Pain Medicine 2015; 16: 1720–1731 Wiley Periodicals, Inc. PRIMARY CARE & HEALTH SERVICES SECTION Simon Holliday, BMed, FAChAM, FRACGP, FACRRM,*,† Simon Morgan, MPH&TM, FRACGP,‡ Amanda Tapley, Biomedical Science(Hons),‡ Adrian Dunlop, PhD, FAChAM,*,† Kim Henderson, BNurs, Grad Dip. Health Soc. Sci.,‡ Mieke van Driel, PhD, FRACGP,§ Neil Spike, MBBS, FRACGP,¶,** Lawrie McArthur, MBBS, FRACGP,†† Jean Ball, BMath, Grad Dip Med Stat,‡‡ Chris Oldmeadow, PhD, BMathematics (Honours),§§ and Parker Magin, PhD, FRACGP*,‡ *School of Medicine and Public Health, Faculty of Health, University of Newcastle, Newcastle, NSW, Australia; †Drug and Alcohol Clinical Services, Hunter New England Local Health District, NSW, Australia; ‡ General Practice Training Valley to Coast, Mayfield, NSW, Australia; §Discipline of General Practice, School of Medicine, The University of Queensland, Brisbane, Queensland, Australia; ¶Department of General Practice, The University of Melbourne, Victoria, Australia; **Director of Medical Education and Training, VMA General Practice Training, Melbourne, Victoria, Australia; ††Director of Medical Education and Training, Adelaide to Outback GP Training, Adelaide, South Australia; ‡‡Clinical Research Design, Information Technology and Statistical Support (CReDITSS), Hunter Medical Research Institute, Newcastle, NSW, Australia; §§Centre for Clinical Epidemiology and Biostatistics, University of Newcastle, Newcastle, NSW, Australia Reprint requests to: Dr. Simon Mark Holliday, Albert St Medical Centre, 78 Albert St Taree NSW 2430 Taree, New South Wales 2430, Australia, Tel: 161 (0)2 6552 5533; Fax: 161 (0)2 6552 4249; Email: simon.holliday@albertstmc.com, Simon.Holliday@hnehealth.nsw.gov.au. This article was published online on 27 June 2015. An error was subsequently identified. This notice is included in the online and print versions to indicate that both have been corrected on 11 July 2015. 1720 Declaration of Conflicting Interests: This study was supported by a competitive grant from the Mental Health, Drugs and Alcohol Office of the NSW Ministry of Health. The project is also funded by the participating educational organisations: General Practice Training Valley to Coast, the Victorian Metropolitan Alliance, General Practice Training Tasmania, and Adelaide to Outback GP Training Program. These organisations are funded by General Practice Education and Training, an Australian Commonwealth Government initiative. Abbreviations: OST 5 Opioid Substitution Therapy; PNCP 5 Persistent Non-Cancer Pain; QUM 5 Quality Use of Medicines; PBS 5 Pharmaceutical Benefits Scheme; ReCEnT 5 Registrar (GP Trainee) Clinical Encounters in Training; MIMWOs 5 Medical Issue managed with opioids; ACSQHC 5 The Australian Commission on Safety and Quality in Health Care; RTPs 5 Regional Training Providers Abstract Objective. With escalating opioid prescribing come individual and public health harms. To inform quality improvement measures, understanding of opioid prescribing is essential. We aimed to establish consultation-level prevalence and associations of opioid prescribing. Design. A cross-sectional secondary analysis from a longitudinal multisite cohort study of general practitioner (GP) vocational trainees: “Registrar Clinical Encounters in Training.” Setting. Four of Australia’s seventeen GP Regional Training Providers, during 2010-13. Subjects. GP trainees. Downloaded from https://academic.oup.com/painmedicine/article-abstract/16/9/1720/1875845 by guest on 28 May 2020 Original Research Article The Pattern of Opioid Management by Australian General Practice Trainees Pattern of Opioid Management Methods. Practice and trainee demographic data were collected as well as patient, clinical and educational data of 60 consecutive consultations of each trainee, each training term. Outcome factors were any opioid analgesic prescription and initial opioid analgesic prescription for a specific problem for the first time. Conclusions. Most opioids were prescribed as maintenance therapy for non-cancer pain. Demographic associations with opioid analgesic prescribing resemble those presenting for opioid dependency treatment. Our findings should inform measures by regulators and medical educators supporting multimodal pain management. Key Words. Opioids; Analgesic; Pain Management; Primary Care; Quality of Health Care; Prescriptions; Persistent Pain; Risk Factors; Aboriginal and Torres Strait Islander Background Pleasure and pain have been improved or treated by opium for millennia. Still, opioid therapeutic use remains a conundrum [1–3]. Opioid management has evolved into several therapeutic models. Opioid Substitution Therapy (OST) is a highly structured and evidence based model of care to minimize harms from opioid dependency [4]. OST remains relatively inaccessible with few (2.1%) Australian doctors authorized to prescribe it [5,6]. Barriers to GPs becoming OST prescribers frequently include stigma, workload, lack of specialist support or perceived threat [7]. The specialty of palliative care developed a model of opioid analgesia based on self-titration and minimal monitoring [8]. The specialty of pain medicine adopted this model to address the under-treatment of chronic pain, regardless of etiology [2,9,10]. The Australian National Pain Strategy claims that less than 10% of Persistent Concurrent with claims of under-treatment, opioid prescribing in Australia increased 15-fold between 1992 and 2012 [15]. During this time pharmaceutical opioids became the most common cause (69%) of accidental opioid overdoses, especially in older age groups [16]. An Australian study tested assertions that most prescribing was for PNCP [17] and found that only a minority of opioid prescribing was for “chronic” conditions 43.9%) or malignant neoplasms (3.5%) [18]. Whether opioid maintenance is indicated in PNCP is increasingly contested [12,19,20] with one recent guideline stating PNCP is currently not an evidence-based indication for long term analgesics [21]. Similarly, a systematic review by the United States Food and Drug Administration found no adequate or well-controlled studies of opioid use longer than 12 weeks [22]. A number of Australian studies have showed that opioid analgesics are frequently prescribed outside Pharmaceutical Benefits Scheme (PBS) subsidization criteria or prescribing guidelines [13,15,21,23,24]. Regulators have responded in various ways. Of disciplinary cases where doctors have been sanctioned by tribunals, 25% involve inappropriate prescribing, mainly drugs of dependency [25,26]. In the USA, the Centers for Disease Control and Prevention is currently establishing prescribing norms to identify and intervene in cases of patients or providers who fall outside these norms [27]. The Australian Commission on Safety and Quality in Health Care (ACSQHC) is similarly investigating mapping variations in opioid prescribing (Dr. Heather Buchan: director: personal communication 11/8/2014). Mapping such variations in opioid prescribing is predicated on detailed and reliable prescribing data. The ACSQHC has established the Quality Use of Medicines (QUM) framework to consider how pharmaceutical options are chosen by clinicians, which medicines are selected and whether the use of these medicines is both safe and effective [28]. To improve access to selected medications for specific indications, the Commonwealth Government provides Australian residents subsidies through the PBS. Data from the PBS is readily accessible to researchers but 1721 Downloaded from https://academic.oup.com/painmedicine/article-abstract/16/9/1720/1875845 by guest on 28 May 2020 Results. Overall, 645 trainees participated. Opioids comprised 4.3% prescriptions provided for 3.8% of patients. Most frequently prescribed were codeine (39.9%) and oxycodone (33.4%). Prescribing was for acute pain (29.3%), palliative care (2.6%) or other indications (68.1%). Most prescribing involved repeat prescriptions for preexisting problems (62.7% of total). Other associations included older patients; prescriber and patient male gender; Aboriginal/Torres Strait Islander status; rural and disadvantaged locations; longer consultations; and generation of referrals, follow-up, and imaging requests. Opioid initiation was more likely for new patients with new problems, but otherwise associations were similar. Trainees rarely reported addiction riskmitigation strategies. Non-Cancer Pain (PNCP) patients, gain access to effective analgesia [11]. The strategy champions access to opioids for “legitimate users” [11]. Notwithstanding the evidence that opioid-related mortality is associated with different variables [10], the strategy proscribes their long term provision in the context of “predominantly psychological factors,” “inappropriate” prescribing or “unsanctioned use” [11]. Such a judgement, reliant on a clinician’s experience, prejudices and knowledge [1], may be especially problematic for “uninitiated prescribers due to the lack of a useful case definition for the various dependent states” [12]. Australian GPs report infrequently effecting such judgements [13], with associated barriers involving the beliefs and stigmatizing attitudes of clinicians and time pressures, similar as towards OST [7,14]. Financial and geographical barriers to accessing multidisciplinary pain management are also involved [11]. Holliday et al. significantly under-estimates prescribing. This is because it does not capture unsubsidized medications, any medications where the dispensed price is below the patient contribution (including many older and cheaper opioids) or “private” or inpatient prescriptions written outside the public community dispensing framework [29]. PBS-data also does not facilitate linkage to individual general practitioner (GP) or patient data [17,29]. Methods This analysis was conducted within the Registrar Clinical Encounters in Training (ReCEnT) project. ReCEnT is an ongoing multisite cohort study of GP trainees’ (registrars’) in-consultation clinical experience. The Australian General Practice Training Program involves a minimum of three years training, of which at least three six-month Full Time Equivalent terms must be spent in general practice rather than hospital-based practice. The general practice component is delivered by 17 geographically defined Regional Training Providers (RTPs) and their affiliated training general practices and leads to Fellowship of either the Royal Australian College of General Practitioners and/or the Australian College of Rural and Remote Medicine. Though GP trainees have recourse to their supervisors for advice and support in an “apprenticeship-like model,” they function as independent practitioners (including for prescribing). The trainees in the ReCEnT study are from four of Australia’s seventeen RTPs spanning four of Australia’s eight states and territories (New South Wales, Victoria, South Australia and Tasmania). One state, Tasmania, commenced rolling-out a real-time prescription monitoring system in 2010, [32] and now approximately one third of Tasmanian general practices have requested it and have access to it (Clinical Associate Professor Adrian Reynolds: Tasmanian clinical director of alcohol and drug services: personal communication 8/4/2015). The study encompasses general practices across all rural-urban classifications from Major City to Very Remote [33]. The ReCEnT methodology has been described in detail elsewhere [34]. Briefly, GP trainees record the details of 60 consecutive clinical consultations with a paper1722 Data in the current analysis is from eight rounds of data collection, 2010-13 and was confined to patients 16 years of age or older reflecting the sparsity of research, specialist treatment centers and GP opioid analgesia prescribing for the younger demographic [11,18]. Outcome Factors The primary outcome factor in this study was the prescription of an opioid for pain management. We selected these opioids using the “N02A” and “N01AH” codes from the N (nervous) section of International Anatomical Therapeutic Chemical [35]. We excluded opioids (codeine) used as cough suppressants (code “R05DA04”) or used for addictive disorders (codes “NO7B A-C”) a priori, as our primary focus was analgesia [35]. The secondary outcome factor was the “initial” prescription of an opioid for analgesia. A prescription was classified as “initial” when used for the first time for that specific medical issue (accepting that the medicine may have been used for that patient for a different medical issue). If a medicine was a continuation or repeat of previous therapy, (signifying historical as well as current prescribing decisions), it was classified as “continuing.” Independent Variables Independent variables collected related to trainee, patient, practice, and consultation. Trainee factors were age, gender, training term, country of medical qualification (Australia/international), and the number of half-days worked per week. Practice factors included practice size (number of full-time equivalent GPs), and if the practice routinely bulk-billed (i.e., the Commonwealth government completely reimburses a practitioner for the consultation leaving no financial cost to the patient). Postcode was used to define practice rurality/urbanicity using the Australian Standard Geographical ClassificationRemoteness Area classification of the practice location [33] and the practice location’s Socioeconomic Index for Area (SEIFA) Relative Index of Disadvantage [36]. Patient factors were age, gender, Aboriginal/Torres Strait Islander status, non-English speaking background and whether new to the practice or the trainee. Consultation factors were duration of consultation in minutes as estimated and recorded by the trainee, the number of medical issues managed, the ordering of imaging, the number of pathology tests ordered Downloaded from https://academic.oup.com/painmedicine/article-abstract/16/9/1720/1875845 by guest on 28 May 2020 Due to the limitations of PBS-level data, calls have been made for population based studies to explore individual level data and identify risky patterns of opioid prescribing particularly to disadvantaged populations [15,26,29,30]. Such studies would inform interventions to promote future QUM but frequently suffer from low GP response rates [31]. Such interventions must involve early-career GPs as they comprise largely “uninitiated” prescribers whose opioid prescribing has been identified as challenging [12] and less guideline concordant [13]. In this study of Australian GP trainees, we aim to map the prevalence, nature and associations of opioid analgesic prescribing by using contemporaneously recorded detailed individual GP trainee consultation data. based encounter form at approximately the mid-point of each six-month general practice training term. As data collection is designed to reflect a “normal” week of general practice, consultations in specialized clinics (e.g., vaccinations) are excluded. Only office-based consultations, not home or nursing home visits, are included. Initial data collection involves demographic, education, work experience and current practice characteristics. The encounter form data encompasses four broad areas: patient demographics, diagnoses or problems managed (hereafter referred to as medical issues), management, and educational training aspects. Pattern of Opioid Management Table 1 Participating trainee, trainee-term, and practice characteristics Variable Trainee variables (n 5 645) Trainee gender Pathway trainee enrolled in Trainee age (years) Trainee year of graduation Trainee-term or practice-term variables (n 5 1426) Trainee training term Trainee worked at the practice previously Trainee works fulltime Does the practice routinely bulk bill Number of GPs working at the practice Rurality of practice SEIFA* n% (95% CIs) or Mean (SD) Male Female General Rural No Yes Mean (SD) Mean (SD) 202 34.1% (30.4–37.8) 425 65.9% (62.2–69.6) 494 77.0% (73.7–80.2) 148 23.0% (19.8–26.3) 155 24.4% (21.1–27.8) 480 75.6% (72.2–78.9) 32.8 (6.6) 2005.1 (5.6) Term 1 Term 2 Term 3 Term 4 No Yes No Yes No Yes 1–4 5–101 Major city Inner regional Outer regional or remote Mean (SD) 557 39.1% (36.5–41.6) 488 34.2% (31.8–36.7) 306 21.5% (19.3–23.6) 75 5.3% (4.1–6.4) 994 70.7% (68.3–73.0) 413 29.3% (27.0–31.7) 302 21.7% (19.5–23.8) 1091 78.3% (76.1–80.5) 1179 83.4% (81.5–85.4) 234 16.6% (14.6–18.5) 454 32.5% (30.1–35.0) 941 67.5% (65.0–69.9) 827 58.0% (55.4–60.6) 424 29.7% (27.4–32.1) 175 12.3% (10.6–14.0) 5.4 (2.8) * Socioeconomic Index for Area (SEIFA) relative index of disadvantage: lower deciles are relatively disadvantaged. or whether a referral or scheduled follow-up was made. Educational factors included whether the trainee sought assistance during the consultation or generated learning goals for subsequent attention. Registrars could code each problem as either “new” or “pre-existing.” The former included initial episodes, exacerbations of recurrent problems, or any problems (regardless of duration) which were being managed for the first time. The medical issues nominated by the registrars were coded according to the International Classification of Primary Care (second edition) (ICPC-2 PLUS) system [37]. This is the international standard for classifying primary care data and the validity of this system has previously been demonstrated [38]. Individual diagnoses/problems are grouped in 17 Disease Chapters and further classified as process codes, symptoms/ complaints, infections, neoplasms, injuries, congenital anomalies, and other diagnoses. Chronic diseases in our study were coded via a classification system derived from ICPC-2 [39]. We also manually reviewed all the coded diagnoses to identify those which related to palliative care or acute pain (taken to include: acute trauma; any injury classified by the trainee as “new”; or issues identified as preoperative or postoperative) with the remainder regarded as PNCP. Statistical Analysis This was a cross-sectional analysis of patient consultations from the longitudinal ReCEnT study. The unit of analysis for both primary and secondary outcomes was the individual medical issue rather than the individual consultation. Percentages of trainees’ medical issues managed with opioids (MIMWOs) were calculated, with 95% confidence intervals. To test associations of an opioid being prescribed, simple and multiple logistic regression were used within a generalized estimating equations (GEE) framework. GEE is a parameter estimation technique that accounts for the lack of independence in the data due to clustering from repeated measures from registrars. In GEEs a working correlation matrix is specified (i.e., a model for the within-cluster correlation) and this is used to re-estimate the regression parameters and standard errors. We have used the compound symmetry working correlation matrix, in this analysis; this assumes the same correlation parameter for all repeated measures, but these methods are robust to misspecification. All independent variables (above) with a p value less than 0.20 and a relevant effect size in the univariate 1723 Downloaded from https://academic.oup.com/painmedicine/article-abstract/16/9/1720/1875845 by guest on 28 May 2020 Qualified as a doctor in Australia Class Holliday et al. Table 2 Opioids prescribed Frequency codeine combinations oxycodone tramadol buprenorphine morphine fentanyl other* Total 1,145 1002 337 203 117 95 33 2,932 Actual % of opioids (95% CI) 39.05 (36.82–41.38) 34.17 (32.09–36.36) 11.49 (10.30–12.79) 6.92 (6.00–7.94) 3.99 (3.30–4.78) 3.24 (2.62–3.96) 1.13 (0.77–1.58) * Other opioids include: dihydrocodeine, hydromorphone, pethidine. analysis were included in the multiple regression model. Variables which had a small effect size and were no longer significant in the multivariate model were removed from the final model as long as the variable’s removal did not change the resultant model. The SEIFA variable was used as a continuous variable with 10 deciles. Consultation duration was also analyzed as a continuous variable. Linearity assumptions for the continuously modelled variables were assessed using graphical inspections of predicted probabilities over the range of the variables values. Odds ratios (ORs) were presented with 95% confidence intervals (CIs). Two regression models were built (using all available data), the first with dependent variable “any opioid prescribed.” This analysis included all medical issues. The dependent variable in the second model was “initial opioid prescribed.” This analysis excluded MIMWOs where the prescription was “continuing.” Statistical analyses used SAS v9.3. Correlates were considered statistically significant if the P-value was < 0.05. Ethics Approval The ReCEnT project has approval from the University of Newcastle Human Research Ethics Committee, Reference H-2009-0323. Results In all, there were 1,426 training term recording cycles (including details of 69,621 individual consultations, 112,890 medical issues and 68,582 medications prescribed) contributed to by 645 individual trainees (response rate 94.0%). The demographics of the participating trainees and practices are presented in Table 1. Overall Findings Opioids were prescribed in 2660 (3.8% of total) consultations and 2,675 (2.4% of total) medical issues, 1724 Opioids were prescribed for 372 individual ICPC-2 PLUS conditions. Our manual review of these found opioid analgesics were provided for acute, palliative and PNCP indications in 29.3%, 2.6%, and 67.7% of total MIMWOs, respectively. For PNCP, the most common MIMWOs was back complaints (609/22.8% of MIMWOs). Some (15/0.6% MIMWOs) opioid analgesics were provided for medical issues classified in terms of addiction rather than pain. Other diagnoses are presented in Table 3. General Prescribing Associations The univariate associations of a MIMWO are presented in Table 4 with the multivariate model presented in Table 5. In the adjusted model, variables associated with MIMWOs at the 5% significance threshold were: Patient factors: Older age (OR: 1.9 CI: 1.7–2.2 & 1.8 CI: 1.6–2.1, respectively for 35–64 & 651 compared to 16– 34), male gender (OR: 1.2 CI: 1.1–1.3), and Aboriginal and Torres Strait Islander status (OR: 2.2 CI: 1.6–2.9). Trainee factor: Male gender (OR: 1.2 CI: 1.6–2.9). Practice factors: Outer regional, remote or very remote location (OR: 1.3 CI: 1.1–1.7) (as compared to major cities) and a lower SEIFA score (i.e., relative disadvantage) (OR: 0.97 for each decile increase CI: 0.950– 0.998). Consultation factors: new medical problems (OR: 0.4 CI: 0.36–0.44), “chronic” problems (OR: 0.6 CI: 0.5–0.7), ordering imaging (OR: 1.4 CI: 1.2–1.7), Table 3 Ten most frequent medical issues for which opioids were prescribed Medical issues (n 5 2675) Frequency % (95% CI) Back complaint Prescription: all Arthritis: all Pain: chronic Fracture Sprain/strain Migraine Pain: post-op Pain: shoulder Pain: knee 609 219 198 172 119 74 58 50 42 40 22.77 (21.19, 24.40) 8.19 (7.18, 9.29) 7.4 (6.44, 8.46) 6.43 (5.53, 7.43) 4.45 (3.70, 5.30) 2.77 (2.18, 3.46) 2.17 (1.65, 2.79) 1.87 (1.39, 2.46) 1.57 (1.13, 2.12) 1.5 (1.07, 2.03) Downloaded from https://academic.oup.com/painmedicine/article-abstract/16/9/1720/1875845 by guest on 28 May 2020 Medication name accounting for 2932 (4.3%) of all prescriptions. One, two or three opioids were provided in 91.1%, 8.5% and 0.5% of MIMWOs respectively. Specific opioids prescribed are presented in Table 2. Excluded from the statistical analysis were 38 OST scripts as were the 18,693 problems for those aged under 16 years which included 42 (0.22%) MIMWOs. Pattern of Opioid Management Table 4 Univariate associations of independent variables with opioid prescriptions Opioid Prescribed Variable Patient Related Variables Age group Aboriginal and Torres Strait Islander Non-English speaking background Patient/practice status Trainee related variables Trainee gender Qualified as doctor in Australia Registrar age Trainee working week Training term Worked at the practice previously Practice related variables Practice size Practice routinely bulk bills Rurality Regional training provider SEIFA Consultation-related variables Consultation Duration Number of medical issues New problem “Chronic” condition Sought help from any source Imaging ordered No (n 5 110215) Yes (n 5 2675) P 16–34 35–64 651 Male Female No Yes No Yes Returning patient New to registrar New to practice 31552 (98.5%) 51828 (97.3%) 24726 (97.3%) 37655 (97.2%) 69551 (97.9%) 103467 (97.7%) 1131 (95.0%) 98361 (97.6%) 6889 (98.2%) 49957 (97.3%) 50059 (97.9%) 7013 (98.3%) 486 (1.5%) 1464 (2.7%) 688 (2.7%) 1099 (2.8%) 1504 (2.1%) 2477 (2.3%) 60 (5.0%) 2424 (2.4%) 126 (1.8%) 1380 (2.7%) 1091 (2.1%) 123 (1.7%) <0.0001 Male Female No Yes mean (SD) Part time Full time Term 1 Term 2 Term 3 Term 4 No Yes 36206 (97.1%) 74009 (97.9%) 26303 (97.3%) 82157 (97.7%) 33 (6.7) 23975 (97.8%) 83742 (97.6%) 44221 (97.8%) 36684 (97.5%) 23678 (97.6%) 5632 (97.5%) 76764 (97.8%) 31951 (97.4%) 1096 (2.9%) 1579 (2.1%) 733 (2.7%) 1893 (2.3%) 33 (6.7) 531 (2.2%) 2063 (2.4%) 995 (2.2%) 959 (2.5%) 575 (2.4%) 146 (2.5%) 1762 (2.2%) 868 (2.6%) Small Large No Yes Major city Inner regional Outer regional or Remote 1 2 3 4 mean (SD) 36312 (97.6%) 71609 (97.7%) 90876 (97.6%) 18420 (97.6%) 63730 (97.9%) 32714 (97.4%) 13771 (96.7%) 40305 (97.5%) 13047 (97.0%) 11096 (97.5%) 45767 (97.9%) 5.3 (2.8) 903 (2.4%) 1697 (2.3%) 2204 (2.4%) 451 (2.4%) 1343 (2.1%) 859 (2.6%) 473 (3.3%) 1019 (2.5%) 405 (3.0%) 288 (2.5%) 963 (2.1%) 4.9 (2.8) mean (SD) mean (SD) No Yes No Yes No Yes No Yes 19 (9.9) 2.1 (1.0) 57443 (96.8%) 52772 (98.5%) 82480 (97.5%) 27385 (97.9%) 93860 (97.7%) 16355 (97.4%) 101634 (97.7%) 8581 (96.3%) 19 (11) 1.7 (0.9) 1897 (3.2%) 778 (1.5%) 2095 (2.5%) 575 (2.1%) 2243 (2.3%) 432 (2.6%) 2347 (2.3%) 328 (3.7%) <0.0001 <0.0001 0.0156 <0.0001 <0.0001 0.0113 0.0986 0.1814 0.0447 0.0146 0.6857 0.8739 <0.0001 0.0004 0.0006 0.0912 <0.0001 <0.0001 <0.0001 0.0153 <0.0001 1725 Downloaded from https://academic.oup.com/painmedicine/article-abstract/16/9/1720/1875845 by guest on 28 May 2020 Patient gender Class Holliday et al. Table 4 Continued Opioid Prescribed Variable Class No (n 5 110215) Yes (n 5 2675) P Follow-up ordered No Yes No Yes No Yes mean (SD) 60385 (98.1%) 49830 (97.1%) 94771 (97.7%) 15444 (97.5%) 96774 (97.8%) 13441 (96.3%) 0.6 (1.6) 1191 (1.9%) 1484 (2.9%) 2280 (2.3%) 395 (2.5%) 2154 (2.2%) 521 (3.7%) 0.2 (0.9) <0.0001 Learning goals Number pathology <0.0001 <0.0001 * Numbers in columns may not add up to n because of missing data. number of pathology tests ordered in the consultation (OR: 0.7 for each one ordered, CI: 0.67–0.79), rates of follow-up (OR: 1.5 CI: 1.4–1.7) and referral (OR: 1.4 CI: 1.2–1.5), consultation duration (by a statistically significant but clinically small amount) (OR: 1.02 CI: 1.017– 1.027), and number of other diagnoses (OR: 0.6 for each extra diagnosis managed CI: 0.55–0.63). The frequency of missing data was 5% for Aboriginal and Torres Strait Islander status, with lower frequencies for all other items justifying a complete-case analysis. In terms of addiction-risk mitigation strategies, no trainee reported contacting a Prescription Monitoring Program and only two MIMWOs involved urine drug screens. There were no reports of trainees seeking authorization to comply with the various State regulations [5]. Initial Prescribing Of opioids prescribed, 37.3% were “initial” prescriptions. The univariate and multivariate associations of an “initial” opioid prescription are presented in Supporting Information Tables 6 and 7. Most associations in the adjusted model remained as for the general prescribing analysis. Differences included associations losing statistical significance: patient gender, rurality, classification as “chronic,” and SEIFA index. One association became statistically significant with more senior trainees (Term 2 versus Term 1) more likely to prescribe opioids (OR: 1.21 C.I. 1.02–1.44). There were two associations which were of opposite direction to general opioid prescribing and reached statistical significance. “Initial” opioids were more likely to be prescribed for a new problem rather than a pre-existing one (OR 1.17 C.I 1.01–1.36) and to a patient new to the registrar (OR 1.25 C.I 1.07–1.45). Discussion Overall Findings: Rates and Characteristics of Prescription Our findings demonstrate GP trainees provided opioids predominantly as recurrent prescriptions for PNCP. How1726 ever, the frequency of MIMWOs (2.4%), was a lower rate than in an Australian national study of established GPs (3.7%) [18]. This could possibly be a response to recent evidence indicating the relative benefits to harms of opioids in PNCP have been over-stated [12,19,21,40]. However, we found scant evidence trainees utilized any addiction-risk mitigation strategies as recommended in guidelines such as urinary drug screening, checking with a prescription monitoring scheme, or seeking authorization as per each state’s regulations [21,30,32]. Additionally there were 15 instances of opioid analgesic prescribing for dependency indications. Malignancy accounted for only 2.6% of MIMWOs in our study (compared to established GPs: 3.5% opioids prescribed [18]). The most frequent specific indication was for back complaints (22.8% MIMWOs versus more established GPs 27.1% opioids prescribed [18]). There is evidence clinicians are increasingly preferring opioid analgesics over nonopioids for back pain [41] despite a recent meta-analysis showing their long-term efficacy is unclear with rates of concurrent substance use disorders and aberrant opioid behaviors as high as 43% and 24%, respectively [40]. As with other studies [18,29], codeine was most frequently prescribed and tramadol the third most. Both rely on the CYP2D6 enzyme to convert into morphine or become active [42]. Because many people may have either ultrarapid or nonexistent rates of CYP2D6 activity, providing a specific dose of either of these may equate to prescribing an unknown dose of morphine [42]. While low-dose codeine compounds may be purchased over-the-counter from pharmacists unmonitored by regulators [30], uncompounded codeine is a “Schedule 8” or a controlled prescription medicine. Between 5% and 20% of all prescribed codeine preparations have been found to be provided to patients identified as “prescription shoppers” [26]. Recent Australian entrants to OST for pharmaceutical dependency identify their primary drug of concern as oxycodone (51%) and codeine (28%) [43]. The second most commonly prescribed opioid, oxycodone has a relatively high abuse liability profile due to higher subjective attractiveness ratings and a lack of negative subjective effects [44]. Increased oxycodone prescribing is associated with opioid-related Downloaded from https://academic.oup.com/painmedicine/article-abstract/16/9/1720/1875845 by guest on 28 May 2020 Referral ordered 0.4024 Pattern of Opioid Management Table 5 Logistic regression model of associations of opioid prescriptions Univariate Adjusted Class OR (95% CI) P OR (95% CI) P Age group Referent: 16–34 Patient gender Aboriginal and Torres Strait Islander Non English speaking background Trainee gender Trainee age Training term Referent: term 1 35–64 651 Male Yes Yes Male 1.79 1.70 1.30 2.12 0.79 1.42 1.01 1.16 1.11 1.17 0.84 1.14 1.20 1.61 (1.61, (1.49, (1.19, (1.63, (0.65, (1.26, (1.00, (1.04, (0.97, (0.95, (0.73, (1.03, (1.06, (1.37, 1.99) 1.94) 1.42) 2.74) 0.96) 1.60) 1.01) 1.29) 1.26) 1.44) 0.96) 1.26) 1.35) 1.89) <0.0001 <0.0001 <0.0001 <0.0001 0.0156 <0.0001 0.0986 0.0068 0.1364 0.1406 0.0113 0.0146 0.0033 <0.0001 1.90 1.80 1.19 2.19 0.84 1.18 0.99 1.11 1.06 1.16 0.95 1.05 1.14 1.34 (1.68, (1.56, (1.09, (1.63, (0.67, (1.03, (0.98, (0.97, (0.91, (0.89, (0.81, (0.92, (0.96, (1.07, 2.15) 2.08) 1.31) 2.94) 1.06) 1.36) 1.00) 1.27) 1.24) 1.52) 1.11) 1.20) 1.36) 1.69) <0.0001 <.0001 0.0002 <0.0001 0.1514 0.0191 0.2311 0.1379 0.4464 0.2729 0.5381 0.4432 0.1438 0.0116 1.24 1.02 0.84 0.97 1.00 0.61 0.45 0.80 1.66 1.53 1.74 1.14 0.74 (1.03, (0.84, (0.74, (0.95, (1.00, (0.58, (0.41, (0.72, (1.46, (1.41, (1.58, (1.03, (0.69, 1.50) 1.24) 0.97) 0.99) 1.01) 0.65) 0.50) 0.89) 1.88) 1.67) 1.92) 1.27) 0.79) 0.0263 0.8601 0.0138 0.0006 0.0912 <0 .0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 0.0153 <0.0001 1.11 0.88 0.95 0.97 1.02 0.59 0.40 0.60 1.43 1.54 1.37 0.95 0.73 (0.88, (0.69, (0.79, (0.95, (1.02, (0.55, (0.36, (0.53, (1.24, (1.38, (1.22, (0.83, (0.67, 1.41) 1.11) 1.14) 1.00) 1.03) 0.63) 0.44) 0.68) 1.65) 1.71) 1.54) 1.08) 0.79) 0.3914 0.2835 0.5980 0.0319 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 0.4135 <0.0001 Qualified as doctor in Australia Worked at the practice previously Rurality Referent: Major city Regional training provider Referent: 1 SEIFA Consultation duration Number of medical issues New problem “Chronic” condition Imaging ordered Follow-up ordered Referral ordered Sought help from any source Number pathology Term 2 Term 3 Term 4 Yes Yes Inner regional Outer regional/ Remote/Very remote 2 3 4 Yes Yes Yes Yes Yes Yes mortality [10] leading to increased restrictions in some jurisdictions [44]. Associations of Overall Prescribing (Including Both Initial and Continuing Prescriptions) The association of opioid prescribing with certain patient, prescriber and practice demographics in our study may be of concern. A systematic review found higher levels of opioid prescribing related to opioid-related harms including mortality [10]. A US analysis showed patient gender and race reliably influenced pain management decisions of almost half (45%) of US medical trainees [45]. Variations in opioid prescribing rates, in another US study, were found to relate more strongly to physician behaviors than either shifting demographics or disease incidence rates [20]. The importance of these findings is that rates of opioid prescribing are directly related to adverse outcomes. Associations of increased trainee prescribing included patient older age, male gender, Aboriginal/Torres Strait Islander status, more rural location and lower socioeconomic location. These associations paralleled the demographics over-represented in OST programs [5], amongst opioid misusers [26,46], and increased opioidrelated mortality [10]. The aged are particularly at risk of opioid-related toxicities such as falls, respiratory depression, delirium and sedation [15,23,24] . The association with older age in our study may have been more prominent if consultations in residential aged care facilities had been included; with an estimated 28.1% residents there taking regular opioids [23]. Our finding of males being prescribed more opioids has been shown in some previous studies [18,20,47]. It is said to reflect prescribers’ subjective judgements due to social categorization [47]. We found that Aboriginal and Torres Strait Islander patients received more opioids. This may reflect a greater symptom burden [11]. Alternatively, US studies indicate race may influence analgesia decisions [3,45,48] especially those made by inexperienced doctors under cognitive stress [47]. But increased prescribing for Aboriginal and Torres Strait Islander patients may reflect increased opioid prescribing to lower 1727 Downloaded from https://academic.oup.com/painmedicine/article-abstract/16/9/1720/1875845 by guest on 28 May 2020 Variable Holliday et al. socioeconomic and more rural populations [29]. In the US, these latter groups have higher rates of selfreported pain, and higher rates of opioid analgesic misuse [46]. Such treatment disparities may reflect lack of access to health services of these groups and the relative expense and geographical inaccessibility of nonopioid or multidisciplinary pain management [11,21,26,46]. The analysis indicates these “uninitiated” prescribers found MIMWOs problematic. The increased recourse to imaging may reflect attempts to ascertain the biophysical basis of the problem to judge the legitimacy of the patient’s pain status or the opioid use [3,14]. The complex needs of patients on opioid analgesics are further indicated by increased rates of referral and follow-up and increased consultation duration despite a decreased number of medical issues addressed per consultation. Increased duration of consultations involving opioids has been found elsewhere, particularly when a pain patient is new to a doctor [52]. Differences in Associations of Overall Prescription versus “Initial” Prescription An “initial” opioid script was more likely to be provided to patients not seen previously by the trainee, and to patients with new problems, some of whom may be “prescription shoppers.” This suggests that for these new pain presentations that trainees may not first trial nonopioid or nonpharmacologic analgesia prior to the initiation of opioids, which is inconsistent with guidelines 1728 Strengths and Limitations Strengths of the study include a large sample size from four states across all rural–urban classifications with a response rate singularly high for a GP consultation-level study [31] and with participant demographics broadly similar to those of Australian GP trainees overall [53]. The contemporaneous recording of detailed patient, prescriber, practice and consultation variables and the diagnostic indication for prescription and whether this diagnostic indication was a new medical issue, all linked to the individual prescription is a particular strength. As is our ability to ascertain all prescribed and recommended opioid medications, not only those that attract a PBS subsidy. Limitations of this analysis include its cross-sectional “consultation snapshot” nature and lack of data on full medication regimens. Also, as a secondary analysis of an existing data set, we do not have data on some variables of particular relevance to this research question: patient mental health status, substance use history pain scores or dosages. This means our analysis does not allow assessment of appropriateness or quality of trainees’ prescribing decisions, and could allow that any associations of opioid prescription were associations of patient pain per se. Implications for Practice and Policy The demographic (rather than strictly clinical) associations of opioid prescribing in our study raise important issues regarding opioid and pain management. Being older, male, or Aboriginal/Torres Strait Islander, and being from rural or remote and lower socioeconomic status locations were all associated with opioid prescription in adjusted models. These groups may thus be at disproportionate risk of poor pain management or opioid-related harms. High rates of prescribing are important determinants of rates of fatal overdose [10] and iatrogenic opioid dependency [54] and correlate with demographics overrepresented on OST programs [5]. Our findings should inform educational interventions addressing barriers to and promoting guideline consistent opioid and pain management [13,14,54]. Our findings also indicate the need for more accessible nonopioid pain management services, especially for geographically isolated patients and socially marginalized groups. Future Research Prescribing habits start early on in a clinical career, so improving QUM starts with the education of those in training. Increasing awareness may reduce stigma and Downloaded from https://academic.oup.com/painmedicine/article-abstract/16/9/1720/1875845 by guest on 28 May 2020 Opioids were mainly prescribed as repeat scripts for preexisting medical issues to patients familiar to the practice. A review of the reasons clinicians fail to deprescribe identified four reasons why trainees may default to careas-usual [49]: nonidentification of the medication as inappropriate; an inertia where continuing to prescribe seems less trouble (e.g., from withdrawal syndromes); hesitance about the management of pain (and its pharmaceuticals); and external factors example, time pressures and patient resistance [2,3,49]. Those on opioid maintenance for analgesia describe poorer physical function and higher psychological distress [29] and so trainees may feel intrapractice consensus is required prior to initiating a change in management. But deprescribing is difficult for most GPs with 89%, in an Australian survey, reporting never or only “occasionally” terminating opioids even when faced with aberrant opioid behaviors [13]. In a large US survey one year after the commencement of long-term opioid analgesia, only 7.5% patients had discontinued them and only 20% had done so when followed for up to 3.5 years [50]. The potential for adverse effects in continued prescribing is highlighted in another US study which showed that duration of opioid analgesia prescribing is strongly associated with risk of opioid use disorders [51]. Amongst Australian entrants to OST for pharmaceutical dependency, two thirds (66%) had initiated opioids for analgesia and 28% continued them for this reason [43]. [17,21,24]. These findings suggest “independent” trainee opioid prescribing decisions for new presentations are made similarly to those made for “inherited” PNCP cases. Pattern of Opioid Management unintentional bias [48] and reduce discomfort managing the complex and competing demands of pain and addiction management [1,8]. The trainee cohort study described in this paper provides a unique opportunity to trial an intervention in an educationally receptive cohort to promote multimodal pain management [34]. Conclusion References 1 Ballantyne JC, LaForge KS. Opioid dependence and addiction during opioid treatment of chronic pain. Pain 2007;129:235–55. 2 Crowley-Matoka M, True G. No one wants to be the candy man: ambivalent medicalization and clinician subjectivity in pain management. Cultur Anthropol 2012;27:689–712. 3 Tait RC, Chibnall JT, Kalauokalani D. Provider judgments of patients in pain: seeking symptom certainty. Pain Med 2009;10:11–34. 4 Farrell M, Wodak A, Gowing L. Maintenance drugs to treat opioid dependence. BMJ 2012;344:e2823. 5 Australian Institute of Health and Welfare 2014. National opioid pharmacotherapy statistics 2013. Drug treatment series no. 23. Cat. no. HSE 147. Canberra: the Australian Institute of Health and Welfare. 6 Australian Institute of Health and Welfare. Medical professionals in Australia in 2013. Australian Government; 2014. Australian Institute of Health and Welfare. http://www.aihw.gov.au/workforce/medical (accessed 3/10/2014) 7 Holliday S, Magin P, Oldmeadow C, et al. An examination of the influences on new south wales general 8 Kircher S, Zacny J, Apfelbaum SM, et al. Understanding and treating opioid addiction in a patient with cancer pain. J Pain 2011;12:1025–31. 9 Arjan GJB, Stijn B, Paul MA, Smith RM, David R. Opioid use after fracture surgery correlates with pain intensity and satisfaction with pain relief. Clin Orthop Relat Res 2014;472(8):2542–9. 10 King NB, Fraser V, Boikos C, Richardson R, Harper S. Determinants of increased opioid-related mortality in the United States and Canada, 1990–2013: A systematic review. Am J Public Health 2014;104(8): e32–42. 11 Pain Australia. National Pain Strategy. Available at: http://www.painaustralia.org.au/the-national-painstrategy/national-pain-strategy.html; 2011 (accessed March 2012). 12 Franklin GM. Opioids for chronic noncancer pain: A position paper of the American Academy of Neurology. Neurology 2014;83(14):1277–840. 13 Holliday S, Magin P, Dunbabin J, et al. An evaluation of the prescription of opioids for chronic non malignant pain by Australian general practitioners. Pain Med 2013;14:62–74. 14 Krebs EE, Bergman AA, Coffing JM, et al. Barriers to Guideline-concordant opioid management in primary care—A qualitative study. J Pain 2014;15: 1148–55. 15 Blanch B, Pearson S, Haber PS. An overview of the patterns of prescription opioid use, costs and related harms in Australia. Br J Clin Pharmacol 2014;1159–66.: 16 Roxburgh A, Burns L. Accidental opioid-induced deaths in Australia 2008. Sydney: National Drug and Alcohol Research Centre, UNSW; 2012. 17 Hall WD, Farrell MP. Minimising the misuse of oxycodone and other pharmaceutical opioids in Australia. Med J Aust 2011;195:248–9. 18 Harrison CM, Charles J, Henderson J, Britt H. Opioid prescribing in Australian general practice. Med J Aust 2012;196:380–1. 19 Chou R, Deyo R, Devine B, Hansen R, Sullivan S, Jarvik JG, Blazina I, Dana T, Bougatsos C, Turner J. The Effectiveness and Risks of Long-Term Opioid Treatment of Chronic Pain. Evidence Report/Technology Assessment No. 218. (Prepared by the 1729 Downloaded from https://academic.oup.com/painmedicine/article-abstract/16/9/1720/1875845 by guest on 28 May 2020 GP trainees have to traverse multiple divergent models of opioid prescribing [21] and our findings indicate that MIMWOs present particular challenges. Trainee GPs are prescribing predominantly for PNCP, an indication lacking evidence of effectiveness or safety [19]. In our study, in only a minority of cases were opioids prescribed for acute pain, terminal care or OST, the indications which are evidencebased. Along with the nonutilization of addiction riskmitigation strategies, these findings suggest prescribing infrequently follows recommended guidelines [1,17,21,24]. Opioid maintenance for analgesia may be putting our patients at risk of toxicities including iatrogenic addiction. The prescribing associations with demographic, environmental, and geographic variables found in this study will inform QUM educational and policy interventions supporting safer opioid management. practitioners regarding the provision of opioid substitution therapy. Drug Alcohol Rev 2013;32:495–503. Holliday et al. Pacific Northwest Evidence-based Practice Center under Contract No. 290-2012-00014-I.) AHRQ Publication No. 14-E005-EF. Rockville, MD: Agency for Healthcare Research and Quality; September 2014. Avaliable at http://www.effectivehealthcare.ahrq.gov/ reports/final.cfm. 21 Hunter Integrated Pain Service. Reconsidering opioid therapy 2014:[1–6 pp.]. Hunter New England Local Health Network, Available at: http://www. hnehealth.nsw.gov.au/__data/assets/pdf_file/0007/ 76039/Reconsidering_opioid_therapy_May_2014. pdf (accessed 27/11/14). 22 United States Food and Drug Administration. FDA/ CDER Response to Physicians for Responsible Opioid Prescribing Partial Petition Approval and Denial 2013 Sept 10, 2013. Center for Drug Evaluation and Research. 23 Veal FC, Bereznicki LR, Thompson AJ, Peterson GM. Pharmacological management of pain in Australian Aged Care Facilities. Age and Ageing 2014; 43(6):851–6. 24 Gadzhanova S, Bell JS, Roughead EE. What analgesics do older people use prior to initiating oxycodone for Non-cancer pain? a retrospective database study. Drugs Aging. 2013;30:921–6. 25 Elkin K, Spittal M, Elkin D, Studdert D. Doctors disciplined for professional misconduct in Australia and New Zealand, 2000-2009. Med J Austr 2011;194:452–6. 26 National Centre for Education and Training on Addiction (NCETA). A Matter of Balance: A background discussion paper to support the development of the National Pharmaceutical Drug Misuse Strategy (NPDMS). Adelaide: Flinders University; 2011 March 2011. 27 Volkow ND, Frieden TR, Hyde PS, Cha SS. Medication-assisted therapies — Tackling the Opioidoverdose epidemic. N Engl J Med 2014;370:2063– 6. PubMed PMID: 24758595. 30 National Drug Strategy (Australia). National Pharmaceutical Drug Misuse Framework for Action (20122015): a matter of balance. Canberra, Australian Capital Territory, National Drug Strategy; 2013,iv: 81. 31 Bonevski B, Magin P, Horton G, Foster M, Girgis A. Response rates in GP surveys - Trialling two recruitment strategies. Aust Fam Phys 2011;40(6):427–30. 32 National Drug and Alcohol Research Centre. A Review of Opioid Prescribing in Tasmania: A Blueprint for the Future. Sydney: University of New South Wales; 2012. 33 Australian Bureau of Statistics. 12160 - Australian Standard Geographical Classification (ASGC). 2010. Available at: http://www.abs.gov.au/ausstats/ abs@.nsf/mf/1216.0 (accessedApril 2011). 34 Morgan S, Magin P, Henderson K, et al. Study protocol: the registrar clinical encounters in training (ReCEnT) study. BMC Family Pract 2012;13:50. PubMed PMID: doi:10.1186/147 1-2296-13-50. 35 WHO Collaborating Centre for Drug Statistics Methodology. WHO Collaborating Centre for Drug Statistics Methodology, Guidelines for ATC classification and DDD assignment 2013. Oslo, Norway: WHOCC; 2012:284. 36 Australian Bureau of Statistics. 2039.0 - Information Paper: An Introduction to Socio-economic Indexes of Areas (SEIFA). 2006. Available at: http:// www.abs.gov.au/ausstats/abs@.nsf/mf/2039.0/ (accessed april 11). 37 Britt H. A new coding tool for computerised clinical systems in primary care–ICPC plus. Australian family Phys 1997;26 Suppl 2:S79–82. PubMed PMID: 9254947. eng. 38 Britt H, Meza R, Mar CD. Methodology of morbidity and treatment data collection in general practice in Australia: A comparison of two methods. Family Pract 1996;13(5):462–7. 39 O’Halloran J, Miller GC, Britt H. Defining chronic conditions for primary care with ICPC-2. Family Pract 2004;21(4):381–6. 28 Australian Commission on Safety and Quality in Health Care, Inc. NTAG. National Quality Use of Medicines Indicators for Australian Hospitals. In: Commonwealth Department of Health and Ageing, editor. Sydney: ACSQHC; 2014:156. 40 Martell BA, O’Connor PG, Kerns RD, et al. Systematic Review: Opioid Treatment for Chronic Back Pain: Prevalence, Efficacy, and Association with Addiction. Ann Internal Med 2007;146(2):116–27. 29 Rogers KD, Kemp A, McLachlan AJ, Blyth F. Adverse selection? A multi-dimensional profile of 41 Mafi JN, McCarthy EP, Davis RB, Landon BE. Worsening trends in the management and treatment of 1730 Downloaded from https://academic.oup.com/painmedicine/article-abstract/16/9/1720/1875845 by guest on 28 May 2020 20 Kao M-CJ, Minh LC, Huang GY, Mitra R, Smuck M. Trends in ambulatory physician opioid prescription in the United States, 1997-2009. PM & R 2014;6(7): 575–5e4. eng. people dispensed opioid analgesics for persistent non-cancer pain. PLoS One 2013;8:13. Pattern of Opioid Management back pain. JAMA Internal Med 2013;173 (17):15731581. doi:10.1001/jamainternmed.2013.8992. 42 Fournier J-P, Azoulay L, Yin H, Montastruc J-L, Suissa S. Tramadol use and the risk of hospitalization for hypoglycemia in patients with noncancer pain. JAMA Intern Med 2015;175:186–93. 44 Wightman R, Perrone J, Portelli I, Nelson L, Likeability and abuse liability of commonly prescribed opioids. J Med Toxicol 2012;8:335–40. 45 Hollingshead NA, Matthias MS, Bair MJ, Hirsh AT, Impact of race and sex on pain management by medical trainees: A mixed methods pilot study of decision making and awareness of influence. Pain Med 2015;16:280–90. 46 Kapoor S, Thorn B. Healthcare use and prescription of opioids in rural residents with pain. Rural Remote Health 2014;14(3):2879. 50 Vanderlip ER, Sullivan MD, Edlund MJ, et al. National study of discontinuation of long-term opioid therapy among veterans. Pain 2014;155(12):2673–9. 51 Edlund MJ, Martin BC, Russo JE, et al. The role of opioid prescription in incident opioid abuse and dependence among individuals with chronic noncancer pain: The role of opioid prescription. Clin J Pain 2014;30:557–64. 52 Henry S, Eggly S. How much time do Low-income patients and primary care physicians actually spend discussing pain? A direct observation study. J Gen Internal Med 2012;27:787–93. 53 General Practice Education and Training Limited. Annual Report to 30 June 2013. Canberra; 2013. 54 Paulozzi L, Mack K, Hockenberry J. Vital signs: Variation among states in prescribing of opioid pain relievers and benzodiazepines-United States, 2012. Washington DC: U.S. Department of Health and Human Services, 2014. Supporting Information 47 Hirsh AT, Hollingshead NA, Matthias MS, Bair MJ, Kroenke K. The influence of patient sex, provider sex, and sexist attitudes on pain treatment decisions. J Pain 2014;15:551–9. 48 Burgess D, van Ryn M, Dovidio J, Saha S. Reducing racial bias among health care providers: Lessons from Social-cognitive psychology. J Gen Internal Med 2007;22:882–7. Additional Supporting Information may be found in the online version of this article at the publisher’s website: Table 6 Univariate associations of independent variables with initial opioid prescriptions Table 7 Logistic regression model of associations of initial opioid prescriptions 1731 Downloaded from https://academic.oup.com/painmedicine/article-abstract/16/9/1720/1875845 by guest on 28 May 2020 43 Nielsen S, Larance B, Lintzeris N, Holliday S, Vanderhaven M, Hordern A, Dunlop A, Haber P, Murnion B, Silsbury C, Johnson J, Demirkol A, Sadler C, Phung N, Burns L, Mattick R, Campbell G, Farrell M, Cohen M, Bruno R, Brown A, Degenhardt L. Pharmaceutical Opioid Dependence: Baseline characteristics from a cohort of treatment entrants National Drug and Alcohol Annual Symposium; September 8, 2014. 49 Anderson K, Stowasser D, Freeman C, Scott I. Prescriber barriers and enablers to minimising potentially inappropriate medications in adults: A systematic review and thematic synthesis. BMJ Open 2014;4(12):e006544-e.