Objective: To define clinical and laboratory variables that suggest the presence of Clostridium difficile colitis and to establish the number of stool specimens needed to reasonably exclude the diagnosis of C. difficile colitis.
Design: Prospective study of consecutive inpatients whose stool specimens were sent to be evaluated for the presence of C. difficile toxin.
Setting: University teaching hospital.
Patients: 268 hospital inpatients in medical, surgical, and gynecology units.
Measurements: Structured history and physical examination; detection of C. difficile toxin by cytotoxin tissue-culture assay with anti-C. difficile antiserum neutralization and by enzyme-linked immunoassay (EIA) for C. difficile toxins A and B; and detection of fecal leukocytes by microscopic examination and by latex agglutination lactoferrin assay.
Results: 43 of 268 consecutive inpatients were positive for C. difficile toxin by EIA or tissue-culture assay. Although toxin was detected by EIA alone in 39 of the 43 patients, it was detected in an additional 4 patients (10%) by tissue-culture assay alone. Univariate and multivariate logistic regression analysis showed that the following clinical and laboratory features were associated with C. difficile toxin positivity: the onset of diarrhea 6 or more days after the administration of antibiotics (odds ratio, 1.38 [95% CI, 1.10 to 3.79]); hospital stay longer than 15 days (odds ratio, 1.33 [CI, 1.09 to 3.95]); the presence of fecal leukocytes determined by microscopy (odds ratio, 2.39 [CI, 1.05 to 5.42]) or lactoferrin assay (odds ratio, 3.74 [CI, 1.80 to 7.76]); the presence of semiformed (as opposed to watery) stools (odds ratio, 2.33 [CI, 1.10 to 4.90]); and cephalosporin use (odds ratio, 2.36 [CI, 1.10 to 5.09]). Toxin-positive patients were no more likely than controls to have had fever, abdominal pain or cramps, leukocytosis, green-colored diarrhea, or blood in the stool or to have received clindamycin or penicillin derivatives.
Of the 43 patients with C. difficile toxin, 34 (79%) had positive results for the toxin on the first stool specimen, 5 (cumulative, 91%) had positive results on the second specimen, and 4 had positive results on the third specimen. Overall, the negative predictive value of the first stool specimen was 97%. All patients who had two or more clinical or laboratory predictors were diagnosed with C. difficile disease when either the first or the second stool specimen was positive for toxin.
Conclusions: Clinicians at the bedside can use readily available clinical and laboratory information to decide which patients are likely to have C. difficile disease and when it is appropriate and useful to order specific diagnostic tests for C. difficile toxin. Such data are also useful in determining the number of stool samples that reasonably excludes the diagnosis of C. difficile colitis.