Development and Validation of a RecurrentClostridium difficileRisk-Prediction Model

    loading  Checking for direct PDF access through Ovid

Abstract

BACKGROUND:

Recurrent Clostridium difficile infection (rCDI) affects 10% to 25% of patients with initial CDI (iCDI). Initiation of new therapies that reduce recurrences rests on identifying patients at high risk for rCDI at iCDI onset.

OBJECTIVE:

To develop a predictive model for rCDI based on factors present at iCDI onset.

DESIGN:

Retrospective cohort study.

SETTING:

Large urban academic medical center.

PATIENTS:

All adult patients with an inpatient iCDI from January 1, 2003 to December 31, 2009.

INTERVENTION:

None.

MEASUREMENTS:

Positive toxin assay for C difficile with no C difficile infection in the previous 60 days constituted iCDI. Repeat positive toxin within 42 days of stopping iCDI treatment defined rCDI. Three demographic, 13 chronic, and 12 acute disease characteristics, and 7 processes of care prior to or at the onset of iCDI, were assessed for association with rCDI. A logistic regression model to identify predictors for rCDI was developed and cross-validated.

RESULTS:

Among the 4196 patients enrolled, 425 (10.1%) developed rCDI. Six factors (case status as community-onset healthcare-associated, ≥2 hospitalizations in the prior 60 days, new gastric acid suppression, fluoroquinolone and high-risk antibiotic use at the onset of iCDI, age) predicted rCDI in multivariate analyses, whereas intensive care unit stay appeared protective. The model achieved moderate discrimination (C statistic 0.643) and calibration (Brier score 0.089). Its negative predictive value was 90% or higher across a wide range of risk.

CONCLUSIONS:

Among patients hospitalized with rCDI, multiple factors present at the onset of iCDI increased the risk for rCDI. Recognizing patients at high-risk for rCDI can help clinicians tailor early treatment and prevention. Journal of Hospital Medicine 2014;9:418–423. © 2014 Society of Hospital Medicine

Related Topics

    loading  Loading Related Articles