The impact of COVID-19 public health measures on the utilization of antipsychotics in schizophrenia in Manitoba – A population-based study

,


Introduction
Access to prescription medications for individuals with chronic conditions was negatively impacted by the coronavirus disease 2019 (COVID-19) pandemic (Esposti et al., 2020;Livori et al., 2023b).This was largely due to global supply chain shortages (Lau et al., 2022), reduced access to hospitals and primary healthcare service which reduced the medications prescribed, dispensed and administered (such as injectable medications) (Livori et al., 2023a;Stephenson et al., 2023), and the avoidance of healthcare due to fear of COVID-19 contagion (Smolić et al., 2023).
People with schizophrenia (SCZ) are a vulnerable group reporting difficulties with accessing medications during the pandemic (Barlati et al., 2022;Marshall et al., 2005;Miron et al., 2022;Penttilä et al., 2014).Additionally, those with SCZ experienced increased mental health issues from isolation related to the pandemic (Civan Kahve et al., 2021;Pajević et al., 2020), low compliance with treatment (Pennington and McCrone, 2018) and a higher likelihood of developing worse outcomes following a COVID-19 infection (Civan Kahve et al., 2021).Having sustained access to the antipsychotics in SCZ is an important issue by considering the high rate of non-adherence in those with SCZ and the negative effects of discontinuing an antipsychotic.Timing of treatment with antipsychotics in first episode of psychosis is of importance as the initiation of an antipsychotic after one week compared to within that the first week is associated with longer hospital admissions (Arribas et al., 2022).Antipsychotic trends in the general population may not accurately reflect the changes among specific populations such as those with SCZ, which indicates the need to perform further investigations specifically in this group.It is also likely that differences in the extent and intensity of COVID-19 public health measures between countries may have contributed to variability in access to medications, which questions the generalizability of the results in one country or region to another (Kisely et al., 2022).
Studies on the utilization of antipsychotics during the COVID-19 pandemic in those with SCZ have been conflicting, and mostly covered the early stages of the pandemic (Leong et al., 2022;McKee et al., 2021).In Australia, among 7873 people with SCZ, there was no significant change in the use of new (prescriptions with medical consultations) and all (new and refill) antipsychotics from April-May 2020, compared to the same period in 2019 (Kisely et al., 2022).Another study in Italy observed a significant decline in the incident use of long-acting injectable antipsychotics in SCZ outpatients in 2020 compared with the previous year (Barlati et al., 2022).In the general population in the USA, incident antipsychotic use was significantly reduced from March--August 2020 compared to a pre-COVID-19 period (January 2018-March 2020) (Kisely et al., 2022).
In the Canadian province of Manitoba, the first COVID-19 case was reported March 12, 2020 (Aboulatta et al., 2022a).Public health measures to limit the spreading of the virus were first enacted on March 13 and included closure of public spaces, businesses and schools, as described previously (Aboulatta et al., 2022a).On March 20, 2020, Manitoba declared a state of emergency for 1-month based on increasing COVID-19 cases.In the healthcare sector, this state of emergency meant that non-urgent hospital visits were restricted and/or suspended, and a one-month drug dispensing limit was implemented.The one-month drug dispensing limit ended May 11, 2020, because of increasingly stable drug supplies (Aboulatta et al., 2022a).
In this study, we aimed to compare the quarterly prevalent and incident dispensing of antipsychotics in Manitoba, Canada among those with SCZ during the COVID-19 pandemic compared with the expected trend in the previous five years.We hypothesized that COVID-19 public health measures led to a reduction in the prevalent and incident use of all antipsychotics among people with SCZ in Manitoba.

Study design and data sources
This was a population-based, repeated cross-sectional study using the Manitoba Population Research Data Repository, held by the Manitoba Centre for Health Policy (MCHP).The MCHP is a collection of several government administrative databases including healthcare, education, residential, and others.It includes person-level data on almost all residents of Manitoba (population ~ 1.4 million) and is not restricted by healthcare coverage, age, or income.An identification number is assigned to each individual in the registry permitting linkage to multiple data sources possible.The validation of these databases within the MCHP for population-based studies is well established (Jutte et al., 2011;Nickel et al., 2022).In the repository, we used the Drug Program Information Network (DPIN) (Manitoba Centre for Health Policy, 2013) for dispensed prescription drugs which included the date of dispensation (from 1995).Medications received in hospitals and nursing stations are not included in the DPIN.The physician claims and hospital discharge abstracts were used to capture the data on diagnoses (from 1970) (Manitoba Centre for Health Policy, 2012) based on the International Classification of Diseases (ICD).For sociodemographic data, we used the Manitoba Health Insurance Registry (age, sex, urban/rural residence at the beginning of each quarter, from 1970) and Statistics Canada (for income quintiles, from 1971).

Population and time frame
This study included all individuals who had received a SCZ diagnostic code (≥1 physician or ≥ 1 hospital claims for ICD-9-Clinical Modification: 295, ICD-10: F20, F21, F23.2 or F25) (Manitoba Centre for Health Policy, 2020) within five years prior to a quarter or year of interest (April 2015 to March 2021) and were listed in the Manitoba Health Insurance Registry (MHIR) for at least one day of coverage that quarter or year were included in this study.Individuals who met this criteria and who were prescribed at least one antipsychotic medication within each quarter from 2015 to 2021 fiscal years (April 1 to March 31) in Manitoba were included in the study.Exclusion criteria excluded any person with missing data on date of birth, sex, or those not covered by MHIR.Individuals who died or emigrated during the year of coverage but had one day of MHIR coverage were included that year but not subsequent years.We included data from April 1, 2015, to March 31, 2021; this range of time was included to cover before (April 1, 2015-March 31, 2020) and during COVID-19 periods (April 1, 2020-March 31, 2021).

Ethical and privacy committee approval
This study was approved by the Human Research Ethics Board of the University of Manitoba and the Manitoba Health, Seniors, and Active Living Health Information Privacy Committee.Informed consent was not required with the use of de-identified administrative data (Health Information Privacy Committee Project Number: 2020/2021-29, January 28, 2022).

Exposure, outcomes, and variables
The exposure was the introduction of the COVID-19 public health measures, which included restricting access to public spaces in the province of Manitoba, first enacted on March 13, 2020 (Aboulatta et al., 2022a).We defined the first fiscal quarter (FQ) of 2020 (2020-FQ1: April 1-June 30) as the primary intervention point, which was two weeks following the implementation of the COVID-19 public health measures in Manitoba (Aboulatta, Kowalec, et al., 2022).The second intervention point was defined as one year later: April 1-June 30, 2021 (2021-FQ1) to evaluate any further effects of the COVID-19 public health measures on drug dispensation.
The outcomes were the quarterly prevalent and incident use of antipsychotic dispensation.Antipsychotics were identified using their Anatomic Therapeutic Classification (ATC) code (antipsychotic agents: N05A, except N05AN).We considered all antipsychotics (Supplementary Table 1) collectively (in all and age and sex sub-populations with SCZ) and the most commonly used antipsychotics in those with SCZ in Manitoba (quetiapine, olanzapine, risperidone, and clozapine).Usage was further categorized by route of administration (either oral or injectable, Supplementary Table 1) and by generation (first-generation vs. atypical, Supplementary Table 1).
The prevalent proportion was defined as the number of individuals who were dispensed medication within each fiscal quarter during the study period per 1000 individuals with SCZ.The incidence proportion was defined as the number of individuals with no dispensation of the drug of interest for at least three years prior to receiving their first prescription per 1000 individuals with SCZ.
The following information was collected on each individual: sex (male/female), age at the start of the quarter of interest (categorized as: ≤18, 19-39, 40-64, 65-79 and ≥ 80 years) (Leong et al., 2022), and residence (urban and rural).Household income was categorized into five quintiles by urban and rural area representing a total of 10 categories ranked from low (1) to high (5) income level.We computed the Elixhauser Comorbidity Index (Elixhauser et al., 1998) per person (Supplementary Table 2) which included 30 comorbidities (excluding psychoses) as a count: 0, 1, 2, 3 or ≥ 4. Polypharmacy was defined two ways: >5 dispensed different medications in three months or > 10 dispensed different medications in three months (with the exclusion of over-thecounter drugs).

Statistical analyses
We described the total SCZ population at the first fiscal quarter of 2020 using the mean (standard deviation) or frequency (percent) on the following characteristics: age, sex, residence, income quantile, comorbidity count, and polypharmacy use.
All analyses compared quarterly antipsychotic utilization (either prevalence or incidence) at two time points during the pandemic to the time trend during the previous five years.For all objectives, a linear autoregression model was fit and the Durbin-Watson statistic up to the 4th order was used for autocorrelation (Supplementary Table 3 and 4).Quarterly prevalent or incident proportions per 1000 SCZ participants was reported using effect estimate, 95 % confidence interval (95%CI) and p-value.The primary analyses included all antipsychotics, and the four most commonly used antipsychotics, which were then stratified by age categories and by sex.Secondary analyses included a comparison by antipsychotic generation and by administration route.A sensitivity analysis was conducted for all objectives to assess the sensitivity of the results to the inclusion criteria and accuracy of SCZ diagnosis, where the SCZ definition was limited by requiring ≥2 physician claims (instead of ≥1) or ≥ 1 hospital claim in the two years (instead of five years) before the quarter or year of interest.
All analyses were conducted using SAS statistical software (version 9.4).Missing data were not imputed and statistical significance was set at p < 0.05.

Results
The population of those meeting the definition of SCZ in Manitoba in the first fiscal quarter ranged over the study period from 8196 (2015) to 9166 (2021, Table 1).At the beginning of April 2020, the average age of the study participants was 46.9 years (SD = 17.9).More than half were male (59.7 %), two-thirds of the participants (74.8 %) resided in an urban region and 28 % were from the lowest urban income quintile.When stratifying by sex (Table 1) and age (Table 2), these estimates were largely unchanged from the full SCZ population.
The quarterly prevalent use of all antipsychotics showed an upward trend during the study with the minimum proportion in April-June 2018 (707.4/1000) and the maximum in January-March 2020 (733.2/1000,Fig. 1).The prevalent use of the four most common antipsychotics were as follows: olanzapine (206.7/1000),risperidone (190.8/1000),quetiapine (174.4/1000), and clozapine (100.9/1000,Fig. 1).All age and sex subgroups showed an upward trend in the quarterly prevalent use of all antipsychotics during the study period (Fig. S1 and S2).The quarterly prevalent use of first-generation and oral antipsychotics showed a decreasing trend during the study period while atypical and injectables showed a decreasing trend (Fig. S3 and S4).
The quarterly incident use of all antipsychotics showed a downward trend, with a maximum in April-June 2016 (4.9/1000) and a minimum in April-June 2020 (2.3/1000, Fig. 2).All age and sex subgroups showed a downward trend in the incident use, except those ≥80 years during the study period, which showed an increase (Fig. S5 and S6).First-generation and atypical antipsychotics showed a decreasing trend in the quarterly incident use during the study, while oral and injectable antipsychotics showed an increasing trend (Fig. S7 and S8).
The quarterly prevalent use of all antipsychotics and the four most  3).However, stratifying by age identified that those 65-79 years of age experienced a significant decrease in the quarterly prevalent use of all antipsychotics (Table 4).
We noted a significant decrease in the quarterly incidence of all antipsychotics in the first fiscal quarter of 2020, when compared to the expected trend (Table 3).Extending the analyses to include one year later (2021-FQ1), this decrease in antipsychotic incidence use was no longer significant (Table 3).In 2021-FQ1, compared to the expected trend, a significant increase in the incident use was seen for risperidone (Table 3), but for the remaining antipsychotics, no significant change was noted, nor for the age and sex stratification (Table 4).
There were no significant changes at either time point during the pandemic for the prevalent use of first-generation and atypical antipsychotics compared with the expected trend (Table 3).However, atypical antipsychotics showed a significant increase in incident use in 2021-FQ1 compared to the expected trend (Table 3).Both oral and injectable antipsychotics experienced non-significant changes in prevalent and incident use in 2020 and 2021, compared to the expected trend in the previous five years (Table 3).
The population of those meeting the definition of SCZ used in the sensitivity analyses was 5642 individuals and were similar to the population used in the primary analysis based on key sociodemographic and clinical indicators (Table 1).Regression estimates from the sensitivity analyses were similar in terms of direction and effect size to that of the original inclusion criteria for SCZ (Table 5), albeit a non-significant decrease was noted in the quarterly incident dispensation of all antipsychotics in the first fiscal quarter of 2020, when compared to the expected trend.

Discussion
This study examined the quarterly prevalent and incident use of antipsychotics in individuals with SCZ during the first two years of the COVID-19 pandemic in Manitoba, Canada.Adequate initiation and continuation of antipsychotic drugs is crucial to reduce symptoms associated with SCZ (Marshall et al., 2005;Penttilä et al., 2014) and studies have indicated that during the COVID-19 pandemic, access to antipsychotics was interrupted for those with SCZ (Barlati et al., 2022;Marshall et al., 2005;Miron et al., 2022;Penttilä et al., 2014).Overall, we noted a significant reduction in the incidence of antipsychotics in SCZ during the first quarter of the first year of the pandemic but not the second year compared with the expected trend.Additionally, in those with SCZ, a significant rise in atypical antipsychotics and risperidone incident use during the second year of the pandemic was observed.We found that those with SCZ who were 65-79 years old had a significant decrease in the prevalent use of antipsychotics in the first fiscal quarter of 2020 compared with the expected trend.This result could be due to higher mortality rate in this age group, considering more comorbidities (Pranata et al., 2021).We did not detect the same result in those ≥80 years, which might be due to smaller population size.
The incident antipsychotic use in the first year of the pandemic was significantly reduced compared to the previous five years, although it does not seem to be clinically significant considering the effect size.A previous study also observed a significant decline in the incident use of long-acting injectable antipsychotics in SCZ during the COVID-19 pandemic compared with the previous year (Barlati et al., 2022).Studies from the general population have also shown a significant reduction in incident antipsychotics from March 13-August 8, 2020, to the previous two years in the USA (Nason et al., 2021).However, other studies were conflicting with ours.An Australian pharmaceutical claims study including 7873 SCZ individuals treated with antipsychotics found no change in the incident use between April-May 2020 to the same period of time in 2019 (Kisely et al., 2022).This study only utilized antipsychotic dispensations that were dispensed ≥3 times during the study, possibly contributing to the conflicting results.Our finding of a significantly decrease incident use in all antipsychotics in SCZ may reflect access to drug supplies was interrupted among new antipsychotics users.This reduction could be due to challenges in connecting with newly diagnosed cases of SCZ (Lacro et al., 2002;Pajević et al., 2020) because of health system access issues, public health restrictions, fear of contagion (Smolić et al., 2023) (which itself could be worsened by uncontrolled psychotic symptoms) as well as limiting immigration to Manitoba from other regions (Aboulatta et al., 2022a).The finding of decrease incident use in the first year of the pandemic was no longer apparent in the second year and may reflect a return to the expected trend when restrictions were lifted.We found a significant prevalent decrease in antipsychotics in the first fiscal quarter of 2020 in 65-79-year-old group, compared to the previous five years.A UK study of 37,500 individuals seen weekly for mental healthcare, assessed the impact of the pandemic on the frequency of antipsychotics mentioned within electronic healthcare  records (Patel et al., 2021).They noted a steady upward trend in the antipsychotic mentions in those 65 and older during the period of study from July 2019 to July 2020 (Patel et al., 2021).However, they did not restrict to only those with SCZ.In our study, a reduction in prevalent antipsychotics prescribed to those 65-79 years old with SCZ in the early pandemic may have been due to challenges in the delivery of virtual healthcare or access to pharmacies, and highlights the potential importance of caregiver support.Although there is the possibility this drop was a chance finding not associated with the pandemic which is in line with other papers cited here.Members of our study team investigated the use of antipsychotic prescribing in the general population of Manitoba during the pandemic but did not find a significant change in the prevalent antipsychotics in this same age range (Leong et al., 2022).
Subgroup analyses found a significantly increased incident use of atypical antipsychotics in 2021-FQ1, relative to the expected trend.The increase in incident use could be the result of new users of antipsychotics that were not captured during the initial few months of the pandemic, as we noted a significant decrease incident use in 2020, but not in 2021.This could also be the result of COVID-19 induced psychosis (Chaudhary et al., 2022).This significant increase may have been driven by the increased incidence of risperidone, an atypical antipsychotic, during 2021-FQ1.Risperidone was the only individual antipsychotic with a significant increase in incident use which could be a chance finding by considering the fluctuations across the study time and changes in other individual antipsychotics assessed.
Like other studies in SCZ (Kisely et al., 2022) and in the general population (Leong et al., 2022), we did not find a change in the prevalent use of antipsychotics during the COVID-19 pandemic compared with the previous five years.This may suggest that for those with SCZ in Manitoba, access to antipsychotics was sustained among existing users despite public health measures that were implemented during the pandemic.
The sensitivity analysis was conducted by modifying the definition of SCZ and found similar results to that of the primary analyses in terms of the direction of effect.However, the significantly decreased incident finding from the primary analyses was no longer significant in the sensitivity analyses.This may have been due to a smaller number of individuals with SCZ included in the sensitivity analyses (N = 5642 in the sensitivity analyses vs. N = 9045 in primary analyses).Using two physician claims instead of one may have increased specificity in the SCZ definition but at the expense of a smaller sample size.
Incidence of atypical and injectable antipsychotics showed no significant change in 2020-FQ1, compared to the expected trend.Another study found that in admitted SCZ cases in the hospital, the long-acting injectable antipsychotics initiation dropped significantly (48.3 %) in the post-COVID-19 period.They covered a period of 12 months before (pre-COVID-19) and 12 months after March 11, 2020 (post-COVID-19) (Miron et al., 2022).This finding is different from ours, as a nonsignificant increase was observed, in the incident injectable antipsychotics in 2020-FQ1, which could be due to the different settings and prescription practices in different regions.
According to the present study, newly diagnosed cases of SCZ, males, and those between 19 and 39 years were more likely to experience interrupted access to antipsychotics due to COVID-19 public health measures.Incident users in this study include newly diagnosed cases of SCZ and chronic users of antipsychotics with ≥3-year interruption in their antipsychotic therapy.Findings from this study are of importance for policymakers to continue the efficient means for ensuring continued access to pharmaceutical drugs in the incident use of antipsychotics.Our results additionally have implications to define the direction of investment in the healthcare sector after the pandemic is over and of the need to direct policies in future pandemics.This study lays the foundation,   *Statistically significant at p < 0.05.
along with many other studies, to understand medicine resource management during the COVID-19 pandemic in Canada.Future research examining the impact of these trends on outcomes such as suicide, hospitalizations, or mortality, are also needed to inform policies.There remains a need for further investigation into the focus on COVID-19 induced measures on adherence to pharmacotherapy specifically in schizophrenia, with consideration for the severity of pandemic waves.
Strengths of this study included the use of administrative databases to describe antipsychotic dispensing which is not subject to reporting bias that may be present in self-reported surveys.Moreover, these databases provided drug dispensing information in a relatively large sample size (N > 9000) that is not restricted to age, income level, or drug coverage, which is often a limitation in other jurisdictions.This reduces the probability of selection bias and enhances the generalizability of our results across Canada and potentially to other countries with universal healthcare access.Compared to similar studies in other countries, the present study covers the entire first year of COVID-19 pandemic, whereas others compared only the first few weeks or months (Barlati et al., 2022;Kisely et al., 2022).We included secular trend and correction for autocorrelation of the prevalent or incident utilization in the model which is expected to result in improved parameter estimation rather than student t-test and regression models (Abeysinghe et al., 2003;Aboulatta et al., 2022b;Leong et al., 2022).This study has a few limitations worth noting, including that our results may not be generalizable outside the Canadian province of Manitoba, or Canada, given the differences in the COVID-19 public health measures by province and internationally.Using administrative databases has limitations such as the potential to misclassify those with a SCZ diagnosis.To address this, we conducted a sensitivity analysis which was largely consistent with the primary results.We did not have access to data on medications received in hospital and nursing stations.In this study, we accessed data on antipsychotic dispensing and not the actual consumption of medicines.Moreover, in Manitoba, prescription fills were temporarily restricted to a one-month supply (March 19-May 1, 2020) in an effort to minimize potential distribution shortages.It is not likely that this policy would affect the observed trends on dispensation but the extent to which this could have influenced the observed trend is not known.We also did not assess the potential adverse effects of COVID-19 public health measures on outcomes in SCZ as well as the association between antipsychotic use and the different waves of COVID-19 pandemic over the time of this study, which may present future directions.

Conclusions
In summary, we found an overall lower incidence of antipsychotic use during the initial period of COVID-19 compared to the expected trend among individuals with SCZ in Manitoba, Canada.While no change in antipsychotic prevalence was observed during the initial period of the pandemic compared to the expected trend.Future research examining the impact of these trends on SCZ outcomes is needed to help inform future policies to ensure access to treatment is not inadvertently interrupted.

Declaration of competing interest
None.

Fig. 1 .
Fig. 1.The overall and individual antipsychotic quarterly prevalence (per 1000) in individuals with schizophrenia in Manitoba, from 2015 to 2021.Grey bars highlight the quarterly periods under investigation representing the first quarter of COVID-19 pandemic and one year into the COVID-19 pandemic.FQ: Fiscal quarter.

Fig. 2 .
Fig. 2. The overall and individual antipsychotic quarterly incidence (per 1000) in individuals with schizophrenia in Manitoba, from 2015 to 2021.* Indicates statistically significant at p < 0.05.Grey bars highlight the quarterly periods under investigation representing the first quarter of COVID-19 pandemic and one year into the COVID-19 pandemic.FQ: Fiscal quarter.

Table 1
Characteristics of the schizophrenia population in Manitoba, in the first fiscal quarter (April-June) of 2020.
a Versus rural residence.b Hospital and physician claims data used to compute the Elixhauser Comorbidity Index on a yearly basis with the following denominators: N (primary analyses): 9053, N (sex-stratified): 5403 Males, 3650 Females; N (Sensitivity analyses): 5644.c Based on concurrent use in 3 months.The population stratified by age is in

Table 2
Characteristics of the schizophrenia population in Manitoba, in the first fiscal quarter (April-June) of 2020, by age.
a Versus rural residence.b Hospital and physician claims data used to compute the Elixhauser Coc Based on concurrent use in 3 months.d Cell suppressed due to N ≤ 5.

Table 3
Time series using linear autoregression models summarizing the association between COVID-19 public health measures and the prevalent and incident antipsychotic use in schizophrenia in Manitoba, in the first fiscal quarter of 2020 and 2021.

Table 4
Time series using linear autoregression models summarizing the association between COVID-19 public health measures implementation and prevalent and incident antipsychotic use in the sub-populations with schizophrenia in Manitoba, in the first quarter of 2020.
* Statistically significant at p < 0.05, N/A = unable to compute due to 0 value.

Table 5
Sensitivity analysis utilizing different criteria to define schizophrenia (N = 5642 individuals with schizophrenia in Manitoba).