Article Text
PDF
Abstract
Objectives Describe new opioid prescription claims, their clinical indications and annual trends among opioid naïve adults covered by the Quebec’s public drug insurance plan (QPDIP) for the fiscal years 2006/2007–2019/2020.
Design and setting A retrospective observational study was conducted using data collected between 2006/2007 and 2019/2020 within the Quebec Integrated Chronic Disease Surveillance System, a linkage administrative data.
Participants A cohort of opioid naïve adults and new opioid users was created for each study year (median number=2 263 380 and 168 183, respectively, over study period).
Intervention No.
Main outcome measure and analyses A new opioid prescription was defined as the first opioid prescription claimed by an opioid naïve adult during a given fiscal year. The annual incidence proportion for each year was then calculated and standardised for age. A hierarchical algorithm was built to identify the most likely clinical indication for this prescription. Descriptive and trend analyses were performed.
Results There was a 1.7% decrease of age-standardised annual incidence proportion during the study period, from 7.5% in 2006/2007 to 5.8% in 2019/2020. The decrease was highest after 2016/2017, reaching 5.5% annual percentage change. Median daily dose and days’ supply decreased from 27 to 25 morphine milligram equivalent/day and from 5 to 4 days between 2006/2007 and 2019/2020, respectively. Between 2006/2007 and 2019/2020, these prescriptions’ most likely clinical indications increased for cancer pain from 34% to 48%, for surgical pain from 31% to 36% and for dental pain from 9% to 11%. Inversely, the musculoskeletal pain decreased from 13% to 2%. There was good consistency between the clinical indications identified by the algorithm and prescriber’s specialty or user’s characteristics.
Conclusions New opioid prescription claims (incidence, dose and days’ supply) decreased slightly over the last 14 years among QPDIP enrollees, especially after 2016/2017. Non-surgical and non-cancer pain became less common as their clinical indication.
- EPIDEMIOLOGY
- PREVENTIVE MEDICINE
- PUBLIC HEALTH
- Chronic Pain
- Prescriptions
Data availability statement
Data are available on reasonable request. The Quebec National Public Health Institute ('INSPQ') is committed to different partners (Ministry of Health and Social Services (MSSS), Régie de l’Assurance Maladie du Québec (RAMQ) and the Health Information Access Commission) to use the Quebec Integrated Chronic Disease Surveillance System du SISMACQ for chronic disease surveillance, to guarantee the security and protection of personals information put in his disposal. Thus, their use in this project was subject to information management framework and stringent rules of confidentiality. The individual data access is not possible outside INSPQ’s offices. However, aggregated data and research results following diffusion rules would be accessible. Statistical programs allowing the indicator creation will be open access.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
STRENGTHS AND LIMITATIONS OF THIS STUDY
This study included all opioid naive adults of Quebec covered by Quebec Public Drug Insurance Plan with new opioid prescription claims and developed algorithm identifying their clinical indication with good consistency.
Because it is administrative data, there could be information bias on new opioid prescription claims and their clinical indication outcomes.
Our results can only be generalised to similar populations (65 years and over, excluding private insurance enrollees, out-of-pocket payers, militaries, prisoners and indigenous people).
Introduction
Several factors created an ‘opioid crisis’ in North America.1 It started with increased deaths due to opioid prescriptions in 1999 in the USA2 3 and Canada4 and continued with deaths due to illicit opioid use in recent year.2 Total apparent opioid toxicity deaths increased from 3.2 to 6.3 per 100 000 persons in Quebec between 2016 and 2022.5
Prescribed opioids are universally recognised as playing a major role in the opioid crisis because they have been associated to opioid diversion, heroin initiation and opioid overdoses.1 6–17 Since the beginning of opioid crisis, several efforts (politics, educational campaigns, guidelines and prescription monitoring programmes) have been developed to reduce inappropriate opioid prescribing in North America.18–25 One of the mainly goals of these efforts was the reduction of opioid initiation for chronic non-cancer pain and opioid overprescribing for acute pain (eg, surgical pain and dental pain). In 2015–2016, a training support for opioid prescribing was implemented among medical students and physicians in Quebec.26 Several studies elsewhere27–30 have reported a decreasing trend of new opioid prescriptions after such interventions. In Quebec, only one population study was conducted ; however, it was limited to the preintervention period (2006–2016) and did not identify the clinical indication of the opioids prescribed.31 Thus, more research is needed to understand what happened to opioid prescribing in Quebec after these efforts (especially after 2016), including for which clinical indications they are prescribed.
The administrative database of the Quebec Integrated Chronic Disease Surveillance System (QICDSS), which includes a field for pharmaceutical services for Quebec’s public drug insurance plan (QPDIP) beneficiaries, allows the surveillance of opioid prescription claims among specific population subgroups (≥65 years and <65 years with some criteria).32 33 Administrative databases are more often employed to monitor the opioid prescriptions than data from patients’34 or prescribers’ surveys35–39 since they have important advantages such as access to sociodemographic and clinical characteristics and, long follow-up periods of opioid prescriptions among large populations at comparatively low cost. Recent administrative data studies elsewhere suggested a methodological approach based on a hierarchical algorithm27 40–43 to identify the most likely clinical indications of new opioid prescriptions. The QICDSS (included QPDIP database), a linked administrative database,32 can be used with a similar approach to understand for which clinical indications, opioids are prescribed.
This study aimed to (1) present the trend of the annual incidence proportion of new opioid prescription claims among opioid naïve adults registered in the QPDIP from 2006/2007 to 2019/2020; (2) describe the new opioid prescription claims (users’ characteristics, dose, day supply and prescribers’ specialty) and (3) build a hierarchical algorithm to identify the most likely clinical indications of new opioid prescription claims.
Methods
The following sections are presented according to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.44
Data source
Data came from the QICDSS.32 The QICDSS is used for public health surveillance of chronic diseases. It was created by linking the following healthcare administrative databases: the health insurance registry (FIPA), the hospitalisation discharge database (Med-Echo), the medical and pharmaceutical services databases and vital statistics. The linkage was made by a unique anonymous identification number related to the universal Quebec healthcare insurance «RAMQ» (≥99% linkage success). The FIPA database contains sociodemographic records, as well as data about the admissibility of individuals to health and drug insurance plans. The Med-Echo database includes information on all acute care hospitalisations and day surgeries that happen in the province of Quebec. The medical services database includes information about fee-for-service billings, that is, the payment claims submitted by health professionals, mainly physicians, to RAMQ. The pharmaceutical services database includes information about all drugs delivered by retail pharmacies under the public drug insurance plan (QPDIP) and covers 90% of individuals aged ≥65 and approximately 37% of those aged 18–64 years. Data on drugs not covered by the QPDIP or related to private insurance drug plans are not included in the pharmaceutical services database.
The creation of the QICDSS was approved by government bodies in legal possession of the databases.
The patient and public involvement
None.
Study design
We conducted a retrospective cohort study among adult residents of the Province of Quebec aged 18 years and over who were covered by the QPDIP between 2006/2007 and 2019/2020. A cohort of opioid naïve adults was created for each fiscal year of the study period and among these, all new opioid users were identified (detail in further section : outcomes).
Study samples
The cohort for each fiscal year was selected according to the following criteria: (1) adults aged 18 years and older; (2) not living in a nursing home or not hospitalised during all of the fiscal year; (3) registered with the RAMQ and QPDIP on 1 April of the fiscal year and during the previous 730 days and (4) opioid naïve during the previous 730 days. Adults covered by the QPDIP were considered opioid naïve, if they did not have (a) an opioid prescription in the pharmaceutical services database or (b) an opioid addiction or overdose diagnostic code in the hospitalisation discharge database or the medical services databases (online supplemental tables 1 and 2) during the previous 730 days. The look-back period of 730 days was chosen because it was the period often used in prior research and considered sufficient in the literature to define an opioid-naïve population.40 43 45 46 This was a purposive population study of Quebec’s general population; thus, sample size was not estimated before the study was conducted. There was no a priori power analysis performed.
Supplemental material
Online supplemental figure 1 presents the flow chart of the study and the median number of opioid naïve adults included each year over the study period.
Outcomes
New opioid prescription claims or users
All opioid drugs sought were those approved by Health Canada47 and filled by opioid naïve adults in retail pharmacies (online supplemental table 1). For each fiscal year between 2006/2007 and 2019/2020, new opioid users were the opioid naïve adults enrolled in QPDIP on 1 April of that fiscal year, who had claimed a prescription for an opioid drug during the fiscal year. The index date was the date of the first opioid prescription filling.
Hierarchical algorithm for the identification of the most likely clinical indications
Prior to the development of the hierarchical algorithm, the categories of clinical indications were chosen according to previous research.27 40–42 The healthcare information related to their identification in administrative databases was defined by the Canadian Classification of Intervention codes or 10th version of the International Classification of Diseases (ICD-10) diagnosis codes in hospitalisation database; or Current Procedural Terminology codes or ICD-9/10 diagnosis codes in medical services database (online supplemental tables 3 and 4 and online supplemental material codes). The categories were defined as follows: surgical pain, obstetrical pain, cancer pain, traumatic pain, dental pain, musculoskeletal pain, neurological pain, abdominal/pelvic pain, upper respiratory diseases, lower respiratory diseases, cardiovascular system diseases, ocular pain, cutaneous pain, other infectious diseases and opioid dependence.
Supplemental material
Since new opioid users could have more than one medical service and healthcare information related to these categories in the administrative databases of the QICDSS, a six-step hierarchical algorithm was developed to define the category with the highest evidence as the most likely clinical indication for a new opioid prescription. The first two steps used opioid prescription claims’ information to classify new opioid users with a buprenorphine-naloxone prescription (step 1) and those with a dentist prescriber (step 2) in the opioid dependence and dental pain categories, respectively. After that, the hierarchical algorithm used healthcare information related to each clinical category as defined above (online supplemental tables 3 and 4 and online supplemental material codes) over a time window close to the index date to assign the most likely clinical indication. The time window was framed for 120 days before the index date and 30 days after. This definition was chosen according to the literature27 41 43 48 and the distribution of the time between all dates of visits available from new opioid users and their index date (online supplemental figures 2–4). Therefore, on step 3, all new opioid users with healthcare information related to only one clinical indication category in the time window were directly classified into this indication. In step 4, the algorithm classified the new opioid users not already classified and having healthcare information related to two or more clinical indication categories in the time window. To do so, a rank was first assigned to each clinical indication category according to the likelihood of opioid prescribing and other ranking found in previous research27 40 43 (online supplemental table 5). Based on this, the algorithm used a step-by-step approach that began with the classification of new opioid users with healthcare information related to the highest-ranking category, that is, those with surgical pain (rank 1). Among them, if the date of the visit of the highest-ranking category (surgical pain) was nearest to the index date, this category was chosen as the most likely clinical indication. If there was another category with a visit date nearest to the index date, this other category was chosen as the most likely clinical indication only if that situation occurred at least 20% of time among cases who had the highest-ranking category (rank 1). Otherwise, we considered that the other category was not important enough when there was surgical pain occurring and surgical pain was chosen as the likely clinical indication. The algorithm pursued this approach with another step aiming to classify all new opioid users who had the next highest category (ie, rank 2: obstetrical pain) until all remaining categories were classified at step 4. Finally, for the new opioid users unclassified at previous step, a supplementary time window was added (120th–365th days before the index date) and the algorithm repeated the steps 3 and 4 to assign the most likely clinical indication in steps 5 and 6. Details on all steps of the hierarchical algorithm are presented in online supplemental figure 5.
Covariables
The following variables were included in descriptive analyses: age, sex, health region,49 area of residence,50 social and Material Deprivation Index51 and a comorbidity index calculated from the adapted model of Charlson and Elixhauser (0, 1, ≥2 comorbidities).52 The prescriber’s specialty was categorised according to the Canadian health professional categories (ie, Quebec’s physician (specialties specified) and other health professionals). For each new opioid prescription claim, the days supply and the daily dose were calculated. Methadone and buprenorphine/naloxone were excluded from these analyses because they are usually prescribed to treat opioid dependence. Pharmaceutical forms other than tablets, capsules and patches were also excluded from these analyses because it was difficult to estimate their dose/unit and they represented less than 4% of all new opioid prescriptions. The number of days supplied was directly entered into the database by the pharmacist. The daily dose (morphine milligram equivalent, MME/day) was obtained by the following formula: strength per unit × (number of units delivered/number of days’ supply)×MME conversion factor.53 54
Statistical analysis
Continuous variables were described by mean and SD or median with first quartile (Q1) and third quartile (Q3). Categorial variables were described by the absolute number and relative frequency. A comparison of characteristics (age, sex, social and material deprivation, census area and comorbidity index) was done between opioid naïve adults and new opioid users. The frequency of new opioid prescription claims was also described by prescriber specialty and their most likely clinical indications. For clinical indication categories, a post hoc category ‘others’ was created by the combination of categories with low frequencies (≤1%) and which were less common indications of opioid prescribing according to the literature55 (ie, ocular pain, other infectious diseases, cutaneous diseases, upper and lower respiratory diseases, cardiovascular diseases and opioid dependence). For days’ supply and daily dose, aberrant data (ie, values of 0 or which were two times higher than the 99th percentile) or missing data were excluded from the analysis.
The annual incidence proportion of new opioid prescriptions was obtained by the following formula:
The CIs were calculated at 95%. For each fiscal year, the age-standardised annual incidence proportion was calculated using direct standardisation and the age structure of the Quebec population in 2011. For trend analysis purpose, each fiscal year was considered as a continuous variable (eg, 2006/2007=2006). Time trend analysis was performed by using the joinpoint regression method.56 57 The joinpoint regression model estimated the average annual per cent change (AAPC) and the APC of trend for all study periods and for segment periods, respectively.57 When a break (deviation) occurred in trend during the study period, a simple predictive linear regression was performed to compute the expected age-standardised annual incidence proportion for each fiscal year after the break time point (details in online supplemental materials). The age-standardised annual incidence proportion (median of all fiscal years) of a region (categorised as quartile class) was also depicted according to the Quebec health regions using one of the R mapping package.58 It depicted with population density with each health region to describe the potential correlation between the incidence and population density. The age-standardised incidence proportion of northern health regions were not depicted because of high risk of information bias (lowest number of QPDIP enrollees). The time trend of age-standardised annual incidence proportions was also stratified by age (<65 years and ≥65 years) and sex (women and men). This age stratification was performed because QPDIP enrollees ≥65 years of age were representative of the population over 65 years in Quebec, but QPDIP enrollees under 65 of age represented only 37% of the population under 65 years in Quebec, and were more deprived and had more comorbidities than the general population under age 65.
Finally, cross-tabulation of the most likely clinical indications with prescriber specialties (all fiscal years) and users’ characteristics (2014/2015 and 2019/2020) was done. These descriptive comparisons were performed to gain a qualitative appreciation of the consistency between the clinical indications identified by the hierarchical algorithm and by prescriber specialties or users’ characteristics. The appreciation was qualitative because it was based on common knowledge about these clinical indications (eg, cancer pain patients were expected to be older and cancer pain was expected to be the main clinical indication of an opioid prescription by an oncologist). All statistical analyses were performed with the SAS software V.9.4, Microsoft Excel 2019, Joinpoint Software V.5.0.2 and R software V.4.3.2.
Results
Overall, 2 337 136 new opioid prescriptions were claimed from retail pharmacies by opioid naïve adults enrolled in QPDIP over the study period corresponding to a median number per year of 168 183 new opioid prescriptions. Table 1 shows the comparison of characteristics between opioid naïve adults and new opioid users for fiscal years 2006/2007, 2014/2015 and 2019/2020. New opioid users were older (61.4±18.4 vs 57.2±19.5 in 2019/2020), slightly more materially deprived (25% vs 24% in 2019/2020), frequently lived in rural area (25% vs 21% in 2019/2020) and had more comorbidities (35.5% vs 18% had over 2 comorbidities in 2019/2020) than opioid naïve adults.
Description of the characteristics of opioid naïve adults and new opioid users for the fiscal years 2006/2007, 2014/2015 and 2019/2020
The absolute number of new opioid prescriptions claimed by opioid naïve adults increased slightly during the study period from 158 853 prescriptions in 2006/2007 to 1 63 511 prescriptions in 2019/2020. The average crude annual incidence proportion of new opioid prescription claims through the study period was 7.45% (95% CI 7.41% to 7.48%) (online supplemental figure 6). The age-standardised incidence proportion of new opioid users decreased (−1.7%) over the study period, from 7.53% (95% CI 7.51% to 7.55%) in 2006/2007 to 5.78% (95% CI 5.76% to 5.80%) in 2019/2020 (figure 1). The average annual percentage change was −1.8% (95% CI −2.4% to −1.2%), p<0.05. A break time point occurred at 2016/2017. An important decrease was observed between 2016/2017 and 2019/2020 (APC=−5.5%–95% CI −8.0% to −3.0%, p <0.05). This deviation created a −1% difference between the expected and the observed age-standardised annual incidence proportion in 2019/2020. The age-standardised annual incidence proportion was higher among opioid naïve adults up to 65 years (9.6%–8.6% among men and 8.5%–7.9% among women) than among those below 65 years (6.6%–4.6% among men and 7.8%–5.6% among women) during the study period. The decrease was stronger among those below 65 years during all of the study period (online supplemental figure 7). Health regions with lowest population density had the highest age-standardised annual incidence proportion of new opioid prescription claims (figure 2).
Time trend of age-standardised annual incidence proportion of new opioid prescription claims between 2006/2007 and 2019/2020 using joinpoint regression method. AAPC, average APC; APC, annual per cent change.
Mapping of age-standardised annual incidence proportion of new opioid prescription claims (median of all fiscal years) by Quebec’s health region (label on region maps=number assigned to the region/population density). Number assigned to the region=01-Bas-Saint-Laurent; 02-Saguenay–Lac-Saint-Jean; 03-Capitale-Nationale; 04-Mauricie et Centre-du-Québec; 05-Estrie; 06-Montréal; 07-Outaouais; 08-Abitibi-Témiscamingue; 09-Côte-Nord; 10-Nord-du-Québec; (data not shown) ;11-Gaspésie–Îles-de-la-Madeleine; 12-Chaudière-Appalaches; 13-Laval; 14-Lanaudière; 15-Laurentides; 16-Montérégie; 17-Nunavik (data not shown) and 18-Terres crie de-la-baie-James (data not shown). Data not shown because of high risk of information bias. Population density=number of residents per km2. Source: https://www.mamh.gouv.qc.ca/organisation-municipale/cartotheque/cartes-regionales/.
Family physicians and surgeons were the most frequent prescribers, at 55.8% and 21.2%, respectively (online supplemental figure 8). The median days supply decreased from 5.0 (Q1–Q3: 3.0–7.0) days in 2006/2007 to 4.0 (Q1–Q3: 2.0–6.0) in 2019/2020(online supplemental figure 9). Finally, the median daily dose decreased from 27 (Q1–Q3: 18–40) MME/day in 2006/2007 to 25 (Q1–Q3: 18–38) MME/day in 2019/2020) (online supplemental figure 10).
The algorithm identified the most likely clinical indication category for more than 95% of new opioid users (96% of the most likely clinical indications identified were in the time window of 120 days before and 30 days after the index date). Among the 2 337 136 new opioid prescription claims, cancer pain (43.8%) and surgical pain (32.9%) were the most predominant clinical indications of the new opioid prescription claims (figure 3). Between 2006/2007 and 2019/2020, the clinical indication for new opioid prescription claims increased for cancer pain from 34% to 48%, for surgical pain from 31% to 36% and for dental pain from 9% to 11%. Inversely, the proportion of prescriptions for musculoskeletal pain decreased from 13% to 2%. Prescriptions for other categories of indications decreased also during the study period (figure 3). The most likely clinical indications identified by the algorithm showed a good consistency with the prescriber’s specialty (figure 4). Indeed, over the study period, oncologists prescribed mostly for cancer pain (75%). Surgeons, ophthalmologists and ENT specialists prescribed mostly for surgical pain (≥80%). Gynaeco-obstetricians prescribed mostly for surgical pain (65%) and obstetrical pain (24%) and rheumatologists mostly for musculoskeletal pain (54%) and cancer (30%) (figure 4). Characteristics of new opioid users by clinical indication are presented in online supplemental tables 6,7. Briefly, new opioid users with clinical indication of cancer pain, musculoskeletal pain and surgical pain were older than those with other clinical indications. Cancer pain, musculoskeletal pain, obstetrical pain, abdominal pain and neurological pain are more often prescribed in women. New opioid users with dental pain and obstetrical pain were more deprived socially and materially. The median daily dose was higher for new opioid users with clinical indications of surgical pain and traumatic pain.
Frequency of the most likely clinical indication of new opioid prescription claims according to the hierarchical algorithm in each fiscal year (2006/2007–2019/2020) and overall.
Frequency of the most likely clinical indication of new opioid prescription claims according to prescriber specialty (all fiscal years N (number)=2 337 136 new opioid prescription claims). Others professionals=others medical specialist in Quebec (cardiologist, pneumologist, hepatogastroenterologist…) and physician outside Quebec. Others=upper respiratory diseases, lower respiratory diseases, cardiovascular diseases, ocular pain, cutaneous pain, infectious diseases and dependence. ENT, ear, nose and throat; NA, not available.
Discussion
We conducted this study to describe new opioid prescription claims and their clinical indications among opioid naïve adults covered by the QPDIP. We found that new opioid prescription claims (annual incidence proportion, daily dose and days supply) decreased, although slightly (drop of 1.7%, 1 day and 2 MME/day, respectively) over the study period. However, the absolute number of new opioid prescription claims remained quite similar between the beginning and end of the study period. The largest decrease in incidence was observed after 2016/2017, reaching 6% drop of the incidence of the previous year at each one unit increase of fiscal year. But, opioid naïve adults living in rural or aged over 65 years still had the highest annual incidence proportion of new opioid prescription claims during the study period. Family physicians were the principal prescriber. The hierarchical algorithm, we developed identified the most likely clinical indication for the majority of new opioid prescription claims. The most common clinical indications were cancer pain and surgical pain. Between 2006/2007 and 2019/2020, the clinical indication of new opioid prescription claims increased for cancer pain from 34% to 48%, for surgical pain from 31% to 36% and for dental pain from 9% to 11%. Inversely, musculoskeletal pain decreased from 13% to 2%. The clinical indications identified by the hierarchical algorithm show good consistency with the prescriber’s specialty and new users’ characteristics.
The annual incidence proportion of new opioid prescription claims in this study (7.5% in 2006 and 5.8% in 2019) was lower than in the western provinces of Canada (9.5% in 2013 and 8.6% in 2018).30 It was also lower than among the US public drug insurance plan enrollees (15% in 2006 to 13.1% in 2018 among Medicare enrollees) and among the commercial insurance plan enrollees in USA (10.3% in 2006 to 9.7% in 2018 in the Optum database and 8.6% in 2006 to 7.6% in 2018 in the IBM MarketScan Commercial Database).28 His decrease observed in this study was also observed elsewhere.27–30 But, it seemed lower than among US public insurance plan enrollees, especially after 2016.28 The median daily dose (27 MME/day in 2006 to 25 MME/day in 2019) and days supply (5 days in 2006 to 4 days in 2019) were smaller than in remained North America (over 35 MME/day and over 6.0 days); but they decreased in same manner (only slightly).28–30 This decrease of opioid initiation (incidence, dose and days supply) in Quebec and elsewhere might be explained by the impact of several efforts to reduce it,18–25 57 especially after 2016.20 26 59 The difference in opioid initiation patterns observed between this study in Quebec and other jurisdictions could be explained by the difference of healthcare system and study population between them.
Although, overall, we found some similarities for the common clinical indications with other studies in North America.40 41 43 There was an important difference in the frequency of clinical indications found. Other studies reported higher proportion of dental pain (23%), musculoskeletal pain (10%–12%) and traumatic pain (6%–11%) and low proportion of surgical pain (17%–20%) and cancer pain (6.5%) than in this study. These differences may be explained by our study sample that had a higher mean age and more comorbidities than in these studies.40 41 Also, the hierarchical algorithm we developed used different category definitions of clinical indications and a different conceptual framework than the other studies.40 41 However, like other studies observed it,40 41 we found that new opioid users’ characteristics were consistent with each clinical indication categories identified by our algorithm. Additionally, we found that clinical indication categories identified by our algorithm were consistent with prescriber specialty.
This study has some important strengths. Our study sample of people covered by QPDIP was large and representative of the Quebec population aged 65 years and older. Clinical indication categories were defined by the best approach available for this kind of research (according to the literature and to expert opinion). They were relatively exhaustive given the healthcare information used as available, and mutually exclusive, because the algorithm is hierarchical. The algorithm integrated concepts from multiple, priorly published algorithms and built on those. There was good qualitative consistency of the clinical indications found with prescriber specialties and user characteristics.
This study is subject to several limitations. First, the study was based on secondary, administrative data and, therefore, subject to information biases. An opioid might be prescribed but not registered in our database (private insurance, pay-out-of-pocket, not filled at a pharmacy, etc). This could have caused an underestimation of the annual incidence proportion. It is also possible that some administrative codes were missing or incorrect. Another limitation is the use of the closest visit and the more probable clinical indication during a longer time to assign the most likely clinical indication to the new opioid prescription, rather than a shorter time window, as used in other studies.40 43 Although this reduced the percentage of unclassified indications, it could have increased the possibility that an incorrect indication was assigned. However, the hierarchical algorithm we developed, classified up to 96% of clinical indication found in a short time window (120 days before and 30 days after the index date). Moreover, our algorithm has not been validated. Its consequent use by other research teams elsewhere may be challenging to apply. Dose and days’ supply described in this study were only based on the first prescription filled and might not be representative of dose and days’ supply based on all prescriptions used during the first episode of opioid initiation. This study did not quantify trends in the prevalence of opioid use, which is important for understanding the potential impact of opioid use on the whole population. Finally, we cannot generalise our results to the complete population of Quebec, particularly the population under 65 years. Because, private insurance enrollees, out-of-pocket payers, militaries, prisoners and indigenous people were not included in our study population.
In Quebec QPDP enrollees, opioid initiation, especially for non-cancer and non-surgical pain decreased importantly through the past 14 years. This highlights the potential effectiveness of guidelines, laws and campaigns about an opioid initiation. The lowest incidence and dose of opioid initiation in Quebec than other jurisdictions can explain the lowest and later burden of opioid overdose crisis in Quebec than elsewhere. But, this decrease and lowest rate of opioid initiation can reversely motivate the illicit opioid seeking by some patients in need of strong pain reliever.60 This study suggests that administrative data can help to monitor opioid initiation and their clinical indications. In a perspective of optimising opioid prescribing, a patient’s first-opioid prescription and its clinical indication are critically important. This information is essential in order to improve prioritisation of some clinical indications or specific prescriber specialties during the promotion of appropriate opioid prescribing, and to consequently evaluate this prioritisation. Conducted routinely, our study design could provide results helping public health policymakers and prescribers to recognise the impact of their policies or practices on opioid initiation overall and across clinical indications. Further on, they could identify which clinical indications are more challenging for an appropriate opioid initiation. Finally, these results may support the implementation of new opioid prescribing policies or guidelines. However, further studies will be necessary to clinically validate the algorithm developed and to explore inappropriate opioid prescribing across different clinical indications.
Conclusion
This study suggests that opioid initiation (incidence, dose and duration), especially for non-cancer and non-surgical pain decreased in the past 14 years among QPDIP enrollees. It makes an important contribution by showing the feasibility of identifying the clinical indication of a new opioid prescription using administrative databases. These findings will allow further studies to validate the algorithm and eventually to investigate inappropriate prescribing by different clinical indications.
Data availability statement
Data are available on reasonable request. The Quebec National Public Health Institute ('INSPQ') is committed to different partners (Ministry of Health and Social Services (MSSS), Régie de l’Assurance Maladie du Québec (RAMQ) and the Health Information Access Commission) to use the Quebec Integrated Chronic Disease Surveillance System du SISMACQ for chronic disease surveillance, to guarantee the security and protection of personals information put in his disposal. Thus, their use in this project was subject to information management framework and stringent rules of confidentiality. The individual data access is not possible outside INSPQ’s offices. However, aggregated data and research results following diffusion rules would be accessible. Statistical programs allowing the indicator creation will be open access.
Ethics statements
Patient consent for publication
Ethics approval
The current study was conducted as a part of the Quebec ministerial plan for the surveillance of multiple topics that received its approval by the Quebec Public Health Ethics Review Board (ISBN: 978-2-550-58576-3). Participants do not need to give informed consent.
References
Footnotes
X @AttissoEug62204, @EhlKroger
Contributors EA wrote the research protocol, conducted the literature review and statistical analysis. He wrote this manuscript and inputted all comments for coauthors and reviewers. LG gave advices on study designing and outcomes measurement. She assisted the manuscript writing. CED gave advices on study designing and outcomes measurement. He assisted the manuscript writing. EK gave advices on study designing and outcomes measurement. She assisted the manuscript writing. SJ was the principal investigator of this project. She planned this project (choice of the topic of interest, research questions, the goals of this project, the study design and outcomes definitions) and assisted the manuscript writing. SJ is the guarantor of this study.
Funding This work (Project 1-13) was funded by Health Canada’s Substance Use and Addiction Programme.
Disclaimer There was independence between the funder and the authors of this manuscript. The funder was not involved at any step of this study. Grant number was not applicable.
Competing interests None declared.
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.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.