Clinical prediction rules to optimize cytotoxin testing for Clostridium difficile in hospitalized patients with...

Although routine testing of hospitalized patients with diarrhea for Clostridium difficile cytotoxin has been advocated as a high-yield procedure, the rationale for this practice has been questioned. To target a low-yield subgroup for whom routine testing could be deferred, we derived a clinical decision rule for predicting results of the C difficile cytotoxin assay in hospitalized adults with diarrhea.

We hypothesized a priori that two variables, antibiotic use (within 30 days prior to testing) and history of significant diarrhea (new onset of > 3 partially formed or watery stools per 24 hour period), would be highly predictive of cytotoxin results, and obtained these data on 480 consecutive patients who underwent diagnostic testing for C difficile at a university hospital and affiliated Veterans Affairs medical center. For more detailed modelling, we recorded symptoms, signs, comorbidity, and other potential causes of diarrhea for 68 test positive patients (cases) and 265 randomly selected test negative patients (controls) within the study cohort.

The overall prevalence of positive cytotoxin assays was 14%. Prior antibiotic therapy (OR = 9.0, 95% CI 2.1-38.4), significant diarrhea (OR = 2.2, 95% CI 1.1-4.7), and abdominal pain (OR = 1.9, 95% CI 0.96-3.7) were independent predictors of cytotoxin assay results. The model discriminated patients with positive and negative assays with a receiver operating characteristic (ROC) area of 0.68; observed and predicted probabilities of a positive cytotoxin assay were well correlated over the entire range of observed probabilities (r2 = 0.86). A decision rule (defined as positive if prior antibiotic use and either significant diarrhea or abdominal pain are present) demonstrated sensitivity and specificity of 86 and 45%. When applied to the entire dataset (N = 480), a simplified a priori rule, defined as positive if both prior antibiotic use and history of significant diarrhea are present, demonstrated sensitivity, specificity, positive and negative predictive value of 80, 45, 18 and 94%, respectively (6% of those predicted to be cytotoxin-negative actually tested positive). Use of this rule would have averted 39% of cytotoxin assays in our study population.

Patients without prior antibiotic use and either significant diarrhea or abdominal pain are unlikely to have positive C difficile cytotoxin assay results, and may not routinely require cytotoxin testing.