JSTOR

Methods

A case patient was defined as a patient who had onset of diarrhea at least 48 hours after admission and a positive assay result for C. difficile toxin between January 1, 2004, and April 30, 2004. We also included in our analysis one patient who received antimicrobials during a prior admission and was readmitted with ileus, peritonitis, extreme leukocytosis, and a history of severe diarrhea. The patient died before his stool could be tested for C. difficile toxin.

To identify risk factors for CDAD, we randomly selected 3 control subjects for each case patient admitted during the study period. Control subjects were matched to case patients by age (within 10 years) and sex. Medical records for the 6 weeks before hospital admission were reviewed for both case patients and control subjects. Antimicrobial use, steroid use, other infections, the patient’s residence before admission (in a nursing home or non–nursing home), and length of hospital stay before the CDAD diagnosis were examined.

A pulmonologist and an infectious disease physician independently reviewed each case of CDAD to determine whether they thought the antimicrobial treatment that the patients received before they acquired CDAD was appropriate. The reviewers evaluated the chief complaint, vital signs, results of chest radiograph, oxygen saturation level, peripheral leukocyte count, and results of sputum Gram stain and culture.

The infection control professional reminded staff to wash their hands with soap and water after caring for patients with diarrhea. In addition, on March 10, 2004, (1) all patients with diarrhea were placed in contact isolation, and tests for C. difficile toxin were ordered; (2) each room was cleaned with a quaternary disinfectant followed by cleaning with a 1:10 dilution of bleach; and (3) physicians were discouraged from using antimotility agents for patients with CDAD.

Univariate and multivariate conditional logistic analyses were performed using exact conditional logistic regression. Also, interrater reliability analysis was performed, and a kappa statistic (κ) was estimated. All statistical analyses were done using SAS version 9.1 (SAS Institute).

Results

In 2003, the hospital’s infection control professional identified only 1 case patient with nosocomial CDAD, and only 3 stool samples from inpatients tested positive for C. difficile during the previous year. However, in January 2004, the infection control professional identified 10 patients with nosocomial disease, and by April 2004, the laboratory had identified 14 unique positive test results from patients. The hospital’s laboratory used the Becton Dickinson toxin CD kit, which detects toxin A, and had not changed its method of detecting C. difficile since 1999.

Fifteen case patients with nosocomial disease were identified, 5 of whom died before discharge. The median age for the case patients was 83 years, and 61% of case patients and control subjects were female. The results of the univariate analysis revealed that CDAD was significantly associated with cephalosporin use ( ), length of stay before diagnosis of CDAD ( ), and admission from a nursing home ( ) (Table 1). We selected our final multivariate model on the basis of the Akaike information criterion (AIC), and it included cephalosporin use ( ) and length of stay before the CDAD diagnosis ( ) (Table 2).

Table 1.  Results of the Univariate Analysis of Case Patients with Clostridium difficile–Associated Disease and Control Subjects

Open New Window

Table 2.  Estimated Parameters and Odds Ratios (ORs) from the Multivariate Logistic Regression Model

Open New Window

Five of the 15 case patients and 4 of the 45 control subjects died (matched odds ratio, 5.1 [95% confidence interval, 1.16‐22.64]). Of the 4 case patients who received loperamide, only 1 survived. However, loperamide use was not significantly associated with death at the .05 significance level ( ).

Before developing CDAD, 9 case patients had received a diagnosis of pneumonia and were given treatment for it. Three other patients had received a diagnosis of other respiratory complaints (eg, shortness of breath) but were given treatment with antimicrobial therapy as if they had pneumonia. Both specialists agreed that 6 of 12 case patients given treatment for pneumonia did not have evidence of this infection. The interrater reliability between the reviewers was good ( ).

Discussion

We investigated an outbreak of severe CDAD in a small rural hospital. Unlike for most other reported outbreaks, the crude mortality rate associated with CDAD was extremely high (33%). Historically, the rate of fulminate C. difficile colitis ranges from 1% to 3%.⁵ In this outbreak, the use of loperamide could have contributed to the high mortality rate; 3 of the 5 patients who died received the medication. Recently, however, other groups have reported high rates of mortality and colectomy associated with CDAD,^1,2 and some severe cases were caused by a specific strain of C. difficile.^3,6 The risk factors we identified are consistent with those previously identified by other investigators (eg, cephalosporin use). However, the small sample size in our study and the fact that patients often were receiving numerous antimicrobial agents during their hospitalizations may have prevented us from identifying additional associations between CDAD and other antimicrobial agents (eg, fluoroquinolone use).

Most of our case patients with CDAD (80%) were given treatment for pneumonia before they developed CDAD. However, we noticed that several of these patients lacked clinical signs and symptoms (such as infiltrates, fevers, and elevated white blood cell counts) consistent with pneumonia. The 2 physicians who independently reviewed the CDAD cases and evaluated the appropriateness of antimicrobial use concurred with our assessment.

The hospital had recently introduced a pneumonia care plan to expedite and standardize the use of antimicrobial agents. This plan required that patients with community‐acquired pneumonia receive an intravenously administered cephalosporin followed by an orally administered fluoroquinolone. Because this outbreak occurred shortly after a severe influenza season, we think physicians frequently incorrectly diagnosed pneumonia in patients with influenza and then treated the patients with the broad‐spectrum agents recommended by the pneumonia care plan. Some of these patients subsequently acquired CDAD. This pattern may have occurred elsewhere, because other investigators have observed higher CDAD rates in winter months.⁷

After our investigation, we recommended that the hospital use a rapid influenza test and that physicians review how they apply the pneumonia treatment pathway. Also, we recommended that physicians continue to avoid giving treatment with antimotility medications to patients who have CDAD.

Recent guidelines recommend that all patients with community‐acquired pneumonia receive treatment within 4 hours of presentation.⁸ Moreover, the Joint Commission on Accreditation of Healthcare Organizations and the Centers for Medicare and Medicaid Services include rapid treatment of community‐acquired pneumonia as a performance measure that can affect accreditation or payment. However, guidelines and performance measures are useful only if physicians make the correct diagnosis. Physicians who feel pressured to give treatment to patients within 4 hours of presentation may start therapy before confirming the diagnosis, and once started, antimicrobial therapy might not be stopped, even if the diagnosis of pneumonia subsequently appears unlikely.

Antimicrobial treatment must be administered as quickly as possible. However, the inappropriate use of antimicrobials can harm patients, and rote adherence to a guideline can cause unintended adverse outcomes. Thus, physicians and hospitals need to focus not only on the timing of antimicrobial treatment for community‐acquired pneumonia but also on making the correct diagnosis.