Overview

We conducted a retrospective interrupted time-series analysis of changes in prescribing of SSHM after the declaration of the COVID-19 pandemic [18]. We used Ontario administrative databases held at ICES (formerly the Institute for Clinical Evaluative Sciences). We used the Narcotic Monitoring System database to identify hydromorphone and opioid agonist therapy prescribing; this database records information for all controlled substances (including prescription opioids) dispensed from community pharmacies in Ontario, regardless of payer. We also used the Canadian Institute for Health Information’s (CIHI) Discharge Abstract Database, the National Ambulatory Care Reporting System, the Ontario Health Insurance Plan (OHIP) database, and the OHIP Registered Persons Database to identify diagnoses and procedures during inpatient hospital admissions and emergency department visits, and to describe demographic characteristics of the cohort (S2 Table). These datasets were linked using unique encoded identifiers and analyzed at ICES. ICES is an independent, non-profit research institute whose legal status under Ontario’s health information privacy law allows it to collect and analyze health care and demographic data, without consent, for health system evaluation and improvement. ICES is a prescribed entity under Ontario’s Personal Health Information Protection Act (PHIPA). Section 45 of PHIPA authorizes ICES to collect personal health information, without consent, for the purpose of analysis or compiling statistical information with respect to the management of, evaluation or monitoring of, the allocation of resources to or planning for all or part of the health system. Projects that use data collected by ICES under section 45 of PHIPA, and use no other data, are exempt from REB review. The use of the data in this project is authorized under section 45 and approved by ICES’ Privacy and Legal Office. All analyses were completed on de-identified data. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for observational studies.

Eligibility and classification

We included individuals with opioid use disorder who received immediate release hydromorphone doses consistent with SSHM prescribing between March 22, 2016 and August 30, 2021. We chose to begin the study period in 2016 because it is when SSHM prescribing first occurred in Ontario. We defined the pre-pandemic and pandemic periods based on dates occurring before versus on or after March 17, 2020, the date a pandemic-related state of emergency was declared in Ontario, respectively [18]. We defined safer supply hydromorphone prescribing based on a definition we developed for a prior study [7]. Specifically, we classified hydromorphone prescribing as consistent with SSHM if the following criteria were met: 1) patients were dispensed doses totaling at least 32 mg per day (using 4 mg or 8 mg oral formulations) for at least 2 of 3 consecutive days; and 2) prescriptions were dispensed daily. We set the index date as the first date of SSHM prescribing. We classified individuals as having opioid use disorder if they met at least one of the following criteria: 1) used opioid use disorder-specific health services in the 2 years prior to the index date; 2) prescription for methadone or buprenorphine in the 4 years before the index date; and 3) opioid toxicity treated in an emergency department or hospital within 2 years before the index date (S1 Table). Individuals were excluded if they were not Ontario residents, had invalid or missing age or sex data, or died on or before the first day of the first study period. We also excluded individuals who had a cancer diagnosis or cancer-related treatment in the 1 year before the index date (S1 Table) because of the potential for hydromorphone prescribed for pain to be misclassified as SSHM. Each individual contributed only one dispensation to the study and were included at their first instance of meeting the cohort definition chronologically during the study period.

Analysis

We summarized baseline sociodemographic and health-related characteristics of individuals prescribed SSHM using descriptive statistics. Individuals initiating safer supply hydromorphone in the pre- versus pandemic periods were compared using Pearson’s chi-squared test for binary measures and Wilcoxon rank sum test for continuous measures.

We counted new dispensations of SSHM in Ontario in successive 28-day periods over the study period. In addition to analyzing new dispensations overall, we performed subgroup analyses by male and female sex (based on sex documented at birth). In periods with 1 to 5 dispensations inclusively, the number was suppressed in accordance with ICES privacy policies. We assigned these months a random integer value within this range with equal probability.

We analyzed these data as an interrupted time series of counts using segmented regression methods and tested for both a change in level and a change in trend related to the COVID-19 pandemic [19, 20].

We addressed three considerations in specifying the regression model. First, we assessed the model for seasonality after visual inspection of trends over time suggested periodic trends in the data. We entered Fourier term coefficients into the model, which allow for modeling period fluctuations as harmonic (sine and cosine) terms [21]. We decided on the number of harmonics by testing model fit after entering one to three pairs of terms and assessing fit using Bayesian Information Criteria (BIC). We used a monthly frequency based on visual inspection of trends over time, a periodogram, and clinical judgment [22]. Second, we assessed two models for over-dispersed count data using the BIC: a Poisson model with robust variance estimators and a negative binomial model [23]. Third, we assessed models for residual autocorrelation through visual inspection of autocorrelation and partial autocorrelation plots [20]. We evaluated overall model fit through both visual inspection of graphs comparing observed to fitted values and graphs of Pearson residuals over time.

We used counterfactual analysis methods to estimate the number of new dispensations attributable to the pandemic. To do so, we used predictive margins to estimate the number of new dispensations in each period within our analysis after March 17, 2020, both with and without pandemic effects and used the delta method to estimate uncertainty bounds around the difference [24, 25].

We conducted three sensitivity analyses. First, we assessed the sensitivity of our results to the imputation of censored values for months with few dispensations by examining the stability of regression parameter estimates for pre-pandemic trends, the level change at the time the pandemic was called, and the pandemic trend over 1000 simulations. We also estimated the number of new dispensations attributable to the pandemic in each of these samples. Second, we selected a change date based on visual inspection of dispensation rates over time, which suggested that the change may have occurred later than the date the pandemic emergency was declared. A later date is consistent with the concept that alterations in dispensation, such as changed prescribing habits, may have been implemented some weeks after the declaration of emergency. Third, we examined an alternative definition for eligibility criteria. Our primary analysis included individuals who were dispensed immediate release hydromorphone prior to their index date because it is known that people with opioid use disorder are sometimes prescribed short acting opioids outside of safer supply prescribing before starting SSHM [7, 26]. In this sensitivity analysis, we excluded individuals who were dispensed 4 or 8 mg tablets of hydromorphone in the 14 days before the index date, which we anticipate was a more specific (but less sensitive) algorithm.

All reported p-values are two-sided with a type I error rate of 5%.

Primary analysis

We identified 1489 initiations of SSHM during the study period, of which 434 (8 per 28-day period) were before and 1055 (56 per 28-day period) after the declaration of the pandemic. The median age was 40 (interquartile range [IQR] 33–48), and individuals initiated on SSHM prior to the pandemic were older than those initiated during the pandemic (median age 42 [IQR 34–50] versus median age 39 [IQR 33–47], respectively; p = <0.001) (Table 1). Overall, most individuals (61.7%; N = 919) were male and more than half (52.9%; N = 788) were in the lowest income quintile, both of which did not differ significantly between the two periods. Individuals initiated on SSHM during the pandemic, compared with individuals initiated before the pandemic, were less likely to have a diagnosis of HIV (5.5% versus 14.1%, p = <0.001); have had an infective complication from infective endocarditis, osteomyelitis, discitis, or skin and soft tissue infection (13.6% versus 26.5%, p = <0.001) (S1 Table); or have received palliative care services in the prior 180 days (≤0.3% versus 13.4%, p = <0.001) (S1 Table); and were more likely to have had an opioid-related overdose in the prior 1 year (29.6% versus 14.1%, p = <0.001) (S1 Table); have been hospitalized for a substance-related disorder in the prior 3 years (52.9% versus 42.2%, p<0.001); or have been prescribed any opioid agonist therapy in the prior one year (82.6% versus 67.7%, p = <0.001).

Download:

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

Table 1. Baseline characteristics of patients initiating safer supply immediate release hydromorphone, stratified by the pre-pandemic (March 22, 2016 –March 16, 2020) and pandemic (March 17, 2020 –August 30, 2021) periods.

https://doi.org/10.1371/journal.pone.0295145.t001

The number of new SSHM dispensations per 28-day period remained low until October 2018 (≤10), after which it began to increase slightly. The peak number of dispensations per 28-day period prior to the pandemic occurred in September, 2019 at 32. The rate increased further after the declaration of the pandemic in March 2020 (Fig 1). The peak number of new dispensations occurred between August 4 and November 23, 2020, with an average of 83 initiations per 28-day period. The best-fitting model was a Poisson regression which included 2 pairs of Fourier terms; this model had no residual autocorrelation and acceptable model fit. Prior to the pandemic, there was a small increase in rate of new SSHM (incidence rate ratio [IRR] 1.002 per 28-day period; 95% confidence interval [95CI] 1.001 to 1.002; p<0.001) (Table 2); this corresponds to an increase of about 2.6% (1.002^365.25/28) per year. There was an immediate increase in new SSHM dispensations at the time the pandemic was declared (relative increase in IRR 1.674; 95CI 1.206 to 2.322; p = 0.002); however, the trend during the pandemic was not statistically significantly different than the trend prior to the pandemic (relative increase in IRR per period of 1.000; 95CI 1.000 to 1.001; p = 0.251). We estimated that there were 511 (95CI 327 to 695) new SSHM dispensations (27 per 28-day period) attributable to the pandemic in the 532 days after the pandemic was declared.

Download:

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

Fig 1. New safer supply hydromorphone dispensations over time in Ontario.

Dots represent the number of safer supply hydromorphone dispensations per 28-day period between early 2016 and late 2021 in Ontario. The vertical dashed line indicates March 17, 2020, the date that Ontario declared an emergency due to the COIVD-19 pandemic. The solid blue line indicates predictions from the best-fitting Poisson regression model and light blue shading indicates confidence bounds around these estimates. The dashed red line indicates the estimated number of dispensations if the pre-pandemic trends had continued during the pandemic (i.e., the counterfactual scenario) and light red shading indicates confidence bounds around these estimates.

https://doi.org/10.1371/journal.pone.0295145.g001

Download:

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

Table 2. Primary, sex-specific, and sensitivity analyses comparing initiations of safer supply hydromorphone before and during the COVID-19 pandemic.

https://doi.org/10.1371/journal.pone.0295145.t002

Sex-specific analyses

When stratified by sex, 570 females (180 [3 per 28-day period] before and 390 [20 per 28-day period] during the pandemic) and 919 males (254 [5 per 28-day period] before and 665 [35 per 28-day period] during the pandemic) were initiated on SSHM. In a subgroup analysis by sex, the best fitting model was a Poisson model with 2 and 1 Fourier terms for males and females, respectively. The pre-pandemic trend for both sexes showed a similar small increase in prescribing per period (IRR 1.002 for both per period; 95CI 1.001 to 1.002 for males and 1.001 to 1.003 for females; p<0.001 for both). The pandemic-related level change in rate was significant for males (relative increase in IRR 1.691; 95CI 1.075 to 2.658; p = 0.023) but not for females (relative increase in IRR 1.289; 95CI 0.912 to 1.821; p = 0.15) (Fig 2). The pandemic trend was not statistically different from the pre-pandemic trend for either sex (p = 0.66 for males, p = 0.90 for females, Table 2).

Download:

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

Fig 2.

New Safer Supply Hydromorphone Dispensations over Time in Ontario for A) male sex and B) female sex. Dots represent the number of safer supply hydromorphone dispensations per 28-day period between early 2016 and late 2021 in Ontario. The vertical dashed line indicates March 17, 2020, the date that Ontario declared an emergency due to the COIVD-19 pandemic. The solid blue line indicates predictions from the best-fitting Poisson model and light blue shading indicates confidence bounds around these estimates. The dashed red line indicates the estimated number of dispensations if the pre-pandemic trends had continued during the pandemic (i.e., the counterfactual scenario) and light red shading indicates confidence bounds around these estimates.

https://doi.org/10.1371/journal.pone.0295145.g002

Sensitivity analyses

Limitations

Our study has some limitations. First, while our case definition of SSHM has face validity, has been previously published, and is based on consultation with safer supply prescribers and characteristic prescribing patterns [7], it was not assessed for criterion validity. Thus, we may have misclassified some individuals receiving immediate release hydromorphone for an indication other than safer supply. Further, individuals in our study could have received hydromorphone prior to the index date, which may have led to an inaccurate start date. It is reassuring that our results remained consistent in a sensitivity analysis using a more restrictive definition. Conversely, in an attempt to ensure specificity of our definition, we may have excluded individuals who were prescribed SSHM, thereby underestimating the number of initiations. Additionally, our case definition was not designed to include initiations that occurred within recovery and isolation hotels established in Ontario in response to the pandemic, which have particular prescribing characteristics not captured in our current definition. Prescribing of SSHM is known to have occurred within these sites but was frequently discontinued upon discharge [15, 37]. Our study therefore applies only to SSHM prescribed outside of institutional settings. We did not include an offset term in our regression model since we anticipated that the number of people with opioid use disorder in Ontario was relatively stable over the time course of our analysis and accurate assessments of the population of people who use opioids is unavailable. Finally, Ontario’s administrative health data does not permit analysis based on gender rather than sex, which is an important consideration that we could not address.