Combined impact of the COVID-19 pandemic and the number of comorbidities on community antibiotic utilisation rates: a population-based retrospective cohort study using linked health administrative data in Quebec, Canada

Abstract

Objectives

To describe community antibiotic utilisation in Quebec from 2018 to 2022 and to measure the combined impact of the COVID-19 pandemic and of the number of comorbidities on utilisation rates.

Methods

Data from the Quebec Integrated Chronic Disease Surveillance System were used to describe monthly (for the overall antibiotics use) and annual (for the main antibiotic classes) changes in antibiotic utilisation rates from 2018 to 2022, stratified by the number of comorbidities (0, 1, 2 and ≥3) and age group (0–17, 18–64 and ≥65 years old). Poisson regression was used to measure the impact of the pandemic and of comorbidities on antibiotic utilisation rates.

Results

The study included an annual average of 424 792 children, 1 761 582 adults 18–64 years of age and 1 490 081 adults at least 65 years old. For each number of comorbidities within each age group, the utilisation rates of overall antibiotics decreased with the arrival of the pandemic and remained low despite the return of respiratory viruses in late summer 2021. This reduction was observed for all major antibiotic classes, except for fosfomycin/nitrofurantoin in adults. The pandemic and respiratory viruses' resurgence periods in adults (≥18 years) without comorbidities were associated with decreases of 25% (95% CI 25% to 25%) and 19% (18% to 19%) (children: 63% (62% to 63%) and 37% (36% to 38%)) in antibiotic utilisation compared with the prepandemic period. In adults with three or more comorbidities, utilisation decreased less, by 13% (12% to 14%) and 7% (6% to 8%) (children: 33% (21% to 43%) and 23% (8% to 35%)), respectively. Children with two comorbidities during the pandemic period also experienced a smaller decrease in antibiotic utilisation than children without comorbidities for the same period.

Conclusion

In Quebec, antibiotic utilisation decreased during the pandemic and remained low despite the resurgence of respiratory viruses in 2021. However, this decrease was less pronounced in individuals with multiple comorbidities.

Strengths and limitations of this study.

• This study was conducted with a large and comprehensive population-based sample.

• A previous study had already demonstrated the possibility of using Quebec Integrated Chronic Disease Surveillance System (QICDSS) data to perform the surveillance of community antibiotic utilisation.

• Previous studies have demonstrated the high quality and validity of the health administrative data used and algorithms employed to identify comorbidities.

• Misclassification was possible if individuals had undiagnosed comorbidities.

• The stage of comorbidities was not documented in the QICDSS.

Introduction

The rapid emergence and spread of antibiotic resistance threaten the effectiveness of antibiotics and health.¹,³ Antibiotic utilisation needs to be documented, as it is the main factor accelerating the emergence of resistance, through the selection of resistant bacterial strains.^{12 4},⁶ Surveillance of antibiotic utilisation is particularly important among people with comorbidities who are more exposed to antibiotics,⁷,⁹ and consequently, to an increased risk of resistance to antibiotics, potentially leading to severe health complications, while their health is already more fragile compared with people without comorbidity.⁸¹⁰,¹⁷ Therefore, devoting special efforts to documenting antibiotic utilisation in this population is essential, as it represents a target for potential interventions that could benefit the entire population by curbing antibiotic resistance.

Several countries (Canada,¹⁸,²⁰ USA,²¹,²⁴ UK²⁵,³¹ and Australia),³²,³⁴ European countries³⁵,³⁷ and other Canadian provinces (British Columbia,³⁸ Ontario)³⁹ have already reported a decrease in community antibiotic utilisation with the arrival of the COVID-19 pandemic. However, to our knowledge, only three studies have addressed in some way whether the impact of the COVID-19 pandemic was similar regardless of the number of comorbidities on community antibiotic utilisation.⁴⁰,⁴² Among these, one relies on self-reported data from a developing country where antibiotics can still be easily obtained without a prescription⁴⁰; another lacks multivariate analyses⁴¹ and the last one uses ecological rather than individual data to estimate the number of comorbidities.⁴²

Thus, our objectives were to describe community antibiotic utilisation rates in Quebec, Canada, from 2018 to 2022, according to the number of comorbidities, and to measure the combined impact of the COVID-19 pandemic and the number of comorbidities on antibiotic utilisation rates. We also estimated the quality of antibiotic utilisation during this period, by using the Access, Watch and Reserve (AWaRe) classification developed by the WHO.⁴³ WHO recommends that at least 60% of antibiotic consumption come from the ‘Access’ group of antibiotics to preserve the efficacy of the other classes.⁴³

Methods

Study design and population

This population-based retrospective study included all individuals covered by the Quebec public drug insurance plan for at least 1 day, between 1 April 2018 and 31 March 2022. While coverage is automatic for people 65 years and older, it is limited for people without a private insurance plan for those under 65 years of age, such as self-employed individuals or those on social assistance.⁴⁴ For the entire Quebec population in 2021, covered individuals represented approximately 28% of children, 38% of adults 18–64 years and 92% of adults 65 years or older.⁴⁵ We excluded residents in long-term care facilities and residents of Nunavik and Terres-Cries-de-la-Baie-James since the public drug insurance plan does not directly cover them.

Data sources and variables

Data were obtained from the Quebec Integrated Chronic Disease Surveillance System (QICDSS), which uses an anonymised unique identifier to link information from five health administrative databases: (1) health insurance; (2) hospital admission and discharge; (3) physician claims; (4) pharmaceutical services (the source for all data related to antibiotic prescriptions) and (5) vital statistics death.⁴⁶ The QICDSS is already well documented elsewhere.⁴⁶

Antibiotic utilisation was determined with all antibiotic prescriptions for systemic use as listed in class J01 of the Anatomic Therapeutic Chemical classification system⁴⁷ and metronidazole, an antibiotic belonging to class P01 also used against bacterial infection. Antibiotic utilisation was measured in prescriptions per 1000 individual-days for each age group and, for oral prescriptions in adults only (≥18 years), in defined daily doses (DDDs) per 1000 individual-days (DIDs). Data from the WHO Collaborating Centre for Drug Statistics Methodology website and the Drug Product Database online query website were used to assign a DDD to each prescription and to identify antibiotic prescriptions, antibiotic classes and total quantities dispensed.^{48 49} As standard DDDs are based on average doses recommended for a 70 kg adult,⁴⁸ DDDs could not be calculated in children. DDDs for parenteral and inhalation prescriptions could not be calculated either due to the lack of information on quantities dispensed (3.0% of total prescriptions among adults in our database, unpublished analyses, 2024). We categorised each prescription using the ‘WHO AWaRe classification of antibiotics for evaluation and monitoring of use, 2021’ available on the WHO website, which organises antibiotics into three categories based on their importance to medicine and their impact on antibiotic resistance, from the low-risk ‘Access’ to the last-resort ‘Reserve’ category.⁴³

To analyse the impact of the pandemic, we divided our data into three periods. The COVID-19 timeline published on the Institut National de Santé Publique du Québec (INSPQ) website shows that the state of public health emergency was declared on 13 March 2020, and the first lockdown in Quebec started on 24 March 2020.⁵⁰ Due to the resurgence of respiratory viruses at the end of the study period, this period was treated separately. Data from the provincial surveillance of respiratory viruses showed that the percentage of respiratory syncytial virus (RSV) positivity rose above 3% (the positivity threshold used by the Centres for Disease Control and Prevention to declare the start of an RSV season)⁵¹ in mid-July 2021.⁵² Based on this information, we defined the period between April 2018 and March 2020 as the prepandemic period, the period between April 2020 and July 2021 as the pandemic period and the period between August 2021 and March 2022 as the respiratory viruses' resurgence period.

The influence of comorbidities on antibiotic utilisation was evaluated according to the number of comorbidities per individual, which was categorised into four groups (0, 1, 2 and ≥3). The list of 31 comorbidities used in this study came from the combined comorbidity index previously validated by Simard et al.⁵³ Validated algorithms based on the ninth and tenth editions of the International Classification of Diseases diagnostic codes and developed specifically for the QICDSS were used to identify comorbidities.^{10 46 53} For a comorbidity to be considered present, at least one diagnostic code had to be recorded in the hospital admission and discharge database, or a minimum of two physician’s claims with the same diagnostic code had to be registered within 365 days, with a gap of at least 30 days between them.¹⁰ We searched for comorbidities diagnosed in each year of the study period based on observations over the previous 5 years.

From the health insurance registry, demographic information was collected, such as sex, age (as of 1 October of each year), area of residence and quintiles of social and material deprivation indices (based on validated algorithms developed by the INSPQ⁵⁴). We excluded individuals with incomplete demographic data (less than 0.4% of individuals, unpublished analyses, 2024). Area of residence and quintiles of social and material deprivation were unknown for 0.4% and 7% of individuals, respectively (unpublished analyses, 2024); these individuals were categorised separately.

Statistical analysis

We described the monthly evolution in antibiotic utilisation rates from April 2018 to March 2022, stratified by the number of comorbidities and by age group (0–17 years, 18–64 years and ≥65 years). Moving averages were used to plot the expected trends of antibiotic utilisation rates during the pandemic and respiratory viruses’ resurgence periods (eg, the average rates for the two prepandemic months of April). We used heatmaps to represent the annual evolution of antibiotic utilisation rates of the five most prescribed antibiotic classes, stratified by age group and number of comorbidities. Similar heatmaps were also produced for azithromycin, given its controversial use in the early pandemic.⁵⁵,⁶⁰ We calculated frequency distributions of prescriber type for fosfomycin and nitrofurantoin prescriptions in younger and older adults, stratified by year and number of comorbidities, to evaluate changes in the prescriber type for these antibiotics specific to urinary tract infections (UTIs) during the pandemic. We computed frequency distributions of prescriptions according to the AWaRe classification, stratified by age group, period and number of comorbidities (in this analysis, the pandemic period was from April 2020 to March 2022 due to the small number of antibiotic prescriptions in the ‘Reserve’ category).

Two Poisson regression methods were used to measure the combined impact of the COVID-19 pandemic and the number of comorbidities on community antibiotic utilisation. For the first method, we performed regression models with indicator variables combining periods and number of comorbidities to obtain the rate ratios of each possible subgroup of the combination (Figure 3). For the second method, we performed regression analyses that included periods, number of comorbidities and interaction terms of these two variables to obtain adjusted rate ratios for each interaction term (table 1). Moreover, this approach enabled us to determine the adjusted rate ratios for each period within strata of the number of comorbidities, which we illustrated using heatmaps (Figure 4). All regression analyses were adjusted for month, sex, age group, area of residence and material and social deprivation indices. These models were also stratified into only two age groups (children and adults) as we were able to combine all adults (≥18 years) into the same models through age group adjustment. Generalised estimating equations were used due to the presence of individuals who provided multiple observations over the years.

Table 1. Multivariate analysis of the impact of the COVID-19 pandemic on antibiotic utilisation rates according to the number of comorbidities, April 2018–March 2022.

Variable	Children (rate of prescriptions per 1000 individual-days)		Adults 18 years or older (rate of prescriptions per 1000 individual-days)		Adults 18 years or older (DID)
	Crude rate ratio (95% CI)	Adjusted rate ratio* (95% CI)	Crude rate ratio (95% CI)	Adjusted rate ratio* (95% CI)	Crude rate ratio (95% CI)	Adjusted rate ratio* (95% CI)
Period (in the 0 comorbidity group)†
Prepandemic period (April 2018–March 2020)	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Pandemic period (April 2020–July 2021)	0.35 (0.35 to 0.36)	0.37 (0.37 to 0.38)	0.74 (0.74 to 0.75)	0.75 (0.75 to 0.75)	0.69 (0.68 to 0.69)	0.70 (0.69 to 0.70)
Respiratory viruses resurgence period (August 2021–March 2022)	0.61 (0.60 to 0.61)	0.63 (0.62 to 0.64)	0.84 (0.83 to 0.84)	0.81 (0.81 to 0.82)	0.76 (0.75 to 0.77)	0.75 (0.74 to 0.75)
Number of comorbidities (in the prepandemic period group)†
0	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
1	1.70 (1.67 to 1.74)	1.75 (1.71 to 1.78)	1.33 (1.32 to 1.34)	1.29 (1.29 to 1.30)	1.55 (1.54 to 1.57)	1.58 (1.56 to 1.59)
2	2.46 (2.34 to 2.58)	2.43 (2.31 to 2.55)	1.71 (1.70 to 1.73)	1.65 (1.63 to 1.66)	2.09 (2.07 to 2.12)	2.16 (2.14 to 2.19)
≥ 3	3.66 (3.31 to 4.05)	3.62 (3.26 to 4.00)	2.62 (2.60 to 2.65)	2.48 (2.45 to 2.50)	3.48 (3.45 to 3.51)	3.73 (3.69 to 3.77)
Interaction terms (number of comorbidities*period)†
0 comorbidity or prepandemic period	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
1 comorbidity * pandemic period	1.00 (0.95 to 1.05)	1.04 (0.99 to 1.09)	0.99 (0.98 to 1.00)	1.00 (0.99 to 1.01)	1.01 (0.99 to 1.02)	1.00 (0.99 to 1.02)
2 comorbidities * pandemic period	1.27 (1.15 to 1.40)	1.33 (1.21 to 1.46)	1.02 (1.01 to 1.03)	1.03 (1.02 to 1.04)	1.06 (1.05 to 1.08)	1.06 (1.05 to 1.08)
≥ 3 comorbidities * pandemic period	1.76 (1.50 to 2.06)	1.80 (1.54 to 2.11)	1.16 (1.14 to 1.17)	1.17 (1.15 to 1.18)	1.20 (1.18 to 1.21)	1.19 (1.18 to 1.21)
1 comorbidity * respiratory viruses resurgence period	0.88 (0.84 to 0.92)	0.93 (0.88 to 0.97)	0.99 (0.98 to 1.00)	1.00 (0.99 to 1.01)	1.03 (1.01 to 1.04)	1.03 (1.01 to 1.04)
2 comorbidities * respiratory viruses resurgence period	1.01 (0.91 to 1.13)	1.07 (0.97 to 1.19)	1.01 (0.99 to 1.02)	1.02 (1.00 to 1.03)	1.06 (1.04 to 1.08)	1.06 (1.04 to 1.08)
≥ 3 comorbidities * respiratory viruses resurgence period	1.16 (0.97 to 1.40)	1.22 (1.03 to 1.46)	1.13 (1.12 to 1.15)	1.14 (1.13 to 1.16)	1.16 (1.14 to 1.17)	1.15 (1.14 to 1.17)

^*The adjusted models for children and adults are adjusted for month, sex, age group, area of residence and material and social deprivation indices. The complete table with rate ratios for the adjustment variables is available in online supplemental table S4.

^†The unadjusted models included periods, comorbidities and interaction terms between periods and comorbidities (allowing the separate impacts of periods, comorbidities and interaction terms to be isolated).

DDD, defined daily dose; DID, DDDs per 1000 individual-days; Ref., reference category.

All heatmaps were generated using Microsoft Excel’s three-level colour scale, which provides a colour gradient to represent data based on their positioning relative to the 10th, 50th and 90th percentiles.

Patient and public involvement

Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

Results

Antibiotic utilisation according to the number of comorbidities (2018–2022)

During the study period, 541 795 prescriptions were dispensed to 620 621 332 individual-days among children (online supplemental table S1). In adults 18–64 years of age, 3 173 031 prescriptions and 25 191 271 DDD were dispensed to 2 573 670 591 individual-days (online supplemental table S2). For older adults, these numbers were 4 307 441 prescriptions, 29 731 878 DDD and 2 177 008 922 individual-days (online supplemental table S3). Sociodemographic data for each age group according to the number of comorbidities are presented in online supplemental appendix 1.

During the prepandemic period, antibiotic utilisation rates reached their highest levels for each number of comorbidities within each age group, with an average of 1.2, 1.4 and 2.2 prescriptions per 1000 individual-days for children, adults 18–64 years and older adults, respectively (figure 1; online supplemental tables S1–S3). Between March and April 2020, rates sharply decreased by an average of 62%, 24% and 23% in the respective age groups (figure 1). During the pandemic period, antibiotic utilisation rates reached their lowest levels, with an average of 0.5, 1.0 and 1.7 prescriptions per 1000 individual-days in children, younger adults and older adults, respectively. With the resurgence of respiratory viruses in the summer of 2021, antibiotic utilisation rates increased but remained lower than before the pandemic with an average of 0.8, 1.1 and 1.8 prescriptions per 1000 individual-days in the respective age groups. These trends were observed across all strata of the number of comorbidities. Antibiotic utilisation rates increased with the number of comorbidities and with age groups (figure 1). The trends were the same for utilisation rates in DID, in adults (online supplemental figure S1).

Figure 1. Evolution of antibiotic utilisation rates among (A) children, (B) adults 18–64 years of age and (C) adults 65 years or older between April 2018 and March 2022, according to the number of comorbidities (prescriptions per 1000 individual-days). Plain lines represent the observed rates, while dotted lines represent the expected rates based on the average of the respective monthly rates within the prepandemic period. The vertical dashed lines delineate the three study periods, as indicated in the timeline.

In the five most prescribed antibiotic classes, utilisation rates also increased with the number of comorbidities (figure 2). The reduction in antibiotic utilisation rates during the pandemic period was also observed across all major antibiotic classes, except for the cephalosporins and tetracyclines classes in children and the fosfomycin/nitrofurantoin class in younger and older adults (figure 2; online supplemental figure S2 for rates in DID). It is important to note that the cephalosporins and tetracyclines classes represented a small percentage (6% and 5%, unpublished analyses, 2024) of the total antibiotic prescriptions in children, which was insufficient to confirm a decrease. Within the stable utilisation of fosfomycin/nitrofurantoin among adults, the proportion of prescriptions issued by physicians decreased, to the benefit of those dispensed by pharmacists (online supplemental figure S3). This trend was particularly pronounced among adults aged 18–64 years and those without comorbidities. Azithromycin utilisation rates also decreased between 2019 and 2020, for each number of comorbidities within each age group (online supplemental figure S4).

Figure 2. Evolution of utilisation rates of the five most prescribed antibiotic classes among (A) children, (B) adults 18–64 years of age and (C) adults 65 years or older between April 2018 and March 2022, by year and number of comorbidities (prescriptions per 1000 individual-days).¹ Years begin on 1 April and end on 31 March the following year.

Overall, the proportions of antibiotic utilisation classified under the ‘Access’ category of the AWaRe classification respected the WHO recommendations for each age group, except in older adults, where the proportion was 54% and 58% in the prepandemic and pandemic periods, respectively (online supplemental figure S5). It is important to note that, before the pandemic, the proportion in younger adults was not at the 60% threshold, but very close to it (59%). During the pandemic period, for any comorbidity or age category, an increase in the ‘Access’ category proportions and a corresponding decrease in the ‘Watch’ category proportions were observed. However, at the expense of the ‘Watch’ category, the ‘Access’ category proportions were systematically higher among individuals without comorbidity.

The combined impact of the COVID-19 pandemic and the number of comorbidities

For each age group, the presence of comorbidities was associated with higher antibiotic utilisation rates compared with the absence of comorbidity after adjustment (table 1, figure 3 and online supplemental table S4). The pandemic and respiratory viruses' resurgence periods were associated with decreased utilisation rates among children and adults after adjustment, compared with the prepandemic period (table 1 and figure 3). However, in table 1, interaction terms with rate ratios higher than 1 indicate that the reduction during the pandemic and respiratory viruses' resurgence periods was less pronounced for that given number of comorbidities. The decreases observed during the pandemic and respiratory viruses’ resurgence periods represent a good approximation of what happened for most individuals. However, the decrease during these two periods was less important in children and adults with three comorbidities or more. Children with two comorbidities were also less impacted by the decrease during the pandemic period. Due to the high statistical power of the study, other interaction terms were statistically significant but had rate ratios close to the null value; we opted not to consider them as clinically significant. Figure 4 offers an alternate presentation of this combined impact on antibiotic utilisation. For instance, the pandemic and respiratory viruses' resurgence periods in adults with three or more comorbidities were associated with adjusted decreases of 13% (95% CI 12% to 14%) and 7% (6% to 8%) in utilisation, respectively (children: 33% (21% to 43%) and 23% (8% to 35%)), compared with the prepandemic period. This is less pronounced than in adults without comorbidities, where adjusted decreases were instead 25% (25% to 25%) and 19% (18% to 19%) (children: 63% (62% to 63%) and 37% (36% to 38%)) for the same periods, respectively. In children with two comorbidities, the pandemic period was associated with an adjusted decrease of 50% (45% to 55%) in utilisation compared with the prepandemic period, which is also less pronounced than the observed decrease in children without comorbidities.

Figure 3. Adjusted rate ratios for the combined impacts of the periods and the number of comorbidities among children and adults (≥ 18 years), April 2018–March 2022. *Models for children and adults are adjusted for month, sex, age group, area of residence and material and social deprivation indices. DDD, defined daily dose; DID, DDDs per 1000 individual-days.

Figure 4. Adjusted rate ratios of each period within strata of the number of comorbidities among children and adults (≥ 18 years), April 2018–March 2022. *Models for children and adults are adjusted for month, sex, age group, area of residence and material and social deprivation indices. DDD, defined daily dose; DID, DDDs per 1000 individual-days.

Discussion

In Quebec, community antibiotic utilisation sharply decreased in all comorbidities and age categories with the arrival of the pandemic and remained low despite a slight increase during the respiratory viruses' resurgence period. These trends were also observed for most of the commonly prescribed antibiotic classes and for azithromycin. However, during the pandemic and the respiratory viruses’ resurgence periods, children and adults with three or more comorbidities, as well as children with two comorbidities during the pandemic period, experienced smaller decreases. The improvements in the ‘Access’ and ‘Watch’ category proportions during the pandemic allowed each age group to meet the WHO recommendation, except for older adults. However, the ‘Access’ category proportions were consistently higher in individuals without comorbidities.

In Quebec, as in other countries and other Canadian provinces, community antibiotic utilisation decreased with the arrival of the pandemic.¹⁸,³⁹⁴¹ Several studies have reported that public health measures against COVID-19 have reduced common respiratory infections.1922 24 34 37,^{39 41} They suggested that this reduction, which includes bacterial superinfections and viral infections often inappropriately treated with antibiotics, was the main factor of the reduction in antibiotic utilisation during the pandemic. Similarly, the decrease in antibiotic utilisation observed in our study coincides with the implementation of public health measures to limit the spread of SARS-CoV-2 (physical distancing, mask wearing, lockdown, curfew, etc.) and with the reduction in respiratory infection activity that was documented during the pandemic.^{50 61} The observed increase in antibiotic utilisation during the summer of 2021 coincided with the resurgence of respiratory viruses.⁵² Additionally, we noticed a decrease in each antibiotic class used for respiratory tract infections (RTIs); however, rates of fosfomycin and nitrofurantoin, two antibiotics primarily prescribed for UTIs, remained stable in adults during the pandemic; this was also reported elsewhere.1922 24 34 37,³⁹ The slight reduction observed in the rates of sulfonamides and trimethoprim for older adults is likely due to this class being commonly used to treat UTIs and, to a lesser extent, RTIs. Other studies that also associated this class with UTIs reported a slight decrease,^{38 39} a small increase⁴¹ or a stability³⁴ in the utilisation of this antibiotic class during the pandemic. The reduction in antibiotic utilisation may also be explained by the shift from in-person to virtual care; however, its impact remains uncertain, given conflicting study results.³⁹⁶²,⁶⁶

An increase in azithromycin utilisation during the pandemic was observed in some European countries, such as France, Greece, Ireland and Italy.³⁶ This increase was probably a result of a preliminary study suggesting a potential efficacy of azithromycin against SARS-CoV-2, which was later invalidated.³⁹⁵⁵,⁶⁰ On the contrary, in Quebec, azithromycin utilisation decreased during the pandemic, as reported elsewhere in Canada,^{38 39} the USA^{22 23} and other European countries.³⁶

With the pandemic arrival, the quality of antibiotic utilisation in Quebec was better and in line with the WHO AWaRe recommendations in each age group, except for older adults, where the proportion has also improved and is now close to the targeted threshold. It will be important to maintain this trend and improve it in older adults to preserve the efficacy of ‘Watch’ and ‘Reserve’ antibiotics, which have a higher risk of resistance and are of greater clinical importance. Overall, the ‘Watch’ category proportions also decreased during the pandemic in Portugal, Ontario and across Canada.^{18 42 67} Moreover, consistent with our observations, the Ontario study reported that, during the pandemic, the reduction in broad-spectrum antibiotic prescriptions, most of which belong to the ‘Watch’ category, was less pronounced among prescribers with more comorbid patients.⁴² Our results are also supported by another research study, which noted that, in general, individuals with multiple comorbidities frequently received broad-spectrum antibiotics.⁹

The combined impact of the pandemic and a high number of comorbidities on antibiotic utilisation observed in this study is consistent with the observations of the other studies conducted in developed countries. The unadjusted study in France showed that, between 2019 and 2020, the reduction of antibiotic prescriptions was less pronounced in individuals with comorbidities (−12.5%) than in those without comorbidities (−20.1%).⁴¹ Similarly, the Ontario study revealed that prescribers with more comorbid patients were less inclined to reduce antibiotic prescribing during the pandemic than prescribers with fewer comorbid patients.⁴² The reasons for the less pronounced reduction in individuals with multiple comorbidities are not well understood, but could be explained by their greater need for antibiotics and/or by the fact that prescribers lower their prescribing threshold for fear of complications, given their frailty. Further studies are needed to analyse whether antibiotic utilisation is justified in this population by verifying concordance with clinical guidelines, which could confirm their actual need for antibiotics to optimise their utilisation and combat resistance.

Our dataset did not allow us to determine whether access to prescribers influenced the reduction in antibiotic utilisation. Our results on antibiotics for UTIs indicated that access to prescribers for these antibiotics was not an issue during the pandemic, but prescriptions issued by physicians decreased and were replaced by pharmacist prescriptions. This might be attributed to legislative changes in January 2021 in Quebec, which allowed pharmacists to prescribe antibiotics for UTIs in women up to 5 years after an initial prescription by a physician (previously limited to just 1 year), provided that no more than one treatment is given per 6-month period or two treatments over 12 months.^{68 69} This situation might also be attributed to the fact that pharmacists may have compensated for the more restricted access to physicians during the pandemic. However, this cannot be the case for all antibiotics since, in Quebec, pharmacists can only prescribe antibiotics for minor conditions that do not require re-evaluation, that have already been diagnosed in the past and for which antibiotics have already been prescribed by a physician or by a specialised nurse practitioner.⁷⁰ Thus, if patients with comorbidities had more stable access to physicians during the pandemic in Quebec, as suggested by our data on UTIs and as observed in Ontario,⁷¹ this may have led them to higher antibiotic utilisation in general, compared with individuals without comorbidities. Further research is needed to determine whether access to prescribers contributed to the overall reduction in antibiotic utilisation and the less pronounced decrease observed in individuals with multiple comorbidities during the pandemic. Pharmacists’ practices should also be evaluated to ensure that antibiotics are not prescribed unnecessarily.

Strengths and limitations

To our knowledge, except for the unadjusted study in France,⁴¹ this study is the first to use individual-level data to measure the combined impact of the pandemic and the specific number of comorbidities on antibiotic utilisation in a developed country. This study included a large and comprehensive population-based sample. A previous study had already demonstrated the possibility of using QICDSS health administrative data to perform the surveillance of community antibiotic utilisation in line with international standards.⁷² Previous studies have demonstrated the high quality and validity of the health administrative data used and of algorithms employed to identify comorbidities.⁵³⁷³,⁷⁵

This study has some limitations. First, since the public drug insurance plan does not cover people under 65 years with private insurance, our study may have an over-representation of individuals with lower socioeconomic status in this age group compared with the general population. However, our regression models showed minor variations in antibiotic utilisation across material and social deprivation indices, and the impact on rate ratios was very limited, regardless of whether adjustments were made or not. Moreover, a previous study comparing age-adjusted utilisation rates from the QICDSS to those of the Public Health Agency of Canada (covering 97% of Quebec’s community pharmacies) found similar utilisation rates, suggesting that any impact on the generalisability of our findings would be limited.⁷⁶ Second, a patient who avoided healthcare could have been classified as having no comorbidities since undiagnosed comorbidities would not have been documented, leading to an underestimation of the association between antibiotic utilisation and the number of comorbidities. Third, as in similar studies, the number of pharmacy services was used to estimate antibiotic utilisation. This method could overestimate the actual antibiotic utilisation if a patient collects their antibiotic prescription from the pharmacy and does not consume it. However, this information is the most reliable source for estimating what the population consumes, as it is recorded directly by pharmacists and is the last point of entry before antibiotics are dispensed to patients. Fourth, by calculating the number of pharmacy services, a treatment split into two services (such as two pillboxes) counts as two prescriptions. Antibiotic utilisation could be overestimated in individuals with comorbidities, who are more likely to use pillboxes. However, this bias does not appear to significantly impact the validity of our results since the trends were the same for utilisation rates in DID. Finally, residual confounding may persist, considering that the quintiles of social and material deprivation are ecological variables and not individual measures of deprivation, but also because we could not consider the type and stage of comorbidities in our analyses.^{7 8 14 77}

Conclusion

In Quebec, the onset of the COVID-19 pandemic and public health measures to limit its spread reduced antibiotic utilisation rates, which remained low despite a slight increase with the resurgence of respiratory viruses in the summer of 2021. However, the decrease was smaller in individuals with several comorbidities. Further investigations are needed to evaluate the impact of this antibiotic utilisation reduction on resistance and guide efforts against antibiotic resistance.

Data availability statement

No data are available. Data used for this study cannot be shared. Any questions can be addressed to the corresponding author.