Risk estimation for recurrent Clostridium difficile infection based on clinical factors.

Abstract

The incidence of Clostridium difficile infection (CDI) has risen dramatically during the last decade. Although patients respond well to medical therapy such as vancomycin, 20%-30% of patients treated suffer a recurrence of CDI. We developed a simple/practical scoring rule (logistic regression model) for recurrent CDI using data from 2 large phase 3 clinical trials. Seventy-seven baseline CDI factors were classified: demographics, comorbidity, medications, vital signs, laboratory tests, severity, and symptoms. Predictors with the highest discrimination in each class (using receiver operating characteristics curve) were selected. For the final model, stepwise selection was performed. Discrimination, calibration, and internal validation were used to assess the model. The final model with a simple scoring rule was developed. It includes 4 independent risk factors that are readily available when the patient makes initial contact: age (<75 vs ≥75 years), number of unformed bowel movements during previous 24 hours (<10 vs ≥10), serum creatinine leves (<1.2 mg/dL vs ≥1.2 mg/dL) and prior episode of CDI (yes vs no). In addition, the model includes choice of treatment (vancomycin or fidaxomicin). The prediction model for recurrence may be useful for treatment decision. NCT00314951 and NCT00468728.