Retrospective Cohort Analysis of Outpatient Antibiotic Use for Clostridioides difficile-Indicated Agents in British Columbia, from 2000 to 2018

Background Clostridioides difficile (CDI) is the most common cause of nosocomial diarrheal infections. Historically, metronidazole was the first-line treatment, but guidelines now indicate oral vancomycin and fidaxomicin as primary antibiotics for initial episodes. A provincial stewardship program has operated in British Columbia (BC), since 2005. Since the program's inception, surveillance of antibiotic use has been ongoing. However, this is the first study to review community-acquired CDI-indicated antibiotic use. Moreover, this study offers the first interpretation of fidaxomicin use in BC since its addition to the provincial formulary. Methods A retrospective cohort analysis included all outpatient dispensations for CDI-related antibiotics from January 1, 2000, to December 31, 2018. Antibiotic dispensations were extracted for metronidazole, vancomycin, and fidaxomicin. Consumption rates were calculated as prescriptions per 1000 population. Rates were examined overall and then stratified by medication, age, and sex. Secondary outcomes of interest included an examination of adherence to provincial special authority criteria; and proportions of outpatient antibiotic use attributable to administrative health records for CDI. Results The average annual rate of prescribing was 18.5 per 1000 population for all CDI-indicated antibiotics. The rate of prescribing increased (15%) over the 19-year study period, from 17.2 to 19.8 dispensations per 1000 population. Metronidazole accounted for the most antibiotics dispensed in every study year; however, by 2018 it demonstrated the most modest increase in use (15%). In comparison, fidaxomicin increased by 226% by 2018. Vancomycin had the highest percentage increase (621%), with the greatest change occurring from 2014 to 2015, correlating to the dissemination of new clinical practice guidelines. Conclusion This is the first study to evaluate outpatient prescribing for CDI-indicated antibiotics, and one of the few studies to examine fidaxomicin since its introduction to Canadian formularies. Although causation cannot be inferred from study results, oral vancomycin, and fidaxomicin use has increased in line with, or in advance-of guidelines.


Introduction
Clostridioides difcile (formerly Clostridium difcile) is the most common cause of hospital-associated diarrheal infections [1]. Pathogenic strains produce toxin A, B, or a binary toxin upon colonization of the gut, which can lead to a range of symptoms from diarrhea, colitis, and in severe cases death [2]. Clostridioides difcile infection (CDI) has traditionally been characterized as nosocomial, or healthcare-associated (HA), with the incidence of HA-CDI being 2.1/10 000 to 6.5/10 000 inpatient days [3]. However, mounting evidence shows that cases of community-associated CDI (CA-CDI) are increasing [4,5]. A case of CA-CDI is defned as either symptom onset in the community or outpatient setting; symptom development at least 12 weeks following a hospital discharge; or symptom onset within 48 hours after admission to a healthcare facility [1]. International data for CDI reports approximately 20-27% of CDI cases are community-acquired, with an annual incidence of 20-30 cases of CA-CDI per 100,000 population [5]. In British Columbia (BC), the incidence of CA-CDI has increased from 16.74 cases per 100 000 population in 2011-2012 to 17.78 cases per 100 000 population by 2016-2017 [3].
Treatment for CDI centers around antibiotic therapy [6,7]. Historically, metronidazole was the frst-line antibiotic treatment for CDI, but updates to IDSA guidelines now indicate oral vancomycin and fdaxomicin as frst-line agents, displacing metronidazole as the primary antibiotic for an initial episode of CDI [6]. Both oral vancomycin and fdaxomicin have been found to be superior to metronidazole against the initial episode, frst recurrence, and nonsevere CDI; with similar treatment outcomes for both fdaxomicin and oral vancomycin in the treatment of severe cases [8]. In comparison to vancomycin, fdaxomicin demonstrates better sustained bactericidal rates [9]. However, oral vancomycin is often still prescribed over fdaxomicin as the latter is up to 14 times more expensive, and often not included within regional formularies for coverage [10]. Beyond bactericidal efcacy, treatment with fdaxomicin and/or vancomycin results in lower rates of CDI shedding and environmental contamination, when compared to metronidazole [11].
A community-based antimicrobial stewardship program has operated in the province of BC, since 2005. Te DoBugsNeedDrugs ™ program educates the public on judicious antibiotic use, while also disseminating clinical resources to healthcare professionals to promote the principles of antimicrobial stewardship and support patient care. Since the program's inception, provincial surveillance of antibiotic use has been ongoing. However, this is the frst study to review CA-CDI-indicated antibiotic use, and moreover, this study ofers the frst interpretation of fdaxomicin use in BC since its addition to the provincial formulary. Te objective of this study is to examine provincial trends in antibiotic dispensing for metronidazole, vancomycin, and fdaxomicin, from 2000 and 2018.

Data Sources.
Canadian citizens and permanent residents receive health coverage through provincial health insurance. As such, the Ministry of Health in BC maintains several health data indices, which contain comprehensive medical information for the population of BC. Community prescription data is contained within the BC PharmaNet, a centralized system linking outpatient pharmacies and dispensations [12]. All antimicrobials were recorded in this system except those used for the treatment of sexually transmitted infections and HIV. Medications dispensed within a hospital, or inpatient care setting, are also not available in the BC PharmaNet. Te medical services plan (MSP) billing system contains records for outpatient care covered by the provincial, universal insurance program [13]. All claims submitted by physicians are available, including diagnostic codes. Hospitalization data are available through the discharge abstracts database (DAD), recording all diagnostic information for provincial acute care hospitals [14]. Inpatient antibiotic data is not available through DAD. Patient demographics (e.g., age, sex) were provided through a consolidation fle [15]. Access to the data provided by the data stewards is subject to approval, but can be requested for research projects through the data stewards or their designated service providers. All inferences, opinions, and conclusions drawn in this publication are those of the authors and do not refect the opinions or policies of the data stewards.

Study Population.
Our study included all residents of BC, from January 1, 2000, to December 31, 2018, that were dispensed a prescription for a CDI-related antibiotic. Antibiotics were classifed based on the Anatomical Terapeutic Chemical (ATC) classifcation system developed by the World Health Organization. Antibiotic dispensations were extracted from PharmaNet, and the following three agents were included as follows: metronidazole (J01XD01), vancomycin (J01XA01), and fdaxomicin (A07AA12). Antimicrobials were limited to oral use only. Multiple prescriptions per subject were permitted in our analyses. All cells with n < 6 were excluded from subsequent analyses to preserve subject anonymity.

Primary Outcomes & Statistical Analyses.
Primary outcomes included the rates of antibiotic use. Consumption rates were calculated as prescriptions per 1000 population per year, using age-and gender-specifc denominator estimates, available for the population through statistics BC. Rates of antibiotic use were examined overall and then stratifed by the three clinically relevant, CA-CDI-related drugs (i.e., metronidazole, vancomycin, and fdaxomicin). For each medication, rates of use were further examined by sex and age category. Age categories included pediatrics (aged <19 years), adults (aged 19-64 years), and seniors (aged >65 years). Change over time was calculated for all rates, comparing the percentage change between the frst and fnal, study years.

Secondary Outcomes.
Although oral vancomycin and fdaxomicin are exclusively used for the treatment of CDI, metronidazole is used as an antiprotozoal agent for a variety of indications. Secondary outcomes examined the proportions of antibiotic use attributable to community and/or hospital physician records for CDI. CA-CDI cases were identifed using ICD-9 code "008" in MSP, while HA-CDI cases utilized ICD-10 code "A04.7" through DAD. A prescription and diagnosis were linked through an algorithm that tied the dispensation date to a physician visit within a 5day lookback period. Prescriptions that did not match a CDI physician record were categorized as "unlinked." Metronidazole has been included within the BC formulary since 2001, with vancomycin being covered since 2004, and fdaxomicin was added in 2014. However, both vancomycin and fdaxomicin receive coverage under "special authority" in BC, with requirements that must be met prior to patient eligibility, and reimbursement. In order to receive provincial coverage, a prescription of oral vancomycin requires the following: (1) a preceding course of metronidazole (excluding documented allergy/intolerance); (2) patient symptoms of moderate to severe disease; (3) patient experiencing a second disease recurrence; or (4) patient initiated on vancomycin as an inpatient. For fdaxomicin, either a documented allergy/intolerance to vancomycin; or an unsuccessful preceding course of oral vancomycin coupled with a high-risk hospitalization is required to achieve "special authority" coverage. Both oral vancomycin and fdaxomicin were examined to see what proportion of dispensations upheld the BC coverage criteria, respectively. A 14-day lookback period was utilized in linking prescriptions in order to allow a complete frst prescription course, secondary physician visit, and subsequent dispensation.

Results
An average of 61 809 unique patients were prescribed an antibiotic indicated for CDI in any given year, with over 1.5 million antibiotic dispensations between 2000 and 2018 ( Table 1). During this period, the population of BC increased by 24% to 5 001 170 in 2018. Regarding patient demographics, the mean age of the cohort increased from 46 to 50 years of age by 2018. Te average patients were adult females living in urban regions. No diferences were observed based on income quintiles.

Overall Rate of Prescribing for CDI-Indicated Antibiotics.
Overall, an average annual rate of 18.5 antibiotic dispensations per 1000 population was observed for all antibiotics. Te rate of prescribing increased (15%) over the 19-year study period, from 17.2 to 19.8 dispensations per 1000 population (Figure 1(a)).
By medication of interest, all three drugs increased in use over the study period, but by varying magnitude On average, the rate of prescribing per 1000 population for fdaxomicin was 0.009 from 2012 to 2018 (data were available later than the other two antibiotics), compared to 18 for metronidazole and 0.5 for vancomycin from 2000 to 2018. Additionally, the rate for fdaxomicin reached a peak of 0.014 prescriptions per 1000 population in 2016 before decreasing to 0.010 prescriptions per 1000 population by 2018.
When observing trends based on age categories, seniors were generally prescribed CDI-related antibiotics at a higher rate than children and adults (Figures 2(a)-2(d)). Tis trend was consistent among all three antibiotics and across the study period. Vancomycin, in particular, had the greatest increase in the rate of prescribing for seniors (528%) over adults (480%), and children (374%) (Figure 2(c)). Fidaxomicin prescribing followed the same trend, but the percentage increase was higher for adults (299%) compared to seniors (10%); no pediatric dispensations of fdaxomicin were identifed (Figure 2(d)). Metronidazole prescribing saw a decrease in pediatrics (20%) and seniors (2%), while adults increased by approximately 6% (Figure 2(b)). However, the general trend in prescribing remained consistent in that seniors were prescribed the antibiotic at a higher rate than the other two age categories.
Sex stratifcation showed a consistent pattern in prescribing (Figure 3(a)). Females were prescribed antibiotics at a greater rate compared to males for all three antibiotics. However, vancomycin and fdaxomicin prescribing rates for males and females showed a stronger positive association compared to the trend observed for metronidazole (Figures 3(b)-3(d)).

Sensitivity Analysis.
A total of 0.37% of metronidazole prescriptions, 6% of vancomycin, and 18% of fdaxomicin were successfully linked to an outpatient physician record for CA-CDI. Similarly, 0.18% of metronidazole prescriptions, 3% of vancomycin, and 5% of fdaxomicin were successfully linked to an inpatient record for HA-CDI (Table 2). Across the study period, 14% of vancomycin prescriptions were linked to prior use of metronidazole. In comparison, 34% of fdaxomicin prescriptions were linked to prior use of vancomycin.

Discussion
Tis is the frst study to evaluate outpatient prescribing for CDI-indicated antibiotics, and one of the few studies to evaluate fdaxomicin prescribing since its introduction to the Canadian marketplace. We observed a 15% increase in total provincial antibiotics over the 19-year study period. A positive overall trend was reported for each antibiotic but was most pronounced for oral vancomycin (621%), followed by fdaxomicin (226%), with metronidazole demonstrating the most modest trend (15%) over the 19-year study period. Although stagnant in use when compared to the other medications, metronidazole was dispensed at rates far and above vancomycin and/or fdaxomicin across all study years. Tese results are corroborated by previous provincial analyses, which report a 97% total increase in prescribing for the J01X ATC class of "other antimicrobials"; comprised of metronidazole, vancomycin, and more [16].
Updates to clinical practice guidelines over time also correspond to changes observed in prescribing rates for the three antibiotics. During the early 2000s, the frst-line antimicrobial therapy for CDI was metronidazole, which remained the guideline-recommended frst-line agent for CDI until 2012, with preferred use for initial mild episodes of CDI. At this time, vancomycin was reserved for use in severe episodes, due to the adequate efcacy of metronidazole for CDI and the risk of vancomycin-resistant enterococci acquisition in hospitals [17][18][19][20][21][22][23]. However, the study results show an exponential increase in vancomycin prescribing beginning in 2004, 6 years prior to the IDSA guideline updates in 2012 (Figure 1(c)). Prior to 2004, the rate of prescription for vancomycin was relatively stable (0.2 prescriptions per 1000 population). Tis increase may be attributable to a change in disease severity secondary to the emergence of a hypervirulent strain in Canada and the United States in 2003 [23][24][25][26]. Tis new strain, the North American pulsed-feld type 1 (NAP1/B1/027), is associated with greater production of toxins compared to control strains [23]. As vancomycin was recommended for the treatment of severe illness, the shift in bacterial strain may account for the increased physician preference during this period [21,27,28]. Vancomycin use increased further after the guideline change in 2012, which recommended it as the drug of choice for most cases of CDI, unless very mild and uncomplicated.
A growing body of literature supports the superiority of fdaxomicin, when compared to both metronidazole and oral vancomycin, for the treatment of CDI and for less patient sequelae [8,[29][30][31][32]. Tis study found that the frst outpatient prescriptions of fdaxomicin were dispensed in 2012, two years prior to formulary incorporation. Despite a steady upwards trend in use, prescribing rates for fdaxomicin remain low. A special authority status, elevated cost of treatment, and a judicious provincial stewardship culture could each infuence the ongoing reservation of fdaxomicin use in outpatient care. No pediatric use of fdaxomicin was identifed throughout the study period; however, this result was expected as the use of this medication remains limited to individuals aged ≥18 years [33]. Overall, the trends of antibiotic use reported in this study refect updates made to clinical practice guidelines throughout the study period. Te most recent IDSA guidelines became available in 2018, reiterating the position of vancomycin and fdaxomicin as frst-line agents, and correlating to the continued increases in the use of both drugs in later study years. Future research eforts should continue to monitor these CDI-indicated drugs in order to monitor guideline adherence. As previous antibiotic use remains a leading risk factor for CDI, particularly fuoroquinolone exposure, it is notable that provincial quinolone use declined by 22%, over the same study period [6,16]. However, provincial rates of CDI continue to trend upward, particularly within the community [3][4][5].
Te average patient was approximately 48 years of age (Table 1). Previous studies have identifed that CA-CDI patients are typically younger than those with HA-CDI, with the median age of CA-CDI patients being 50 years, in comparison to 72 years for HA-CDI [34]. In another study from the US, Chitnis et al. also reported a median age of 51 years for confrmed CA-CDI patients [35]. Te age-based trends reported for CDI-indicated antibiotics are supported by the literature surrounding patient cohorts and associated ages.
Te sensitivity analyses of vancomycin and fdaxomicin found that only 14% of vancomycin prescriptions were linked to a preceding metronidazole dispensation, while 34% of fdaxomicin prescriptions were linked to a prior vancomycin dispensation. Tese percentages were lower than expected, especially given the criteria necessary to achieve "special authority" status include an inefective but adequate preceding course of treatment for each medication. With respect to vancomycin, a lower linkage may have been achieved as some patients may not have been prescribed metronidazole within the permissible 14-day lookback period. Other conditions for oral vancomycin under "special authority" include moderate to severe CDI; recurrence, or continuation of therapy following inpatient care [36]. As these additional criteria do not specify the prior dispensation of metronidazole, the above scenarios are likely hypotheses underlying "unlinked" oral vancomycin use. Similarly, for fdaxomicin (34% linkage), further "special authority" criteria dictate its use if it is prescribed by an infectious disease physician or gastroenterologist; documented allergy/intolerance for vancomycin, or severe illness and risk of hospitalization [37]. Moreover, there may be some scenarios in which a 14-day lookback window is insufcient to best provide an accurate linkage of fdaxomicin to a prior vancomycin dispensation. Tis study ofers the frst highlevel interpretations of preceding medication use for CDIrelated "special authority" antibiotics. However, further research is required to properly elucidate the provincial "special authority" criteria, subsequent adherence, and proportions attributable to each of the aforementioned criteria.
Overall, only 0.53% of the total antibiotics were linked back to an outpatient CDI-related physician record, with 0.27% linked to an inpatient record. Fidaxomicin reported the highest proportions of linkage across both healthcare (a) shows the rate of total antibiotic prescribing for each sex; (b-d) show the rate of rate prescribing for metronidazole, vancomycin, and fdaxomicin, respectively, for each sex. Table 2: Percentage of CDI-indicated antibiotics linked to the outpatient (ICD-9 diagnostic code "008") database and the discharge (hospital) database (ICD-10 diagnostic code "A04.7") throughout the entire study period, 2000-18.
Type of antibiotic Count linked to "008" Percentage linked (%) Count linked to "A04. 7 (Table 2). Given the wide-ranging indications for metronidazole use (e.g., reproductive system, skin and soft tissue, and sexually transmitted infections), the rates reported are not surprising. Te iteration of clinical practice guidelines which decentred metronidazole as a frst-line agent and the introduction of superior treatment options for CDI are also potential explanations. However, the low linkage of vancomycin and/or fdaxomicin, which are primarily indicated for CDI, highlights the need for further research into our provincial administrative billing codes for accuracy. Moreover, it is integral to understand which billing codes are being utilized for CA-CDI cases in order to identify cases of infection within our provincial data. As a retrospective study using administrative health data, there are limitations to consider. Unflled prescriptions would not be identifed within dispensation records, as such our discussion of antibiotic prescribing may be an underrepresentation of provincial use. Moreover, the absence of accurate culture, lab, and allergy data limited our characterization of "special authority" use to preceding antibiotic dispensation records only. Furthermore, our use of physician billing codes (ICD-9/10) may be subject to misclassifcation bias, as CDI case identifcation was reliant on accurate coding by physicians. It is notable that Canadian primary care physicians' claim data have a high positive predictive value for the diagnosis of common infections [38]. However, this may not be a valid extrapolation to CDI. Finally, annual cases of CDI were not identifed within this study, as accurate case identifcation of CA-CDI, is not yet been validated using outpatient administrative health data. Antibiotics indicated for the treatment of CA-CDI were reported and examined for changes over time relative to CDI clinical practice guidelines.

Conclusion
CDI cases in outpatient care continue to rise, as the infection has expanded beyond nosocomial and into community circulation. In tandem, since 2000, this study reports an overall increase in the outpatient dispensation of CDIindicated antibiotics. Although causation cannot be inferred from study results, oral vancomycin and fdaxomicin use has increased over the study period, in line with, or in advance of guideline changes made throughout the 19-year study period. Furthermore, research is necessary to elucidate adherence to provincial "special authority" criteria in relation to both medications. Moreover, antibiotics dispensed in the community are typically issued to a younger cohort than those reported in HA-CDI cases, highlighting the importance of reframing CDI, and associated antibiotic use as a widespread outpatient concern rather than a nosocomial risk for seniors.

Data Availability
Te underlying data for this research is protected by the BC Ministry of Health and cannot be made publicly available.

Conflicts of Interest
Te authors declare that they have no conficts of interest regarding the publication of this article.