our study, were associated with a threefold higher risk of severe complications.More than one-third of hospital antibiotic use is unnecessary or inappropriate (20,21).Patients diagnosed with C difficile infection warrant targeted attention by antibiotic stewardship programs.In contrast, there was no detectable hazard associated with the use of antimotility agents, which theoretically can increase colonic exposure to C difficile toxins A and B. Given that the use of these medications is uncommon, our study was not powered to test this association.Nevertheless, our findings are consistent with postmarketing surveillance information for loperamide and diphenoxylate, which revealed no adverse outcomes among patients who were concurrently receiving metronidazole or vancomycin (22).
Although our focus was on modifiable predictors, we did detect several nonmodifiable characteristics associated with a higher risk of C difficile complications including recurrent episodes of infection, hypotension, confusion, leukocytosis and hypoalbuminemia.These findings were consistent with the literature, and confirm that these factors identify higher-risk patients who may benefit from more aggressive monitoring and treatment.
Our study was limited to a single tertiary care centre, and processes of care may differ at other institutions.However, we believe that our data provide broad lessons for the care of C difficile patients that are likely generalizable to most hospital settings.As with all retrospective studies, there was a potential for nondifferential misclassification of predictor and outcome variables that would attenuate OR estimates toward the null.We attempted to overcome this issue by selecting objective measurements.Moreover, a prospective study of care processes may not be more helpful due to inevitable Hawthorne effects.A prospective study, however, would be especially helpful in accurately determining the appropriateness of other antibiotic therapies and defining attributable causes of death.Finally, we lacked data regarding C difficile strains; however, this information is not available to clinicians and, therefore, would not be relevant to risk prediction in the setting of routine patient care.
Given the frequency, severity and early onset of C difficile complications, it is important to identify patients who are at increased risk of these outcomes.In this regard, useful markers include confusion, hypotension, leukocytosis and hypoalbuminemia.However, the simplest and most effective way to optimize the management of patients with C difficile infection is to consider the use of vancomycin as first-line therapy for moderate to severe cases, and discontinue or avoid the use of unnecessary antibiotics.

Figure 1 )
Figure 1) Timing of Clostridium difficile complications.The median time to C difficile-related complications was three days following microbiological diagnosis (interquartile range one to nine days).Fifteen (16%) of the complications preceded the date of diagnosis

aCKNOWlEDGEMENTS:
The authors acknowledge Ted Winkle for abstraction of pharmacy treatment data, and Laura Rosella for statistical advice.FiNaNCial SuPPORT: Dr Nick Daneman is supported by a Clinician Scientist award from the Canadian Institutes of Health Research, and a Patient Safety Scholar Award from Sunnybrook Health Sciences Centre.CONFliCTS OF iNTEREST:The authors have no conflicts of interest to declare.

Table 1 Complications among 365 episodes of Clostridium difficile infection Complication C difficile episodes with complication(s), n (%)
*Potassium level lower than 2.5 mM; † Requiring blood transfusion of one or more units of red blood cells; ‡ Denominator does not include patients already in intensive care unit at diagnosis; § C difficile episodes involving at least one of the above complications

Table 2 baseline characteristics of patients with uncomplicated or complicated episodes of Clostridium difficile infection baseline characteristic C difficile episodes P*
Data presented as n (%) unless otherwise indicated.*Reflects Wilcoxon ranksum test for continuous variables and c 2 test for differences in proportions

Table 4 Investigation and treatment in uncomplicated and complicated episodes of Clostridium difficile infection
Data presented as n (%) unless otherwise indicated.*ReflectsWilcoxonranksum test for continuous variables and c 2 test for differences in proportions;† From positive test result to physician order for vancomycin or metronidazole.CT Computed tomography

Table 3 Clinical presentation in uncomplicated and complicated episodes of Clostridium difficile infection
Data presented as mean ± SD unless otherwise indicated.*Reflects Wilcoxon rank-sum test for continuous variables and c 2 test for differences in proportions