Does Choice of Control Group Affect the Association of Antibiotics with Clostridium difficile-Associated Diarrhea?

Background: The incidence and severity of Clostridium difficile-associated diarrhea (CDAD) has been increasing in the past decade.  Use of certain antibiotics (ABX) has been associated with CDAD among other factors.  Interventions aimed to reduce inappropriate use of these ABX have shown to reduce CDAD rates.  By identifying high-risk ABX, antimicrobial stewardship program (ASP) can design strategies to decrease CDAD rates.

Objective: To determine if different control groups result in different ABX association with CDAD that can be used as ASP intervention target to reduce CDAD.

Methods: Antibiotic exposure (> 30 ABX alone and in combination) was compared between CDAD case and 5 sets of CDAD-negative controls admitted from 7/09 to 10/09 (within 30 days of diagnosis) at our 800 bed facility.  Each control cohort was matched to case by admission date and admission type, age, total length of stay (LOS) and/or LOS to event [i.e., case LOS to positive C difficile (CD) matched to control LOS to discharge].  Logistic regression analyses were performed for each case-control cohort to determine ABX independently associated with CDAD.

Results: A total of 126 cases were compared to 252 controls in each cohort.  Case patients were 42% male; have mean age of 67 years, length of stay (LOS) to CD of 7 days, and total LOS of 16 days.  Of all cases, 111 (88%) were on ABX, with 28 (22%) on >3 ABX.  Most commonly used ABX were intravenous vancomycin (VA; 37%), piperacillin/tazobactam (PT; 34%), fluoroquinolones (FQ; 28%), high-generation cephalosporins (HGC; 23%), and last resort drugs (LRD, e.g., polymyxin B, linezolid; 13%).  Controls were significantly younger (except age-matched controls), have shorter LOS (except LOS-matched controls), and less likely to be on any ABX.  Case patients were consistently more likely than patients in all 5 control cohorts to be on VA [3-16%, p <0.01, odd ratio (OR) range 3-21], PT (2-12%, p <0.01, OR 4-26), FQ (3-13%, p <0.01, OR 3-12), HGC (1-11%, p <0.01, OR 2-37), and LRD (0-3%, p<0.01, OR 4-undefined).  Logistic regression analyses reveal PT (p <0.01, OR 4-17), FQ (p <0.01, OR 2-9), and HGC (p <0.01, OR 2-28) use were independently associated with developing CDAD.  These 3 ABX represent a majority of the ABX use in case patients.

Conclusions: Regardless of matching criteria used to identify control patients, use of PT, FQ, or HGC remained independently associated with CDAD. The choice of control seems less important given the overall large burden of ABX use in CDAD patients compared to those with no CDAD.  ASP-initiated interventions to reduce inappropriate use of these agents and validation of these interventions to decrease CDAD rates at our institution are ongoing.   Our project was not designed to identify all factors associated with CDAD, but rather a simple and feasible (stripped down) exercise that could be performed by facilities regardless of resources as a planning tool for ASP targets and activities.