Time trends and prescribing patterns of opioid drugs in UK primary care patients with non-cancer pain: A retrospective cohort study
Abstract
Citation: Jani M, Birlie Yimer B, Sheppard T, Lunt M, Dixon WG (2020) Time trends and prescribing patterns of opioid drugs in UK primary care patients with non-cancer pain: A retrospective cohort study. PLoS Med 17(10):
e1003270.
https://doi.org/10.1371/journal.pmed.1003270
Academic Editor: Zirui Song, Massachusetts General Hospital, UNITED STATES
Received: April 20, 2020; Accepted: September 11, 2020; Published: October 15, 2020
Copyright: © 2020 Jani et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The data used for this paper are available through The Clinical Practice Research Datalink (CPRD) (https://www.cprd.com/, contact for data queries: [email protected]) for researchers who meet criteria for access to confidential data.
Funding: The authors received no specific funding for this work. The work is supported by the Versus Arthritis Centre for Epidemiology, the authors’ host institution (grant number 20380; WGD Principal Investigator). MJ’s work was supported by an NIHR academic clinical lecturership and a Presidential Fellowship. The funders listed had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: MJ is a member of the Medicines and Healthcare products Regulatory Agency (MHRA) Opioids Expert Working Group. WGD has received consultancy fees from Google and Bayer, unrelated to this work.
Abbreviations:
aOR,
adjusted odds ratio; CDC,
Centers for Disease Control and Prevention; CPRD,
Clinical Practice Research Datalink; GP,
general practitioner; MMI,
morphine milligram equivalents; NHS,
National Health Service; OR,
odds ratio
Introduction
The sharp increase in prescription opioid use for non-malignant pain in the US, Canada, and several European countries [1–3] has led to concerns of a similar epidemic in the UK. Opioids have now become the leading cause of accidental death and unintentional injury in the US [4]. In the UK, opioid-related deaths have been increasing over the last few decades, the majority of which are non-intentional [3,5,6]. Alongside this, a rise in opioid prescribing, based on national population-level prescribing datasets, has been reported (for all indications including cancer) [7,8]. A recent Public Health England analysis revealed 13% of the UK adult population had 1 or more prescriptions of opioids dispensed between 2017 and 2018 [9].
Opioids are associated with several serious adverse outcomes that are believed to be dose and potency dependent [10]. The escalation rate to higher doses and more potent opioids is likely to also contribute to long-term prescriptions, which in turn may be associated with opioid dependence, addiction, and overdose [11]. Until recently, several commonly prescribed opioids did not have a recommended maximum dose, despite minimal evidence of benefit in non-chronic pain at higher doses. This may lead to considerable variation in opioid prescribing in the context of chronic pain following initiation, including transitioning to stronger opioids, higher dose, or combination opioids, or not reducing dose in a timely manner. The longitudinal opioid pathway of patients commencing opioids for non-cancer pain, the scale of dose escalation/tapering, and ensuing long-term use remain unexplored.
Variation in opioid prescribing across UK regions has been described recently on a population level based on data from clinical commissioning groups [8]. Furthermore, physician prescribing behaviour has been described to be one of the key drivers of rising opioid use [12]. However, the influence of region, practice, and prescriber requires interpretation within their context, by accounting for individual patient characteristics. No studies to our knowledge have investigated the extent to which regions, practices, and individual general practitioners (GPs) vary in opioid prescribing, accounting for the patient (case) mix nor the implications of such variations for long-term opioid prescribing. Identification of what individual patient characteristics are associated with long-term opioid prescribing in primary care would allow prescribers to exercise vigilance and explore alternatives to opioids where appropriate in certain patient subgroups.
The study objectives were to (i) describe trends of the most commonly prescribed opioids for non-cancer pain in UK primary care over a 12-year period (2006–2017) in new users, (ii) assess the transition of morphine milligram equivalents (MME; accounting for dose, opioid type, and sequence of use) in the first 2 years after first prescription, (iii) quantify and identify risk factors for the transition from new user to long-term opioid user, and (iv) quantify the variation of long-term use attributed to region, practice, and prescriber, accounting for patient mix and chance variation.
Methods
Data source
We conducted a retrospective observational study from 1 January 2006 to 31 December 2017 using the Clinical Practice Research Datalink (CPRD), a database of anonymised UK primary care electronic health records. In the UK, most patients are registered with a GP, who are often the first point of medical contact and act as ‘gatekeepers’ in the national healthcare system. The majority of opioids in the UK therefore are prescribed in primary care. CPRD collects de-identified patient data from a network of general practices across the UK, providing a longitudinal, representative UK population health dataset. One in 5 practices in the UK contribute data to CPRD through an opt-in system (established for over 30 years). CPRD is one of the largest research databases of longitudinal primary care records in the world and contains information from >14 million registered patients. Prescriptions are recorded electronically, and clinical data including diagnoses are documented using Read codes. Only data that had undergone quality checks by CPRD and were ‘up to standard’ were used in this study. This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 STROBE Checklist).
Study population
Patients aged ≥18 years without prior cancer who were new users of opioids were identified, in order to establish an incident user cohort prescribed an opioid for non-cancer indications. A 24-month ‘wash-out’ period prior to the index date was used to identify new users. Patients with a previous history of a malignancy Read code up to 10 years prior to the index date were excluded, with the exception of non-melanoma skin cancer. Follow-up start was defined as the date of the first opioid prescription for a given individual (index date). Patients stayed in the cohort until end of follow-up, death, or they left the practice. Patients on methadone were excluded because, in the UK, it is primarily prescribed as an opioid addiction treatment and not consistently prescribed by GPs. S1 Fig describes the derivation of the cohort.
Covariates
Baseline characteristics such as age, sex, ethnicity, comorbidities that comprise the Charlson Comorbidity Index [13], smoking, and deprivation score were measured using data from the year prior to index date. Socioeconomic status was assessed using linked data for Townsend deprivation score, a composite measure of material deprivation based on UK census data [14]. Townsend deprivation score incorporates 4 variables: unemployment (as a percentage of those aged ≥16 years who are economically active), non-car ownership (as a percentage of all households), non-home ownership (as a percentage of all households), and household overcrowding.
To identify risk factors for long-term opioid use, we identified additional a priori variables, based on clinical knowledge and published literature. All diagnoses were identified using Read codes 1 year prior to first opioid prescription. The US Centers for Disease Control and Prevention (CDC) has identified certain factors associated with opioid misuse such as previous substance use disorder, major depression, and use of psychotropic medications [15], which we defined in CPRD. Psychotropic medications included antiepileptics, antihistamines, antiparkinsons, antipsychotics, anxiolytics, hypnotics, and sedatives. Other factors evaluated included prior history of suicide or self-harm, alcohol excess, major surgery in the last 1 year, and pain conditions such as back pain, migraine, and fibromyalgia. Rheumatological disorders included rheumatoid arthritis, systemic lupus erythematosus, myositis, and giant cell arteritis (defined by Charlson Comorbidity Index score [13]). The use of psychotropic medications, benzodiazepines, and gabapentinoids was defined as any use in the 1 year prior to the index date, including the date the first opioid was prescribed. Prescriber, general practice, and regional information for each patient was obtained to examine variation at each level in opioid prescribing.
Opioid drug preparation and exposure
Opioid exposure data were prepared using a drug preparation algorithm published previously [16]. The decisions made to prepare the data are described in S2 Fig. Classes of opioids were divided into weak opioids (codeine, dihydrocodeine, meptazinol), moderate opioids (tramadol, tapentadol), and strong opioids (morphine, oxycodone, fentanyl, buprenorphine, diamorphine, hydromorphone, pethidine). Tramadol and tapentadol were classed as moderate-strength opioids, as despite their low MME they are phenotypically distinct from conventional weak opioids due to their dual mechanism as a partial serotonin-norepinephrine reuptake inhibitor. Combination formulations, such as co-codamol, were classed according to their active opioid ingredient. If patients were on ≥1 opioid at index, we categorised them into a separate combination opioids group.
To allow direct comparison of doses and opioid potencies across different drugs and formulations we calculated MME for each prescription. MME/day was defined as the daily dose for each prescription multiplied by the equivalent analgesic ratio as specified by the CDC [15]. For transdermal buprenorphine and fentanyl formulations, strength per hour and the duration of delivery rate of the formulation was considered in the dose calculation to avoid underestimation of daily MME. An episode of long-term opioid use was defined as at least 3 opioid prescriptions issued within a 90-day period, or ≥1 opioid prescription lasting at least 90 days, in the first year of follow-up, not including the first 30 days after the index date. When defining long-term use, we ignored the first 30 days following the index date to allow for acute pain treatment.
Statistical analysis
Descriptive statistics were used to assess the baseline characteristics of the cohort, stratified according to opioid strength at initiation. Although a prospective analysis plan has not been included, the objectives of this study were determined at the outset of the planned study according to unmet need in the literature/clinical relevance and were not adapted subsequently. No data-driven changes to the analyses took place after obtaining the data during the statistical analysis stage.
Prescribing trends over time.
To evaluate trends of opioid prescribing over time, the rate of prescription for each opioid drug was calculated by calendar year by dividing the number of prescriptions per year for the cohort (numerator) by the number of eligible patients registered in CPRD per year (denominator). Raw denominator numbers of patients registered were provided by CPRD in April 2018 and prepared for use (S1 Fig).
Transition of opioids over 2 years.
Patients were stratified into 4 categories according to the average MME/day in the first 6 months after index date to incorporate the type, potency, and dose of the opioid. MME categories were as follows: low, <50 MME/day; medium, 50–119 MME/day; high, 120–199 MME/day; and very high, ≥200 MME/day. For instance, a prescription of 30 mg codeine 4 times a day would equal 18 MME/day. An oxycodone prescription of 40 mg 4 times/day equates to 240 MME/day. In patients on a combination of opioids, MME was calculated for each drug, and the sum was taken as the MME/day. Stacked plots and Sankey diagrams were created to quantify visually the sequential transition of MME/day, in 6-month bands, over a 2-year time window from the index date. One plot was generated for each of the MME dosage categories derived from the first 6 months’ exposure (low, medium, high, and very high).
Transition from new user to long-term opioid user.
A multi-level random-effects logistic regression model was used to examine the association of different patient characteristics with the odds of becoming a long-term opioid user. Person-level characteristics investigated include age, sex, ethnicity, deprivation score, and the comorbidities outlined above. More details about the approach are provided in S1 Appendix. To examine opioid variation amongst prescribers, general practices, and regions after adjusting for patient case mix, we used a nested random-effects structure (i.e., prescribers nested within practices and practices nested within regions). The approach introduced by Snijders and Bosker [17] was followed to obtain the explained variation at each level of the hierarchy. Furthermore, the posterior distributions of the prescriber-, practice-, and region-level random effects were simulated using the REsim function in the merTools package [18] for the fully adjusted models. The adjusted random-effects estimates along with 95% confidence intervals were then ranked and plotted on an odds ratio (OR) scale as well as percentage value. To express the adjusted estimates as a proportion, we used the transformation described in S1 Appendix. ORs with the lower end of the 95% CI > 1 were associated with a higher risk, and ORs with the upper end of the 95% CI < 1 were associated with a lower risk, of long-term opioid use. ‘High-risk’ regions, practices, or prescribers were defined as those where the entire adjusted 95% CI lay above the population average (i.e. 1). The risk of becoming a long-term opioid user attributed to a specific practice was then plotted against the proportion of high-risk prescribers within each practice to evaluate the influence of a high-risk prescriber on the practice (S5 Fig). All analyses were performed in STATA version 14.0 and R version 3.5.0.
Results
We identified 1,968,742 new users of opioids who met our inclusion criteria, of which 88.2% were initially commenced on a weak opioid, 8.5% on a moderate opioid, 2.6% on a strong opioid, and 0.7% on combination opioids (Table 1). The highest proportion of new opioid users for the weak, moderate, and combination opioid categories were aged 35–54 years, whereas patients who were started on strong opioids were older: 31.5% of strong opioids were prescribed to patients ≥85 years (compared to 4.1% and 3.3% in the weak and moderate opioid groups, respectively). Proportionally more patients commencing strong opioids had Charlson Comorbidity Index score ≥ 4 (7.5% compared to <2% in other opioid groups). Townsend deprivation quintiles were represented in similar proportions across weak, medium, strong, and combination opioids. The proportion of patients on all types of opioids was slightly lower in the most deprived category, between 11% and 16%. The strong opioid group had the highest proportions of patients on prior benzodiazepine, gabapentinoid, or psychotropic medications (Table 1).
Population-level opioid prescribing patterns
The most commonly used opioids were codeine, dihydrocodeine, and tramadol. Over a 12-year period, 2006–2017, codeine use increased 5-fold, from 484 to 2,456 prescriptions per 10,000 population/year. Dihydrocodeine, tramadol, and fentanyl prescriptions increased between 2006 and 2012, and plateaued thereafter until end of 2017. Within the strong opioids group, oxycodone prescribing rose approximately 30-fold, from 5 to 169 prescriptions per 10,000 population/year over 12 years. Morphine prescriptions also rose considerably, from 18 to 422 prescriptions per 10,000 population/year between 2006 and 2017 (Fig 1).
Fig 2. Transition of MME dosage categories over a 2-year period from index date, stratified by dose category in the first 6 months.
Each panel represents the index daily morphine milligram equivalents (MME). Each bar represents the proportion of patients that transition to a different level of MME, stay in the index MME category, or come off treatment over the 2 years of follow-up. The MME value thresholds were chosen considering differences in recommendations between international guidelines. In the UK, the Faculty of Pain Medicine suggests harms outweigh benefits when patients exceed 120 MME/day [27].
Variation of long-term opioid use by prescriber, practice, and region
In our new user cohort, 14.6% became long-term opioid users in the first year after the index date. In the fully adjusted model, a number of individual factors were identified as being associated with a higher odds of long-term opioid use including older age, social deprivation, fibromyalgia, suicide/self-harm, excess alcohol, gabapentinoid use, psychotropic use, major surgery, and initial dose (Fig 3) (all p-values < 0.001). The strongest association was seen in those who were ≥75 years, who were 4.6 (95% CI 4.5 to 4.7, p < 0.001) times more likely to become a long-term opioid user compared to those who were <35 years.
Fig 3. Factors associated with long-term opioid use using a multi-level model accounting for clustering of individuals within prescriber, practice, and region.
CVA, cerebrovascular accident; MMI, morphine milligram equivalents.
Fig 4. Level of variation among regions, practices, and prescribers in terms of the proportions of new opioid users with long-term opioid use.
Each horizontal line represents the point estimate with 95% confidence interval by region (left) or practice (middle) or prescriber (right). Regions, practices, and prescribers with 95% confidence intervals entirely above or below the population average (red vertical line) are indicated in blue. For instance, the adjusted proportion of long-term users for the North West region (15.8%) is significantly higher than the population average (14.6%). The largest variation is seen among practices and prescribers. The proportion of long-term opioid users for some practices reached up to 23.2%. The proportion of long-term opioid users for the highest risk prescriber was 37.2%.
Discussion
In this large national cohort of opioid-naïve patients in CPRD, we found a substantial increase in opioid prescribing for non-cancer pain between 2006 and 2017. Of the patients who started on high (120–199 MME/day) or very high dose opioids (≥200 MME/day), 10.3% and 18.7%, respectively, remained in the same MME/day category or higher at 2 years. We identified a number of patient-specific factors associated with long-term opioid use not previously identified in a UK population, most notably high initial dose/potency of opioid, fibromyalgia, rheumatological conditions, history of depression, prior gabapentinoid/psychotropic use, and history of major surgery. A wide variation in the risk of long-term opioid use was observed by prescriber, practice, and region. In addition, a regional divide in long-term opioid use risk was found, with the North West, Yorkshire and the Humber, and South West regions associated with the highest levels of long-term opioid use. Whilst there was a small proportion of prescribers (3.5%) who had significantly higher prescribing practices, their opioid prescribing rates were considerably higher in comparison to the population average. After adjusting for case mix, certain prescribers within a practice could be observed to be driving their entire practice towards high long-term opioid prescribing.
Comparison with previous studies and interpretation
To our knowledge this is the largest UK study evaluating opioid prescribing for non-cancer pain, with patient-level data to ascertain total amount of drug prescribed in terms of MME/day. We addressed a number of key questions quantifying the variance in prescribing at the regional, practice, and prescriber level. The finding of an overall rise in opioid prescribing is complementary to a recent study using National Health Service (NHS) digital pharmacy claims data demonstrating a 34% increase in opioid prescriptions between 1998 and 2016 [7]. This study however included prescription data on all opioids, including those prescribed for cancer pain, and lacked individual-level data, and high dose MME definitions were based on presumptions of daily dose. Our results are consistent with a previous cross-sectional CPRD study that reported a rise in morphine, oxycodone, fentanyl, and buprenorphine prescribing between 2000 and 2010 in a non-cancer population that we were able to extend both in time frame and in the range of opioids included [19]. In our study, codeine, morphine, and buprenorphine prescriptions in particular continued to rise until the end of 2017. Since 2013 in the UK, national regulations have been designed to improve the use and monitoring of controlled drugs such as opioids [20]. An increase in tramadol, oxycodone, and fentanyl prescriptions continued until 2012, following which prescribing plateaued, suggesting that GPs may have already started to reduce use of new opioids for these medications earlier. However, such regulations did not change prescribing patterns for morphine or buprenorphine.
Clinicians have an opportunity to be vigilant about what type of patient may become a long-term opioid user. A number of individual features associated with increased odds of long-term opioid use were identified. Older age and social deprivation were associated with an incremental increase in risk of long-term opioid use (Fig 3). Clinical-commissioning-group-level deprivation has been associated with higher population-level opioid prescribing using NHS digital data [8]. In the US, substance abuse, depression, and psychotropic medicines have been associated with an increased risk of opioid misuse [15,21], and we found these factors to also be associated with an increased risk of long-term opioid use in opioid-naïve patients in the UK. Additionally, benzodiazepines and gabapentinoid use also significantly increased odds and may be a surrogate for chronic pain severity. Concomitant use with opioids has also been associated with an increased risk of death [22,23]. Additionally, alcohol excess, fibromyalgia, rheumatological conditions, diabetes, and prior major surgery were significantly associated with a higher odds of long-term opioid use. In the US, new opioid users, especially post-surgery, have been shown to be a vulnerable population both for new persistent use and for developing opioid dependence/overdose [24,25]. Therefore, addressable patient-level factors and the existence of certain vulnerable groups at a higher risk of long-term use warrant increased awareness in prescribing clinicians.
An important finding in our study was that 14.6% of new users became long-term users over a 1-year period. In patients who were started on high doses of opioids (≥120 MME/day), a considerable proportion continued on higher doses throughout the following year. We also observed a wide variation at the practice and prescriber level in the adjusted odds of a new opioid user becoming a long-term opioid user (Fig 4), and the propensity of a practice being ‘high risk’ for its patients becoming long-term opioid users was being driven by a few prescribers in some cases. Of the 3 levels, prescribers had a bigger influence on long-term use than practice or region in our study. Whilst variation in prescribing between provider and practice has not been explored previously within a national UK setting, a US study examined the extent to which emergency physicians varied in rates of opioid prescribing and the implications of that variation for long-term opioid use. It was reported that prescribing rates varied widely between low-intensity and high-intensity prescribers (7.3% versus 24.1%), with long-term opioid use significantly higher among patients treated by high-intensity prescribers [26].
There are a few possibilities why prolonged opioid use may occur, in addition to ongoing appropriate prescribing for patients with clinical need. The variability in prescribing may in part be explained by unclear guidance regarding best practice in managing non-cancer pain. The advice regarding MME/day thresholds beyond which tapering should occur varies internationally [15,27]; therefore, GPs may not be aware of which patients to intervene with. Currently there is considerable heterogeneity in guidance internationally regarding dose thresholds that warrant caution, which vary between 50 and 200 MME/day [15,28,29]. In the US, national guidelines advise precautions and reassessment of patients exceeding 50 MME per day, and that prescribers should avoid increasing dose to 90 MME or more per day [15]. The Faculty of Pain Medicine in the UK suggests harms outweigh benefits when patients exceed 120 MME/day [27]. There is minimal guidance based on scientific evidence on how best to reduce/discontinue opioids in chronic pain. Tapering could fail to happen because clinicians are guided by patients, who may understandably fear worsening pain or withdrawal symptoms, may lack adequate social/healthcare support, or could perceive a lack of effectiveness of non-opioid pain relief options [30]. Alternatively, transition to long-term opioid use could be driven by ‘clinical inertia’ in some instances [26], where prescribers continue providing repeat prescriptions, assuming drug effectiveness without regular review.
The adjusted odds for long-term opioid use in opioid-naïve patients was highest in the North West, Yorkshire and the Humber, and South West regions of England (Fig 4). Regional UK variation in population-level opioid prescribing between the North and the South of England has been observed in recent studies [7,8]. A previous study using NHS digital data showed that 9 out of 10 of the highest prescribing areas in the country were located in the North of England, and there was an association with social deprivation [8]. Health is known to be worse in the North of England, and a strength of the present study was that we were able to account for case mix also, while previous studies have not. Whilst chronic pain severity was not measured, there is no known significant regional variation in the prevalence of chronic pain across strategic health authorities [31]. Therefore, it should not account for the observed regional differences in long-term opioid prescribing. We found, even after adjusting for deprivation (which has been linked to chronic pain [31]), the North of England/South of England disparities in long-term opioid use continued to exist.
References
- 1.
Smolina K, Gladstone E, Morgan SG. Determinants of trends in prescription opioid use in British Columbia, Canada, 2005–2013. Pharmacoepidemiol Drug Saf. 2016;25:553–9. pmid:26947145 - 2.
Boudreau D, Von Korff M, Rutter CM, Saunders K, Ray GT, Sullivan MD, et al. Trends in long-term opioid therapy for chronic non-cancer pain. Pharmacoepidemiol Drug Saf. 2009;18:1166–75. pmid:19718704 - 3.
Jani M, Dixon WG. Opioids are not just an American problem. BMJ. 2017;359:j5514. pmid:29212773 - 4.
Rudd R, Aleshire A, Zibbell J, Gladden M. Increases in drug and opioid overdose deaths—United States, 2000–2014. MMWR Morb Mortal Wkly Rep. 2016;64:1378–82. - 5.
Office for National Statistics. Deaths related to drug poisoning in England and Wales: 2016 registrations. Newport: Office for National Statistics; 2017 [cited 2018 Apr 23]. https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/bulletins/deathsrelatedtodrugpoisoninginenglandandwales/2016registrations. - 6.
Office for National Statistics. Opioid drug deaths by cause, 1993 to 2015. Newport: Office for National Statistics; 2015 [cited 2019 Jul 22]. https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/drugusealcoholandsmoking/adhocs/006149opioiddrugdeathsbycause1993to2015. - 7.
Curtis HJ, Croker R, Walker AJ, Richards GC, Quinlan J, Goldacre B. Opioid prescribing trends and geographical variation in England, 1998–2018: a retrospective database study. Lancet Psychiatry. 2019;6:140–50. - 8.
Mordecai L, Reynolds C, Donaldson LJ, de Williams ACC. Patterns of regional variation of opioid prescribing in primary care in England: a retrospective observational study. Br J Gen Pract. 2018;68:e225–33. pmid:29440012 - 9.
Taylor S, Annand F, Burkinshaw P, Greaves F, Kelleher M, Knight J, et al. Dependence and withdrawal associated with some prescribed medicines. London: Public Health England; 2019. - 10.
Gomes T, Mamdani MM, Dhalla IA, Paterson JM, Juurlink DN. Opioid dose and drug-related mortality in patients with nonmalignant pain. Arch Intern Med. 2011;171:686–91. pmid:21482846 - 11.
Dunn KM, Saunders KW, Rutter CM, Banta-Green CJ, Merrill JO, Sullivan MD, et al. Opioid prescriptions for chronic pain and overdose. Ann Intern Med. 2010;152:85. pmid:20083827 - 12.
Rhodes D. NHS accused of fuelling rise in opioid addiction. BBC News. 2018 Mar 15 [cited 2018 Apr 23]. http://www.bbc.co.uk/news/uk-england-43304375. - 13.
Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40:373–83. pmid:3558716 - 14.
UK Data Service. 2011 UK Townsend Deprivation Scores. Colchester: UK Data Service; 2011 [cited 2019 Jul 19]. https://www.statistics.digitalresources.jisc.ac.uk/dataset/2011-uk-townsend-deprivation-scores. - 15.
Dowell D, Haegerich TM, Chou R. CDC guideline for prescribing opioids for chronic pain—United States, 2016. JAMA. 2016;315:1624–45. - 16.
Pye SR, Sheppard T, Joseph RM, Lunt M, Girard N, Haas JS, et al. Assumptions made when preparing drug exposure data for analysis have an impact on results: an unreported step in pharmacoepidemiology studies. Pharmacoepidemiol Drug Saf. 2018;27:781–8. pmid:29667263 - 17.
Snijders TAB, Bosker RJ. Multilevel analysis: an introduction to basic and advanced multilevel modeling. 2nd edition. Thousand Oaks (CA): SAGE Publications; 2012. - 18.
Knowles J, Frederick C. merTools: tools for analyzing mixed effect regression models. Version 0.5.0. Comprehensive R Archive Network; 2016 [cited 2020 Sep 29]. https://cran.r-project.org/web/packages/merTools/index.html. - 19.
Zin CS, Chen LC, Knaggs RD. Changes in trends and pattern of strong opioid prescribing in primary care. Eur J Pain. 2014;18:1343–51. pmid:24756859 - 20.
Department of Health. Controlled drugs (supervision of management and use) regulations 2013: information about the regulations. London: Department of Health; 2013 [cited 2020 Sep 29]. https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/214915/15-02-2013-controlled-drugs-regulation-information.pdf. - 21.
Fleming MF, Balousek SL, Klessig CL, Mundt MP, Brown DD. Substance use disorders in a primary care sample receiving daily opioid therapy. J Pain. 2007;8:573–82. pmid:17499555 - 22.
Park TW, Saitz R, Ganoczy D, Ilgen MA, Bohnert ASB. Benzodiazepine prescribing patterns and deaths from drug overdose among US veterans receiving opioid analgesics: case-cohort study. BMJ. 2015;350:h2698. pmid:26063215 - 23.
Gomes T, Juurlink DN, Antoniou T, Mamdani MM, Paterson JM, van den Brink W. Gabapentin, opioids, and the risk of opioid-related death: a population-based nested case–control study. PLOS Med. 2017;14:e1002396. - 24.
Brummett CM, Waljee JF, Goesling J, Moser S, Lin P, Englesbe MJ, et al. New persistent opioid use after minor and major surgical procedures in us adults. JAMA Surg. 2017;152:e170504. pmid:28403427 - 25.
Brat GA, Agniel D, Beam A, Yorkgitis B, Bicket M, Homer M, et al. Postsurgical prescriptions for opioid naive patients and association with overdose and misuse: retrospective cohort study. BMJ. 2018;360:j5790. pmid:29343479 - 26.
Barnett ML, Olenski AR, Jena AB. Opioid-prescribing patterns of emergency physicians and risk of long-term use. N Engl J Med. 2017;376:663–73. pmid:28199807 - 27.
Faculty of Pain Medicine. Opioids aware: a resource for patients and healthcare professionals to support prescribing of opioid medicines for pain. London: Faculty of Pain Medicine; 2018 [cited 2018 Apr 24]. http://www.fpm.ac.uk/faculty-of-pain-medicine/opioids-aware. - 28.
Moisset X, Trouvin AP, Tran VT, Authier N, Vergne-Salle P, Piano V, et al. [Use of strong opioids in chronic non-cancer pain in adults. Evidence-based recommendations from the French Society for the Study and Treatment of Pain.] Press Med. 2016;45:447–62. pmid:27067628 - 29.
O’Brien T, Christrup LL, Drewes AM, Fallon MT, Kress HG, McQuay HJ, et al. European Pain Federation position paper on appropriate opioid use in chronic pain management. Eur J Pain. 2017;21:3–19. pmid:27991730 - 30.
Frank JW, Levy C, Matlock DD, Calcaterra SL, Mueller SR, Koester S, et al. Patients’ perspectives on tapering of chronic opioid therapy: a qualitative study. Pain Med. 2016;17:1838–47. - 31.
Bridges SL. Chronic pain. In: Craig R, Mindell, editors. Health survey for England–2011. Volume 1. Health, social care and lifestyles. London: Health and Social Care Information Centre; 2012 [cited 2020 Sep 29]. https://files.digital.nhs.uk/publicationimport/pub09xxx/pub09300/hse2011-ch9-chronic-pain.pdf.