Introduction

Community-acquired pneumonia (CAP) is a leading cause of mortality and morbidity for older adults. The most commonly identified pathogens include Streptococcus pneumoniae, as well as respiratory viruses [1, 2]. Despite microbiologic testing, no pathogen is detected in up to 62% of inpatient cases [1, 3, 4]. Deteriorations in functional status following illness occurs in 10% of patients post-recovery, jeopardizing quality of life [5]. Moreover, despite the polysaccharide pneumococcal vaccine being available in Canada through publicly-funded immunization programs since the 1990s, annual rates of CAP have continued to rise, with 212 additional incident cases per 100,000 in the province of British Columbia (BC) by 2018 (compared to 2010), and 239 in Ontario (ON) [6]. Highest infection rates continue to be reported in the elderly populations [7]. Adherence to clinical guidelines in antibiotic treatment selection has been associated with improved patient outcomes, and reduced healthcare costs [8–10]. Given the elevated risk of mortality in this population, understanding appropriate antibiotic use is essential to protect this vulnerable population.

The etiology of CAP is not homogenous [6]. However, unlike the case of viral respiratory tract infections, delays in antibiotic treatment for CAP increase the risks for adverse outcomes [9, 11, 12]. As such, the goals of antimicrobial stewardship for CAP include the optimization, rather than reduction, of empiric antibiotic prescribing. The evaluation of antibiotic prescribing is often contingent on patient factors not available within routine, administrative health data, as such, appropriate outpatient use has been examined in limited scope [13]. Efforts to evaluate prescribing quality are further limited by the absence of standardized nomenclature [14]. In 2018, a Canadian antimicrobial stewardship program published categories of antibiotic prescribing aimed to address aforementioned limitations; provide more clinically relevant and granular interpretations of prescribing quality; and offer a nuanced lexicon with which to engage [15].

This retrospective cohort study is the first in North America to examine if antibiotic prescribing was appropriate for CAP in older adults, by agent, dose and duration. The use of clinical practice guidelines to define discrete categories of prescribing quality offer a novel, more objective interpretation of appropriate prescribing, when compared to previous efforts of expert opinion elicitation. Our objectives were to determine the quality of empiric antibiotic prescribing for CAP, across two Canadian provinces, and expand interpretations of appropriate prescribing by agent, dose, and duration.

Methods and materials

Results

Over the study period 118,606 and 317,823 total episodes of CAP were identified in BC and Ontario, respectively (Table 1). Of these episodes, 46% were prescribed an antibiotic in BC, and 52% in Ontario, resulting in 219,480 total episodes of CAP issued antibiotics (Fig 1). The average age for BC patients was 78 years and 77 years in Ontario, with slightly more females for both provinces (BC: 53.3%; ON: 55.1%). In both provinces, the overall rate of CAP episodes was 28 cases per 1000 population. The average number of physician visits per episode was 1.75 and 1.35, per province respectively. These episodes can be attributed to 51,981 unique patients in BC, and 138,157 in Ontario.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Counts of CAP episodes, prescribed episodes and unique patients.

https://doi.org/10.1371/journal.pone.0292899.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Cohort characteristics.

https://doi.org/10.1371/journal.pone.0292899.t001

Category of prescribing quality

Clinically appropriate prescribing accounted for the most antibiotics issued to older adults, for CAP, across all study years (BC: 60.7%; ON: 74.2%) (Fig 2). These prescriptions were mainly composed of non-first line agents dispensed in the presence of a clinical factor that could preclude the use of a guideline recommended agent (BC: 96.4%; ON: 97.9%). A complete list of these factors and their relevant cohort counts are available in Table 1. The remaining clinically appropriate prescriptions (BC: 3.6%; ON: 2.1%) can be attributed to the use of first-line agents, in the presence of chronic kidney disease, re-classified from undertreatment. In both provinces, rates of clinically appropriate prescribing significantly decreased over the study period (Table 2; Fig 3). By 2018, a 20% (RR: 0.80; 95% CI: 0.77–0.83) reduction in BC, and 15% (Rate Ratio [RR]: 0.85; 95% Confidence Interval [CI]: 0.84–0.87) reduction in Ontario, was observed in clinically appropriate prescribing, when compared to the first study year (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Percentage of outpatient antibiotic prescribing for CAP, by category of prescription quality and year.

https://doi.org/10.1371/journal.pone.0292899.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Rate of overall antibiotic prescribing for outpatient CAP in Canadian seniors, by age category and province.

https://doi.org/10.1371/journal.pone.0292899.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Patterns of antibiotic prescribing by category of quality for outpatient CAP in Canadian seniors.

https://doi.org/10.1371/journal.pone.0292899.t002

Guideline adherent reported the smallest proportions of antibiotic prescription in each study year, for both provinces (BC: 1.69%; ON: 0.36%) (Fig 2). These prescriptions were fully concordant by agent, dose and duration. Despite a two-fold and four-fold increase by 2018, in BC and ON respectively, guideline adherent prescribing remained the least prescribed category of antibiotic use (BC: 6.93 prescriptions per 1000 population; ON: 1.86 prescriptions per 1000 population) (Table 2; Fig 3).

Effective but unnecessary prescriptions, which were adherent by agent but excessive by dose and/or duration were the second-most identified category; accounting for 19.9% of all prescribing in BC, and 10.8% in Ontario (Fig 2). Although both provinces demonstrated increasing rates over time, “effective but unnecessary” prescribing was roughly double in BC, across all study years (Table 2; Fig 3). By 2018, when compared to the first study year, Ontario (RR: 2.94; 95% CI: 2.78–3.10) experienced a greater increase than BC (RR: 1.79; 95% CI: 1.68–1.91) in this category. Excess duration of therapy was the hallmark characteristic for effective but unnecessary prescribing categorization (BC: 92%; ON: 99%). The remaining effective but unnecessary prescriptions (BC: 8%; ON: 1%) were also excessive in treatment length, in addition to a dose beyond the average daily recommendation.

Undertreatment prescriptions were also adherent in agent, but subtherapeutic by dose and/or duration. This category followed as third most identified in both provinces, and as most comparable across provinces for magnitude of overall antibiotic utilization (BC: 10.98%; ON: 10.83%) (Fig 2). In Ontario, “under treatment” prescribing increased year-on-year with 35 prescriptions issued per 1000 population, by 2018. Contrastingly, BC “under treatment” antibiotic use peaked in 2016, before trending down in later study years with 35 prescriptions issued per 1000 population by 2018 (Table 2). Across both provinces, the majority of “under treatment” prescriptions (ON: 94%; BC: 79%) can be attributed to subtherapeutic dosing in the presence of excessive length of therapy. In BC, much of the remaining prescriptions are tied to either short duration of therapy (9%), subtherapeutic dosing (5%), or both (6%); while the residuals in Ontario are split evenly (2%) across the same categories.

Not recommended prescribing was the only category of prescribing to remain stable over time, in both provinces (BC—RR: 1.05; 95% CI: 0.95–1.17; ON—RR: 0.93; 95% CI: 0.86–1.01). These prescriptions were composed of non-guideline adherent antibiotic agents, in the absence of all clinical factors that would have precluded their use. Overall, “not recommended” antibiotic use was higher in BC (7%) than Ontario (4%), and this was true across each study year as well.

Discussion

This is the first study to evaluate appropriate prescribing for outpatient CAP, by agent, dose and duration. Our study found that “guideline adherent” prescribing, which are concordant with first line agents and recommended dose and duration, accounted for only 2% of prescriptions in BC, and less than 1% in Ontario. However, the majority of prescribing in both provinces was deemed “clinically appropriate”, wherein comorbidities justified the use of non-first line agents (BC: 61%; ON: 74%). The breakdown of appropriate prescribing between guideline adherent and clinically appropriate may be attributable to the study population—as older adults may present more routinely with comorbidities, requiring deviations from guideline adherent therapies. Not recommended prescribing was minimal in both provinces, with 4% of all prescriptions in Ontario, and 7% in BC, discordant with guidelines in the absence of any patient clinical justifications. Clinical practice guidelines, and literature preceding our study period provide evidence that shorter treatments are non-inferior for CAP [8–10, 23–25]. Regardless, duration was the leading determinant for effective but unnecessary prescription, with excessive lengths of therapy (≥7 days).

Given the highest incidence rates for CAP are amongst older adults (≥65 years), and the positive association between increasing age and risk, ensuring judicious use of antibiotics for this population is paramount. Multiple studies have also identified associations between increased in-hospital mortality and therapies used for a preceding episode of community acquired pneumonia [33, 34]. Chakrabarti et al. report that pre-treatment with antibiotics for CAP remain an independent risk factor for increased in-hospital mortality (Odds Ratio [OR]: 1.43, 95% CI: 1.19–1.71) after adjusting for age, and various comorbidities. This finding however may simply reflect a lack of response to previous empiric oral therapy being a poor prognostic factor. Johnson et al. found that 75% of antibiotics issued in the community, prior to a subsequent hospitalization for pneumonia, were appropriate, with 98% of these prescriptions reporting adequate dose, but with no additional analysis of duration of therapy. Appropriate prescribing prior to hospitalization conferred a 38% relative mortality reduction for patients. With respect to these data, our study found high rates of appropriate prescribing for CAP, which is favorable for Canadian older adults, in the event of subsequent hospitalization.

First line agents for this study included: amoxicillin, amoxicillin clavulanate and doxycycline. In BC, doxycycline accounted for a much larger proportion of prescriptions issued (43%), when compared to Ontario (4%). Previous work by this team had identified a provincial difference in tetracycline prescription, with BC using six times more than Ontario [35]. A potential explanation for this inter-provincial difference may be the shift to tetracyclines as a preferred agent in BC clinical resources, while beta-lactams remain primary agents in Ontario. With respect to non-first line agents, both the categories of clinically appropriate, and not recommended were examined by major ATC class. Macrolides and quinolones were the most selected classes of antibiotics for both categories. Macrolides were the most prescribed class of antibiotic in the absence of clinical justifications (i.e., not recommended antibiotic use). In BC, 48% of all not recommended prescribing were macrolides, with 62% in Ontario. Although macrolides are recommended alternatives in the presence of patient comorbidities—our study identified increased prescription in the absence of a host of patient factors, as listed in Table 1. Moreover, the use of beta-lactam monotherapy has been determined to be non-inferior to beta-lactam macrolide combination treatment [36]. Positively, the prescribing of quinolones and macrolides for outpatient treatment of community acquired pneumonia, in the absence of patient clinical justifications, accounts for a minority of overall prescribing in both provinces (BC: 6%; ON: 4%). It is possible that not recommended prescribing reflect antibiotic prescription in individuals with penicillin allergy. Although allergy data were unavailable for this study, reported population rates for beta-lactams are approximately 1 to 5% and may explain the use of non-first line agents in the absence of other examined comorbidities [37, 38].

This study has limitations inherent to all retrospective studies using administrative health data. Foremost, the absence of lab data to confirm the presence of infections, and case identification based on physician billing and ICD-9 codes may be subject to misclassification bias. However, Canadian physician claims data have reported high positive-predictive values for the diagnosis of common respiratory infections [39]. Moreover, codes utilized to identify CAP did not eliminate non-bacterial pneumonia. In BC, a unique diagnostic code is available for viral pneumonia (ICD-9: 480) however, available OHIP codes do not distinguish on etiology. As such, in order to ensure internal validity, all available pneumonia codes were included across both provinces and our reduced proportion of episodes prescribed (BC: 46%, ON: 52%) is likely attributable to the inclusion of cases that do not warrant antibiotic therapy. As only CAP-related codes were utilized in defining episodes of infection, and did not account for the potential of concurrent diagnoses, which may impact patient and prescribing misclassification. The absence of laboratory results led to a blanket assumption of non-atypical infection. The use of chronic kidney disease as a proxy for unknown creatinine clearance could also have underestimated our proportion of clinically appropriate prescribing, reclassified from under treatment. In a validation of Canadian outpatient data to determine chronic kidney disease, Fleet et al. identified only 33% sensitivity in older adults [40]. As such, it may be that the reported 4% (BC) and 2% (ON) attributed to the use of first-line agents, in the presence of chronic kidney disease re-classified from undertreatment, underestimates the true proportions. Allergy data was also unavailable and limited our characterization of non-first-line antibiotic prescribing. Current estimates report a range of 1–5% for beta-lactam allergies, and the absence of this data may have overestimated not recommended antibiotic prescriptions [37, 38]. With respect to antibiotic prescribing, unfilled prescriptions are not included within records, and our discussion of antibiotic use may be an underrepresentation of true provincial magnitude. Furthermore, prescription parameters of dose and duration were pulled from dispensation records, with no knowledge of patient adherence and/or presence of physician instructions (e.g. wait and watch stewardship efforts). Future research efforts should also engage with laboratory data to identify cases of atypical infection to elucidate elevated rates of macrolide and quinolone use.

Conclusion

Community acquired pneumonia in older adults confers increasing risk for mortality, or functional status. Understanding the quality of antibiotic prescribing in this population is integral to maximize positive patient outcomes, while minimizing antibiotic-related adverse events, and the impact of prescribing on rates of bacterial resistance. This is the first study to review the quality of prescriptions issued for CAP, by agent, dose and duration, with the majority of antibiotic prescription clinically appropriate. However, the continued use of beta-lactam penicillins in place of macrolides or quinolones, and encouraging 3 to 5-day shorter course treatments present a clear direction for future stewardship efforts.

Supporting information

Acknowledgments

BC data were extracted, anonymized, and made available to researchers by Population Data BC Access to data provided by the Data Steward(s) is subject to approval, but can be requested for research projects through the Data Steward(s) or their designated service providers All inferences, opinions, and conclusions drawn in this publication are those of the author(s), and do not reflect the opinions or policies of the Data Steward(s).

The study was completed at the ICES Western site. This document used data adapted from the Statistics Canada Postal Code^OM Conversion File, which is based on data licensed from Canada Post Corporation, and/or data adapted from the Ontario Ministry of Health Postal Code Conversion File, which contains data copied under license from ©Canada Post Corporation and Statistics Canada. Parts of this material are based on data and/or information compiled and provided by Ontario Ministry of Health, MLTC and CIHI. However, the analyses, conclusions, opinions and statements expressed in the material are those of the author(s), and not necessarily those of MOH, MLTC and CIHI. We thank IQVIA Solutions Canada Inc. for the use of their Drug Information File.