Impact of the COVID-19 pandemic on prescription drug use and costs in British Columbia: a retrospective interrupted time series study

Strengths and limitations of this study

  • This study used real-world data of every prescription drug dispensed at a community pharmacy or hospital outpatient pharmacy in British Columbia over the study period, and irrespective of the drug insurance payer.

  • Interrupted time series (ITS) is one of the strongest quasi-experimental research designs that has been increasingly used in drug policy research. ITS has the distinct advantage of being methodologically rigorous and the graphical display of the ITS results are easily interpretable.

  • The pandemic events and the responses to it impacted medication dispensing of select medication subgroups over the study period.

  • Medication dispensing does not necessarily reflect medication consumption and treatment adherence.

  • The categorisation of drugs into large therapeutic categories and subgroups may have obscured subtle variations in the dispensing of individual drugs.

Introduction

The COVID-19 pandemic led to concerns about both the overuse and the underuse of prescription medications. In terms of underuse, pharmaceutical supply chains were disrupted, and fears of drug shortages and hoarding led to numerous policy shifts.1–4 The proliferation of misinformation and fraudulent data about the potentially harmful effects of certain medications and the potential for others to treat COVID-19 received considerable media attention.5 Consequently, patients may have stopped effective cardiovascular medications for reasons not supported by evidence.6 In terms of overuse, the demand for hydroxychloroquine and azithromycin soared following anecdotal reports of their effectiveness against COVID-19.7 Moreover, employment losses associated with COVID-19 may have had adverse health impacts on the population,8 particularly since the pandemic revealed and exacerbated health inequalities.9

To date, data-driven analyses of the impact of the pandemic on prescription drug utilisation have shown mixed results for specific classes of medicines.10 For example, two studies from the USA have shown different results for benzodiazepine use, with one study in California showing no major shift11 and another showing an increase.12 In Ontario, modest increases in the use of psychotropics13 and opioids were reported among nursing home residents.13 14 In contrast, no utilisation changes were found for antidepressants and benzodiazepines in the Canadian population as a whole.15 The discrepancies in these findings can largely be attributed to the duration of the post-COVID observation period, the drug use measure, the health system disparities and the COVID-19 pandemic policy response differences.

To contain the spread of COVID-19, most countries implemented travel and mobility restrictions, physical distancing measures and intermittent lockdowns, and instituted surveillance with testing and contact tracing.16 17 In March 2020, healthcare delivery was reconfigured to accommodate physical distancing and virtual visits. Although virtual care has been established for decades in Canada, the COVID-19 pandemic drove its uptake.18 19 In British Columbia (BC), pharmacists’ scope was expanded to permit them to extend and transfer prescriptions, and physicians’ prescribing practices changed to ensure that patients could continue to receive their medications, including virtual renewals and some longer maintenance prescriptions.18 20 To safeguard against hoarding and shortages of drug supply,3 4 many BC pharmacies imposed temporary 30-day limits on prescription length, which were gradually lifted about 5 weeks into the declaration of the COVID-19 pandemic. This temporary drug rationing translated into increased prescription volumes and additional dispensing fees for more frequent refills.21 22 Compounded with a public health emergency related to the COVID-19 pandemic, BC has been distinctively going through another public health emergency related to escalating opioid overdoses and related deaths since April 2016. At the intersection of these dual public health emergencies, a risk mitigation approach was released to guide healthcare providers in supporting their patients with substance use disorders while minimising community spread of COVID-19.23 These guidelines provided clinical guidance and practice updates to providers for prescribing safe supply pharmaceuticals for these patients.23

While prior studies have investigated drug-specific or disease-specific impacts, it remains important to understand how COVID-19 and associated mitigation measures impacted overall prescription drug utilisation in ambulatory care settings. The health system dynamics, including the reconfiguration of healthcare delivery, the actual and anticipated COVID-19-related drug shortages and the COVID-19 policy mitigation measures, certainly affected medication use. These shifting patterns have yet to be unravelled empirically. Therefore, we assessed changes in dispensing trends across all therapeutic drug categories, both overall and by individual drug classes in BC. We hypothesised that prescription drug use shifted from typical trends following the onset of the pandemic and its associated policy responses, particularly for drug groups used to treat chronic diseases, infections, and mental health and substance use disorders.

Methods

Study context

While physician visits, including diagnostic tests and hospitalisations, are universally publicly funded in Canada and free at the point of care, prescription drugs are paid for through a mix of private and public insurance coverage and out-of-pocket payments.24 In BC, the Fair Pharmacare programme provides universal public coverage of outpatient prescription drugs. For most residents, coverage is based on income-based deductibles and copayments.25 Many residents also have private insurance that pays a portion of these charges and generally covers drugs that are not on the public formulary.

Study setting

We conducted a retrospective population-based study to assess the changes in total dispensing volume and costs for prescription drugs. Our study cohort included all individuals registered for BC’s publicly funded medical services plan (MSP) for at least 1 day in a week between 1 January 2018 and 28 March 2021. These individuals would have filled at least one medication prescription and irrespective of the drug insurance payer (public and/or private insurer). We excluded those who receive drug funding through the federal government, such as First Nations and veterans, as these data were not made available to us. The percent of the population this would have excluded from the total sample is about 3%.

Data sources and measures

We used two linked administrative databases available through Population Data BC.26

The MSP consolidation file provided registration and demographic information on individuals registered with the provincial medical services plan.27 We linked this with PharmaNet data, using unique encoded identifiers, to obtain information on every prescription dispensed at a community pharmacy or hospital outpatient pharmacy in BC, including cost-related data for both public and private payers.28 PharmaNet is one of Canada’s most comprehensive and reliable pharmacy claims databases and has been extensively used in pharmacoepidemiology and health services research.29 30 Finally, we used the Health Canada-Drug Product Database to obtain the Anatomical Therapeutic Chemical (ATC) classification for each drug.31 These databases are described further in online supplemental appendix 1A.

Supplemental material

We classified prescriptions using the ATC anatomical main groups and further examined select therapeutic and pharmacological subgroups with high prescription volume32 that we hypothesised a priori might be impacted by the pandemic. These subgroups included ACE inhibitors, angiotensin receptor blockers (ARBs), antidiabetic drugs, antibacterials, antivirals, antipsychotics, antidepressants, anxiolytics, opioids and drugs used in addictive disorders such as drugs used in nicotine, alcohol or opioid dependence (online supplemental appendix 1B).

Our primary analysis focused on two major outcomes. We calculated weekly rates of prescription drug dispensing and costs, both overall and stratified by medication subgroups, over 169 weeks. The denominator for rates was the weekly number of individuals who are actively registered with the provincial medical insurance plan (ie, MSP) for at least 1 day in each study week. The rate of prescription drug dispensing was defined as the weekly number of prescriptions dispensed per 100 000 individuals. Prescription drug costs were the sum of the ingredient cost and dispensing fees. These costs were calculated as the total amount paid by public and private (ie, private supplemental insurance or out-of-pocket payment) payers to dispense the prescription drug. The rate of prescription drug costs was defined as the weekly drug spending per 100 000 individuals. We reported the relative changes in post-COVID-19 outcomes expressed as the ratio of the observed to the counterfactual value, had the COVID-19 pandemic not occurred. Moreover, we estimated the average pharmaceutical therapy duration defined by the number of days supplied divided by the number of prescriptions dispensed each week.

Statistical analysis

Using interrupted time series (ITS) analysis, we assessed changes in the level and trend of our outcomes after the start of the pandemic period.33 ITS is arguably one of the strongest quasi-experimental research designs that have been increasingly used in drug policy research.34 In our study, we had population-level data on every prescription dispensed in a community pharmacy over 169 time points, which makes the ITS the most appropriate research design in capturing the study outcomes. We compared the observed dispensing trends during the pandemic with expected dispensing (ie, counterfactual), estimated based on secular trends from 1 January 2018 to 15 March 2020.33 The counterfactual assumption in an ITS model is that these trends would have continued absent the COVID-19 pandemic. We assessed our models for non-stationary and autocorrelation using standard methods (eg, autocorrelation function (ACF) and partial autocorrelation function plots, and Durbin-Watson test). We controlled for any observed autocorrelation using appropriate adjustments in a generalised least squares model.35 We adjusted our models for weeks that included the pandemic declaration and seasonal variations related to the annual renewal of deductible amounts by adding stockpiling and seasonality terms to applicable models. All analyses were carried out using SAS V.9.4 and R V.3.6.1.

Patient and public involvement

Throughout this research, we engaged multiple stakeholders from government entities and academic institutions in the identification of research questions and communication of study results.

Results

As of the first week of 2018 and 2021, our weekly cohort size ranged between 4 298 768 and 4 370 568 individuals registered with MSP for at least 1 day a week, with relatively stable demographic characteristics over the study years. Of those actively registered with MSP in 2021, 52% were female, with a mean (SD) age of 46.2 (22.7) years (online supplemental figure S1 and table S1).

Overall utilisation

The model coefficient estimates for the level and trend changes in prescription dispensing rates and costs with their 95% CIs are available in online supplemental tables S2A and S2B. As shown in figure 1, we observed a significant stockpiling of 19.3% (p<0.01) in overall prescriptions dispensed on the week of the pandemic declaration. This was followed by an immediate level decrease of 2.4% (p<0.01) in overall dispensations. A sustained weekly increase followed that restored pre-pandemic trends by the end of January 2021. Overall expenditures followed a similar pattern and returned to pre-pandemic trends by mid-January 2021 (p<0.01).

Figure 1
Figure 1

Weekly dispensing rates and costs for overall medication prescriptions from 1 January 2018 to 28 March 2021 in British Columbia (BC), Canada. The black vertical dotted line represents the declaration of the state of emergency related to COVID-19 in BC. Counterfactual trends are indicated with a red dotted-trend line and were projected based on secular trends before the onset of the pandemic.

Class-specific results

A number of class-specific results were largely consistent with our overall findings. This was particularly the case for cardiovascular, respiratory system drugs, alimentary tract and metabolism, systemic hormonal preparations, and genitourinary system and hormones (online supplemental figure S2). Notably, we observed the largest stockpiling behaviour in respiratory drugs (+60.7%), cardiovascular system drugs (+20.5%), and alimentary tract and metabolism drugs (+20.5%) at the onset of the pandemic declaration. Across cardiovascular system subgroups, we found an overall decrease of 8.8% (p<0.01) in ARB prescriptions by the end of the study period associated with an 8% (p=0.05) marginally significant cost decrease. While we did not detect a statistically significant change in the dispensing rates of ACE inhibitors, we found an overall significant cost increase of 17.8% (p<0.01) in associated prescription costs by the end of March 2021. Similarly, the dispensing rates of antidiabetic drugs did not change significantly, whereas the costs had risen weekly, yielding an overall 3.7% (p=0.02) increase by the end of March 2021 (online supplemental figure S3).

We did note changes in utilisation in two of our classes of interest. First, we noted decreases in both prescriptions and costs for anti-infectives for systemic use, with an immediate level decrease of 31.9% (p<0.01) in prescriptions and a cost decline of 33.5% (p<0.01) (online supplemental figure S4). The declines in outpatient antibacterial and antiviral dispensing may be clinically significant. Our models detected immediate level decreases in antibacterial dispensing trends (30.3%, p<0.01) and costs (31.1%, p<0.01) and antiviral dispensing trends (22.4%, p<0.01) that remained below the counterfactuals by the end of the study period (figure 2, online supplemental figure S4).

Figure 2
Figure 2

Weekly dispensing rates for antibacterial and antiviral prescriptions in British Columbia, January 2018–March 2021.

In contrast, we found overall increases in nervous system drugs. While we observed a marginal immediate level decrease of 1.6% (p=0.04) in dispensing nervous system drugs, a sustained increase in trend followed, yielding an overall increase of 7.3% (p<0.01) by the end of March 2021. This trend corresponded to an overall increase of 12.1% (p<0.01) in associated prescription costs (figure 3). Utilisation changes for antipsychotics, anxiolytics, antidepressants and drugs used in addiction disorders were largely consistent with our overall findings. We found an overall increase of 2.3% (p<0.01) in antipsychotic prescriptions, yielding an overall cost increase of 15.1% (p<0.01) by the end of March 2021. For anxiolytics, our models detected an overall increase of 5% (p<0.01) in dispensing, corresponding to a 4.2% (p<0.01) increase in associated dispensing costs by the end of March 2021. While the dispensing rates of antidepressants remained relatively constant, the prescription costs experienced a sustained weekly increase yielding a 3.9% (p=0.02) relative cost increase by the end of March 2021. For drugs used in addiction disorders, we detected an immediate level decrease of 11.9% (p<0.01), followed by a week-to-week increase to return to expected levels by mid-March 2021. However, we did observe large changes for opioids, with a dispensing rate increase of 43.3% (p<0.01) by the end of March 2021 and a corresponding cost increase of 25.6% (p<0.01) that exceeded the expected levels throughout the first year of the pandemic (figure 4, online supplemental figure S5). The large magnitude of this effect size suggested that opioid dispensing may be clinically significant. Finally, we noted a systematic decrease in the days of therapy per prescription for opioids. This was contrasted with no or transient changes in the days of supply for other medication subgroups (online supplemental figure S6).

Figure 3
Figure 3

Weekly dispensing rates and costs for nervous system prescriptions in British Columbia, January 2018–March 2021.

Figure 4
Figure 4

Weekly dispensing rates for antipsychotics, anxiolytics, opioids, antidepressants and drugs used in addiction disorders in British Columbia, January 2018–March 2021.

Discussion

In this population-based study, we found that the COVID-19 pandemic and related mitigation measures led to an abrupt minor decrease in overall prescription drug dispensing that had largely reverted to pre-existing trends by the first year of the pandemic. Within these overall trends, however, there were heterogeneous effects across therapeutic drug categories and subclasses. For example, we observed sustained increases in drug dispensing for nervous system drugs that persisted through the first year of COVID-19. Relative decreases were detected for antibiotic and antiviral outpatient dispensing, but the sustainability of these declines beyond the first year of the pandemic remains unclear. This contrasted with only minor long-term changes for cardiovascular, respiratory system and alimentary tract and metabolism medications. We also observed significant stockpiling on the week of the state of emergency declaration for COVID-19 across every therapeutic drug category, with the largest stockpiling in respiratory system medications.

The transient increases in prescription drug dispensing and costs suggested that patients may have overstocked medications to avoid pharmacotherapy disruptions. Following the initial rise, dispensing rates and costs for overall prescription drugs fell, likely due to the ability to deplete supplies from early stockpiling and then returned to expected levels. Despite early concerns that the pandemic would exacerbate existing medication shortages due to reliance on international production of active pharmaceutical ingredients and increased demand for some drugs, no significant disruptions in medication access appear to have occurred in the first year of the pandemic in BC. Data-driven research suggested that the Canadian market did not experience major increases in drug shortages early in the pandemic.36

Overall, our findings for medications for chronic conditions are consistent with pharmacotherapy being maintained over time. Despite unfounded early controversies around the effectiveness of ACE inhibitors and ARBs among patients with COVID-19,5 13 37 their fills did not appear to substantially decline in BC and elsewhere.6 14 We did, however, note some cost increases associated with commonly prescribed medications for chronic conditions, such as ACE inhibitors and antidiabetic drugs. These increases in medication costs, borne by both private and public payers, are likely associated with additional dispensing fees due to shorter prescription lengths at the outset of the pandemic. Another factor contributing to the increased costs of ACE inhibitors is likely the variability in the cost of case mix drugs within this drug subgroup throughout the study period. This suggests possible price increases for these drugs over time or switches to brand-name medications. Our findings showed that the duration of medication supplied dropped at the onset of the pandemic for major drug classes and re-stabilised in 5 weeks, reflecting the provisional pharmacists’ rationing to avoid drug hoarding.

The observed decreases in outpatient antibacterial and antiviral dispensing may have some clinical significance. Nonetheless, a longer follow-up period would have better depicted whether these decreases were sustained or faded out beyond the first year of the pandemic These changes could have resulted from artifactual reduced contact with prescribers and less antibiotic prescribing in primary care,38 39 as well as real reductions in seasonal infections resulting from COVID-19 containment measures.40 41 Our findings align with research showing substantial declines in outpatient antibiotic prescribing.42 Still, they contradict prior studies that reported no changes14 in or increased consumption of antimicrobials in residential care homes and hospitals.43 These differences reflect the distinct prescribing practices across healthcare settings, with many patients hospitalised with COVID-19 inappropriately receiving antimicrobials.44 Finally, while antiviral compounds have been repurposed for possible treatment against COVID-19,7 we did not find evidence of their increased use in community settings.

In contrast, the increases in the dispensing rates of antipsychotics and anxiolytics and the constant patterns in antidepressant dispensing in the first year of the pandemic were not very notable. We did not examine individual drug dispensing across these subgroups which may have obscured the therapy indication for their use. Survey data reported increased anxiety and depressive disorder rates during the first wave of the COVID-19 pandemic due to prolonged and restrictive social restrictions and loss of employment.45–47 These studies mainly used self-reported measures48 that are subject to social desirability bias and may have, in some instances, captured the pre-pandemic baseline of those reporting psychiatric disorders. Our findings are consistent with those conducted over a similar follow-up period. Specifically, studies conducted in the USA throughout the first and second COVID-19 waves found persistent increases in Z-hypnotics and serotonergic12 and relatively constant changes in the dispensing of most psychotropic medications among privately insured cohorts,11 findings that underscored the impact of COVID-19 on mental health. In our study, the increases, although minor, in the rates of medication dispensing of nervous system drugs through BC’s second waves might be partly related to the effects of the pandemic on mental health.49 Alternatively, these modest changes may reflect patients’ reluctance to initiate pharmacological treatment or an increased likelihood of discontinuing pharmacotherapy.19 Previous research estimated that over 50% of primary care patients do not adhere to antidepressant therapy,50 and medication initiation for antidepressants and anxiolytics was suboptimal among new patients during the early months of the COVID-19 pandemic.51 These patients could have resorted to other treatment options, with reported increases in physician-based mental healthcare services through the first year of the pandemic.19 52 53 More research is therefore needed to understand clinical diagnosis variations between in-person and virtual care, therapy initiation and discontinuity, and therapy indication for individual drugs among these patients in light of the increased prevalence and burden of depression and anxiety.47

In parallel, some of the documented changes in opioid dispensing could have occurred, in part, following the rapidly developed risk mitigation guidance released in March 2020 to provide safe supply pharmaceuticals to people with substance use disorders.23 The examination of trends in other drug categories suggested that changes in opioid dispensing were specifically associated with changes in opioid prescribing and did not reflect a broader pattern, particularly with the decreases in therapy duration for which dispensing fees could be substantial. In the absence of medication review studies assessing the clinical context of opioid prescribing, dosages, forms, duration and type of prescription opioids, these surges warrant further examination to discern the impact of these dispensing changes, particularly with the enduring surges in opioid-related overdose deaths in BC.54 55 The increases in opioid dispensing have potential implications for clinical practice and policy and should be used in conjunction with other health resource utilisation to make inferences about drug safety and health system outcomes.

Strengths and limitations

Our study used large and comprehensive routinely collected data that captured all dispensed outpatient medications during the first year of the COVID-19 pandemic in a population-based cohort. It highlighted the utility of real-world data in generating real-world evidence on prescription drug utilisation in this population of patients and underscored its potential to drive healthcare decisions in response to changing environments.

While a single time-series study design is susceptible to history threats, most policies introduced in BC at the time were responses to the COVID-19 pandemic. Although the ITS analysis enabled us to project a linear counterfactual trend based on the pre-pandemic secular trends, this is likely to represent a conservative bias in terms of a return to the pre-COVID-19 period. The direction of the patterns in drug dispensing, particularly for opioids and antibacterials, should be the subject of further investigation, as they likely resulted from response policies. Nonetheless, they could be used as a proxy indicator for disease development (ie, tracer conditions) across the population that requires more attention. While we did not examine the impact of employment loss on access to drugs, the relatively stable rates of private insurance coverage56 suggested that people did not lose access to medicines due to unemployment. A limitation of data is that third-party paid amounts are not captured in PharmaNet. As such, we could not characterise if the prescription costs portion borne by patients were out-of-pocket payments or financed through supplemental private drug insurance. Another key limitation of this study is that the number of prescriptions dispensed does not necessarily reflect the consumption of those medications and treatment adherence. Lastly, our categorisation of drugs into large therapeutic categories and subgroups may have obscured subtle variations in the dispensing of individual drugs.

Conclusion

The COVID-19 pandemic and associated policy responses have had a heterogeneous impact on prescription drug dispensing. While dispensing trends have returned to expected levels for most drug classes, the sustained increases in dispensing nervous system drugs, particularly opioids, throughout the first year of the pandemic deserve further attention and continued monitoring. Moreover, the sustainability of the relative decreases in outpatient antimicrobial dispensing beyond the first year of the pandemic remains to be studied. As data become more available, monitoring the long-term impact of these prescribing trends on patient outcomes and health systems outcomes is warranted.

Data availability statement

Data may be obtained from a third party and are not publicly available. 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).

Ethics statements

Patient consent for publication

Ethics approval

This study was approved by the University of British Columbia behavioural research ethics review board (H20-01932).

This post was originally published on https://bmjopen.bmj.com