Evaluating the Risk Factors for Hospital-Onset Clostridium difficile Infections in a Large Healthcare System

Abstract

In this multicenter retrospective cohort study of over 1 million patients at 150 US hospitals, proton pump inhibitors increased the odds of a patient having hospital-onset Clostridium difficile infection as did third and fourth generation cephalosporins, carbapenems, and piperacillin/tazobactam. These findings support appropriate prescribing of acid-suppression therapy and high-risk antibiotics.

Clostridium difficile infection (CDI) remains a persistent challenge, with more than 450000 infections and 29000 related deaths per year in the United States [1]. The number of CDIs is projected to increase, to some extent caused by greater spread of hypervirulent strains resistant to commonly used antibiotics [2].

Certain risk factors are associated with increased likelihood of CDI, such as increased age, female sex, admission from a long-term acute care (LTAC) facility, immunosuppression, length of stay, and exposure to antibiotics [3]. Many studies have included exposure to antibiotics as a composite risk factor, with a possible association between CDI and the use of fluoroquinolones, third-generation cephalosporins, clindamycin, macrolides, and carbapenem [4].

Proton pump inhibitors (PPIs) have been associated with a 70%–80% increase in the risk of developing CDI and a 2.5-fold increase in the risk of recurrent infection in meta-analyses of observational studies [4–6]. However, these analyses also found statistically significant heterogeneity among the studies included, as well as differing definitions of CDI. A population-based study of 385 patients showed an apparent increased risk of CDI associated with the use of PPIs and H2 antagonists, but this finding did not persist after adjusting for age and comorbid conditions [7]. Other studies have also shown little or no association between PPIs and the risk of CDI [8, 9].

There is a need for a more robust investigation of CDI risk factors and the relationship between PPI use and CDI. Accordingly, this study used a very large database of over one million patient encounters to analyze the association of hospital-onset (HO) CDI with the use of PPIs and other acid suppressing drugs and frequently used antibiotics.

METHODS

This retrospective cohort study used data from 150 hospitals across the United States associated with a large healthcare system (Hospital Corporation of America). Adults at least 18 years old discharged between October 1, 2015, and September 30, 2016, were included. Patients were excluded if pregnant or delivering a baby, had a positive stool test for Clostridium difficile collected on hospital day ≤3, or had missing data elements. International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnosis codes were used for clinical diagnosis. The study was approved by the University of Tennessee Health Science Center Institutional Review Board.

Patients were identified from the National Healthcare Safety Network (NHSN) database and classified as HO CDI positive and HO CDI negative. Demographic, medication, and laboratory data were collected from a centralized clinical data warehouse of electronic health record data. All patient and location identifiers were removed prior to analysis.

HO CDI was defined as a positive stool test for toxigenic Clostridium difficile or toxin-coding genes from an unformed stool collected on hospital day >3. The day the positive stool sample was produced was considered as the day of CDI onset for this study. This definition was based on a laboratory test data only and was not a clinical case definition, in alignment with the CDC classification of HO CDI [10].

Initial variable selection was determined based on a critical review of the literature related to Clostridium difficile. Variables were analyzed using t-tests and χ² tests in order to determine which ones differed the most between HO CDI positive and HO CDI negative groups. These analyses were reviewed by clinical experts to make the final determination of what variables should be included in the regression model.

Variables were grouped into 3 categories (demographics, acid suppressing/protecting medications, antibiotic medications) in a multivariate logistic regression model. Acid suppressing/protecting medications included antacids, H2 antagonist, PPIs, and sucralfate (see Table 1). Antibiotic medications included aminoglycosides, ampicillin and sulbactam, carbapenems, cephalosporins (1st, 3rd, 4th generation), clindamycin, fluoroquinolones, macrolides, metronidazole, penicillins, piperacillin/tazobactam, sulfamethoxazole/trimethoprim, and tetracyclines (see Table 1).

Statistical analyses were performed using SAS, version 9.4 (SAS Institute Inc., Cary, NC). To check for multicollinearity between the variables, a correlation matrix was created between the parameter estimates of a fitted model using PROC GENMOD in SAS. All correlations were found to be weak; therefore, the assumption of independent variables was met. A multivariate logistic regression model was used to evaluate the risk-factors related to CDI, generating the coefficients for the model as well as the odds ratios using PROC LOGISTIC in SAS. Statistical significance was defined as P ≤ .05.

RESULTS

Of the 1237537 patients included in the analysis, 4587 patients (0.37%) were classified as HO CDI positive. The HO CDI positive patient population was slightly older with more patients admitted from LTAC and a higher case mix index (3.44 vs 1.65; P < .001). The odds of a patient having HO CDI increased by 0.5% for each year in age. The odds of CDI in females was 1.2 times as likely as in males. Admission from a LTAC, comorbidities of diabetes, Crohn’s disease, or ulcerative colitis, days in the ICU, and antibiotic days of therapy all significantly increased the odds of a patient having HO CDI.

A larger percentage of patients with HO CDI were receiving acid suppressing/protecting medications (Table 1). PPIs increased the odds of HO CDI by 44% (P < .001), whereas the use of H2 antagonists increased the odds ratio for HO CDI to a lesser, but still significant, extent (13% increase, P < .001). The odds of HO CDI were significantly higher in patients taking carbapenems, third and fourth generation cephalosporins, metronidazole, and piperacillin/tazobactam, as well as patients on multiple antibiotics. In contrast, treatment with tetracyclines, macrolides, or clindamycin significantly reduced the odds of HO CDI (Table 1). The combination of PPIs with FQs, third generation cephalosporins, fourth generation cephalosporins, clindamycin, or carbapenems did not significantly alter the odds of HO CDI.

DISCUSSION

In this large data set, PPIs and other acid suppressing/protecting medications increased the odds of HO CDI, as did certain classes of antibiotics. Our results are consistent with other published observations that PPIs have a greater effect on increasing HO CDI than other acid-controlling medications [11, 12]. This study also confirms previous findings that carbapenems and third-generation cephalosporins increase the odds of HO CDI [13, 14], as well as fourth-generation cephalosporins, and piperacillin/tazobactam. In contrast, we found that clindamycin, macrolides, and tetracyclines significantly reduced the odds ratios for CDI. No additional effect was not found for any combination treatments, which did not support the meta-analysis that concluded the use of PPIs and antibiotics together lead to a greater risk than PPIs alone [4].

The clindamycin findings are somewhat surprising in that clindamycin has been classified as a high risk for development of a CDI in both the hospital and in the community [14, 15]. This finding may be due to reduced use of clindamycin over time. The tetracycline and macrolide findings are consistent with literature supporting the notion that tetracyclines have potential protective effect against developing CDI [16], although we did not confirm the protective effect of the use of a combination of doxycycline with other high-risk antibiotics.

We also did not find an increase in the odds of HO CDI with fluoroquinolones. This could be due to fluoroquinolone stewardship, which has been shown to decrease the rate of CDI, or the circulating strain of Clostridium difficile. Because our study did not perform any strain typing, we cannot definitively determine what strains were prevalent in our study population. It is possible that the circulating strain type or changing prescribing patterns could have impacted the observed relationship of antibiotics to the odds of HO CDI.

This retrospective cohort study used clinically collected data and therefore must contend with a number of uncontrolled factors, which may have contributed to the results such as outpatient medications and over the counter substances (e.g., probiotics), dose, exposure duration, lifestyle factors, and prior hospitalization or medication use. For example, it is possible that for some antibiotics (e.g., fluoroquinolone and tetracyclines), administration prior to hospitalization allows for adaptive responses in the microflora to occur so as to either enhance or reduce the incidence of in-hospital CDI. Our data did not allow for the identification of patients with a history of CDI or antibiotic use prior to hospitalization; therefore, there is a possibility of confounding from unidentified patients who would potentially have baseline increased odds of CDI [17]. Furthermore, patients were identified through the NHSN database where data might be electronically or manually entered and the potential for human error exists. In addition, the NHSN definition is a Lab-ID definition and does not rely on chart review or symptoms.

In summary, this study leveraged a large database and confirms and supports previous studies demonstrating an increased odds of HO CDI associated with the use of PPIs and certain high-risk antibiotics. It is interesting to note that both acid suppressing medications and antibiotics affect the gastrointestinal microbiome [18, 19]. Whether or not there are common pathways by which these medications increase the risk for development of a CDI remains to be determined. Nonetheless, the results support the need for stewardship practices around both high-risk antibiotics and medications that alter gastric acid regulation. Furthermore, the impact of de-prescribing acid suppression therapy coupled with antibiotic stewardship could greatly reduce the incidence of HO CDI.

Notes

Acknowledgments. The authors acknowledge individuals for their expertise and support in this study: Kenneth Sands, MD; Karla Miller, PharmD; Julia Moody, MS; Kacie Kleja, MS, MBA; and Jonathan B. Perlin, MD, PhD, MSHA, FACP, FACMI.

Potential conflicts of interest. All authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References