Clostridium difficile (C. difficile) is the predominant infectious cause of healthcare-associated diarrhea and one of the most common types of healthcare-acquired infection, resulting in prolonged hospital stays, higher mortality, and increased healthcare costs. Exposure to antibiotics is the most important risk factor for C. difficile infection (CDI), presumably through the disruption of the normal fecal flora. Although a number of approaches have been proposed to contain outbreaks of CDI, such as improved hand hygiene and antibiotic stewardship, little is understood about how these interventions alter the dynamics of C. difficile transmission and acquisition and contribute to its control.
C. difficile transmission is dependent on the interactions of innumerable factors and processes. The design of policies to control nosocomial CDI is aided by an understanding of these interactions and the relative impact of different control strategies on C. difficile transmission dynamics. With this in mind, our objectives were to (a) describe variation in antimicrobial prescribing and perform a facility-level analysis of associations between antibiotic use patterns and CDI rates at VA hospitals nationwide; (b) incorporate these patterns and associations into our agent-based simulation of nosocomial C. difficile transmission; (c) use the simulation to evaluate and compare alternative and novel policies for C. difficile control in VA hospitals, including antibiotic stewardship; and (d) explore the impact of these intervention strategies under varying conditions, including the introduction of an epidemic C. difficile strain.
We refined a high-fidelity agent-based computer simulation of nosocomial C. difficile transmission created as part of a previous project. Our analyses of antibiotic prescribing and CDI rates were based on large, nationwide databases of VA patient data. The combination of individual- and hospital-level data from more than 150 VA hospitals made it feasible to fit models that separately estimated direct effects of antimicrobial agents on CDI risk from their indirect effects mediated through person-to-person spread. Hierarchical mixed effects models were used to characterize the association of CDI rates to patient and hospital level factors. Results of these analyses were incorporated into the simulation, which was then used to assess the various strategies and factors that impact C. difficile transmission through a series of simulation experiments. Traditional quantitative epidemiologic methods were used to analyze simulation results, with a focus on C. difficile incidence and transmission rates as outcomes. Dynamic cost-benefit analyses were performed by projecting C. difficile incidence rates and costs under the various alternative policy regimes.
For Aim 1, we used VA data to categorize the 14 antibiotics that together accounted for over 95% of antibiotic daily doses in 113 acute care VHA facilities as: 1) high risk for CDI, 2) medium risk for CDI, 3) neutral effect on CDI, and 4) used to treat CDI. Use was highest for medium risk antibiotics, roughly equal for high risk and neutral antibiotics, and lowest for CDI treatment antibiotics. The facility rate of patient-days with an antibiotic was similar for high risk, medium risk, and neutral antibiotics (mean rates were 215, 200, 170 antibiotic days per 1000 patient-days for high, medium, and neutral risk, respectively) and approximately three times the rate of antibiotics to treat CDI. We then assessed the risk of developing CDI upon exposure to these three antibiotic classes by calculating the hazard ratio (HR) for each class, compared to no exposure to any class. The HR for developing CDI while being treated with high, medium, and neutral risk antibiotics was 2.03, 1.72, and 1.26, respectively, suggesting that some antibiotics in the neutral risk category do carry some small risk of CDI. In the immediate period after stopping the antibiotic, the HR of developing CDI was 2.06, 1.98, and 1.06 for high, medium, and neutral risk classes, which suggests that the risk of developing CDI is high even after stopping high and medium risk antibiotics.
We also performed analyses of the antibiotic data by traditional antibiotic class, rather than by categorizing by CDI risk, in order to determine the association between exposure to the different antibiotics classes and the development of CDI. To do this, we also calculated the HR for the development of CDI for the time period during antibiotic treatment as well as the immediate post-antibiotic period once the treatment had been discontinued. The antibiotic classes with the highest HRs included the carbapenems, third- and fourth-generation cephalosporins, and penicillins at 5.74, 3.39, and 2.03, respectively, during treatment and 4.14, 2.62, and 2.19, respectively, in the post-antibiotic period. Interestingly, the classes thought to contribute the most to CDI risk, fluoroquinolones and clindamycin, had lower HRs. In fact, while fluoroquinolones had HRs of 1.36 and 1.78 in the treatment and post-treatment periods, respectively, clindamycin had a HR of 0.85 while on therapy and 1.34 during the post-antibiotic period. This suggests that fluoroquinolones may be less of a risk for CDI than previously thought, and that clindamycin may actually be protective for CDI while on therapy, though the HR is not statistically significant, likely owing to a fairly low level of clindamycin use in the VA system.
Additional analyses focusing on intravenous (IV) vancomycin showed that, although the receipt of IV vancomycin increased the risk of CDI among patients receiving a CDI-neutral antibiotic nearly 2-fold, the absolute increase in risk was very small; thus, IV vancomycin is not a substantial contributor to CDI risk among inpatients on CDI-neutral antibiotics, a finding that will help to clarify this controversial topic. Other analyses of the data also showed large variations in the management of common infectious diseases in the choice of empiric agents and duration of therapy. Specifically, while empiric antibiotic use during the first 2 hospital days was frequent for all infectious diagnoses, type of empiric coverage and treatment duration varied significantly among facilities for all conditions. Additionally, our work investigating practice variation in the propensity to start and stop antibiotics revealed that even VAs with similar levels of antibiotic use showed marked differences in their rates of starting and stopping antibiotics.
This work supported the parameterization of our CDI simulation model for Aim 2. Day-level distributions were used to reflect realistic antibiotic prescribing patterns in the agent-based model. Following this, we used the simulation to explore the impact of altering antibiotic use through a simulated antibiotic stewardship program, by 1) changing overall facility antibiotic usage by changing the antibiotic start/stop assumptions and 2) changing the percentage of antibiotics prescribed from the high- and neutral-risk CDI categories to more in the low-risk category. Results of this work showed decreases in C. difficile transmissions and CDI incidence with shifts in prescribing, with a greater decrease with shifts from high-risk to neutral risk than from neutral risk to low-risk. These results are consistent with theory and results from other publications, and suggest that a more fruitful target for stewardship efforts is on the high-risk, high usage facilities.
Our work thus far has shown the degree of and changes in inpatient antimicrobial prescribing that have occurred across the VA over the past few years, noting the increase in prescribing of high-cost, high-risk, broad-spectrum agents, and the significant variation in prescribing that exists across all VA hospitals. It also showed that these high-risk antibiotics have a strong relationship with CDI incidence, and that lowering the prescribing of these high-risk antibiotics through antibiotic stewardship efforts can have a significant impact on C. difficile transmission and infection among Veterans. This work revealed useful targets for VA antimicrobial stewardship interventions to reduce inappropriate and excessive antibiotic prescribing, particularly to control CDI in VA care facilities and reduce the risk to Veterans. This information will be valuable for informing antimicrobial stewardship efforts in VA facilities nationwide.
- Stevens VW, Khader K, Nelson RE, Jones M, Rubin MA, Brown KA, Evans ME, Greene T, Slade E, Samore MH. Excess Length of Stay Attributable to Clostridium difficile Infection (CDI) in the Acute Care Setting: A Multistate Model. Infection control and hospital epidemiology. 2015 Sep 1; 36(9):1024-30.
- Rubin MA, Jones M, Leecaster M, Khader K, Ray W, Huttner A, Huttner B, Toth D, Sablay T, Borotkanics RJ, Gerding DN, Samore MH. A simulation-based assessment of strategies to control Clostridium difficile transmission and infection. PLoS ONE. 2013 Nov 21; 8(11):e80671.
- Staes C, Jacobs J, Mayer J, Allen J. Description of outbreaks of health-care-associated infections related to compounding pharmacies, 2000-12. American journal of health-system pharmacy : AJHP : official journal of the American Society of Health-System Pharmacists. 2013 Aug 1; 70(15):1301-12.
Mental, Cognitive and Behavioral Disorders, Health Systems, Cardiovascular Disease, Infectious Diseases
Prevention, Treatment - Efficacy/Effectiveness Clinical Trial, Treatment - Comparative Effectiveness
Bipolar Disorder, Care Coordination, Comparative Effectiveness, Computational Modeling, Computer Simulations, Depression, Healthcare Algorithms, Outcomes - Patient, Patient Safety, Risk Factors, Schizophrenia, Serious Mental Illness