Development of a Brief Patient-Administered Screening Tool for Prescription Opioid Dependence for Primary Care Settings

Abstract

Objective

To develop a short, patient-administered screening tool that will allow for earlier assessment of prescription opioid dependence (often referred to as addiction) in primary care settings.

Design and Setting

Cross-sectional analysis (N = 1,134) from the two-year time point of the Pain and Opioids IN Treatment (POINT) cohort was used in the scale development.

Subjects

Participants who completed two-year interviews in the POINT study, a prospective cohort study that followed people with chronic noncancer pain over a five-year period, and who were prescribed strong opioids for a minimum of six weeks at baseline.

Methods

An advisory committee provided advice on wording and content for screening in primary care settings. Univariate logistic regression identified individual items that were significantly associated with meeting ICD-11 criteria for prescription opioid dependence. Exploratory and confirmatory factor analysis (EFA and CFA) were conducted, and items were reduced to identify a small item set that were discriminative and shared a simple underlying structure.

Results

Sixty-four variables associated with ICD-11 criteria for prescription opioid dependence were initially identified. Four rounds of EFA were performed, resulting in five items remaining. CFA identified two possible four-item combinations, with the final combination chosen based on greater item endorsement and the results of goodness-of-fit indices.

Conclusions

Addressing prescription opioid dependence is an important part of the global public health challenge surrounding rising opioid-related harm. This study addresses an important initial requisite step to develop a brief screening tool. Further studies are required to validate the tool in clinical settings.

Introduction

There have been considerable increases in opioid prescribing, particularly in high-income countries including the United States, Canada, and Australia [1]. This marked increase in prescription opioid use is associated with considerable related harms, including mortality and morbidity [2–6], as well as prescription opioid dependence and use disorder, also referred to as addiction [7,8].

The United States appears to be worst affected; there are growing concerns with rising rates of dependence and rapidly increasing mortality [9–11]. More than one in three people in the United States were prescribed an opioid in 2015, and an estimated one in three people prescribed opioids for chronic noncancer pain (CNCP) met Diagnostic and Statistical Manual of Mental Disorders (DSM)-5 criteria for opioid use disorder [12, 13]. In 2017, there were 58 opioid prescriptions for every 100 Americans, and 17% of the population was prescribed an opioid in the past year [14]. Overdose deaths involving prescription opioids were five times higher in 2017 than in 1999 [14]. Similar trends have also been observed in Australia, where between 2013 and 2017 more than 3 million adults were prescribed opioids annually, and 1.9 million adults were initiated on opioids per year [15]. Increasing treatment demand and opioid-related mortality have been reported [16, 17].

To identify those at risk of opioid-related harm, a number of screening tools have been developed, including the Current Opioid Misuse Measure (COMM) [18], Opioid Risk Tool (ORT) [19], Screener and Opioid Assessment for Patients with Pain (SOAPP) [20], Prescription Opioid Misuse Index (POMI) [21], and the Opioid Related Behaviors In Treatment (ORBIT) [22]. Although several of these screening tools (e.g., SOAPP, COMM, ORT, ORBIT) identify aberrant drug-related behavior among patients prescribed opioids, few are specifically intended to identify opioid use disorders or dependence. For example, the SOAPP is intended to predict risk before opioid prescribing, whereas the COMM [18] and the briefer COMM-9 [23] are intended to detect aberrant behaviors. Structured clinical assessments for opioid dependence are not brief enough to allow for routine use in primary care settings.

One tool that does specifically measure opioid dependence is the Rapid Opioid Dependence Screen (RODS) [24], an eight-item tool that was designed for use in clinical and research settings. The first item in the tool assesses lifetime use of eight types of opioids (such as heroin, methadone, buprenorphine, oxycontin, and other opioid analgesics), whereas the remaining questions measure physiological, behavioral, and cognitive factors associated with opioid use [24]. Although the RODS has been shown to be a relatively quick and effective tool for measuring opioid dependence, it is limited by the fact that it is validated among a small sample (N = 97) of newly incarcerated, HIV-positive individuals and therefore its generalizability to other populations is unknown. Furthermore, the language used in this tool, for example, feeling “dope sick,” may have negative connotations, and this language lacks relevance for those using prescription opioids for chronic pain (vs illicit opioids).

The Severity of Dependence Scale (SDS) is another widely used tool that can be adapted for and used to measure dependence in various substances including opioids [25]. Although it has been validated as a dependence measure for alcohol, benozodiazepine, cocaine, amphetamines, cannabis, etc. [26–31], it has not been specifically validated for prescription opioids. Further, as the SDS was developed to assess illicit opioid dependence, the questions center only on aspects of illicit substance use and do not collect information that is salient to dependence in the context of therapeutic use, such as escalating doses and associated adverse effects.

There appears to be a need for a specific screening tool for dependence on prescription opioids with content and language that are relevant to patients receiving opioids for CNCP. Given the time pressures of primary care environments, an ideal tool would be brief, patient-administered, and facilitate targeting of more detailed assessment where needed, while accurately identifying those who are unlikely to develop dependence on their prescribed pain medication. Such a tool may provide valuable information in addition to clinician judgment and prescription monitoring systems and allow for earlier identification of emerging concerns. The current study therefore aims to develop a short, patient-administered tool that will allow for earlier targeted assessment of prescription opioid dependence in primary care settings such as general practice and community pharmacy settings.

Methods

Advisory Committee

The need to develop a brief patient-administered screening tool was established through an advisory committee meeting attended by health care professionals from primary care settings (general practice and community pharmacy settings), pain and addiction experts, a consumer representative (CNCP consumer), and implementation scientists. The advisory committee was formed for a study that aimed to implement routine screening for prescribed opioid dependence in primary care settings, initially focusing on community pharmacy [32]. The advisory committee reviewed existing tools including the SDS [25], Prescribed Opioid Difficulties Scale (PODS) [33], and a codeine dependence scale [34] and came to a consensus on the language and questions that would be optimal and identified features of existing tools that made them unsuitable for all primary care settings for patients prescribed opioids for CNCP. Despite the brevity of the SDS, the committee felt the content of the questions was not relevant enough for routine use for all patients prescribed opioids for CNCP and that the focus on nonmedical use may be a barrier to use in a therapeutic setting. The advisory committee felt that a purpose-designed tool was required. Feedback on the codeine dependence scale indicated that questions that focused on aspects of medication use were perceived to be most relevant and acceptable, and this feedback guided the choice of questions throughout the study.

Participants and Procedure

Data from the Pain and Opioids IN Treatment (POINT) cohort were used in the scale development. The POINT study is a prospective cohort study (N = 1,514) that followed people with CNCP over a five-year period and who were prescribed strong opioids for a minimum of six weeks at baseline. Additional inclusion criteria included being 18 years or older, competent in English, mentally and physically able to complete telephone and self-complete interviews, without memory or comprehension difficulties, living with CNCP (defined as pain present daily for a period of three months or more), and prescribed a Schedule 8 opioid (an Australian classification of medicines that are subject to additional regulatory controls regarding their manufacture, supply, distribution, possession, and use) [35]. A history of injecting drug use was not an exclusion criterion; however, those currently prescribed opioids for opioid substitution therapy for heroin dependence were not eligible, nor were those taking opioids for cancer pain. Recruitment of patients was via community pharmacies across Australia. Participating pharmacists approached people who were prescribed a Schedule 8 opioid for CNCP for a period of more than six weeks. Potentially eligible participants were screened by a research assistant. Upon meeting eligibility criteria, participants received study information and were invited to provide informed consent, and those who consented were scheduled for the baseline interview.

Full details of the cohort protocol [36] and characteristics of the cohort are published elsewhere [8, 37]. The current study only used existing data from the two-year wave of the cohort, where interviews were conducted with 1,278 participants (84% of the baseline cohort). The two-year time point was chosen as there were additional items included to measure opioid dependence that were not captured at other time points. Of these, 144 participants had not used opioids in the past 12 months and were excluded from the analyses, leaving a sample size of 1,134. The sample comprised 43% males, had a median age (interquartile range [IQR]) of 60 (51–69) years, and 43% were unemployed. The sample had been prescribed opioids for a median (IQR) of seven (4–13) years and most commonly reported back or neck pain (85%), arthritis/rheumatism (76%), or frequent headaches or migraines (33%), with most participants (87%) reporting more than one pain condition. One in 10 met International Classification of Diseases (ICD)-10 criteria for (9.4%) prescription opioid dependence, and one in five met DSM-5 criteria for opioid use disorder (18%) [7]. In the cohort, around one in three reported benzodiazepine use [38], and around half used antidepressants [39].

Participants were reimbursed $50 for their time for participation in the two-year interview. Ethical approval was obtained from the Human Research Ethics Committee of the University of New South Wales (HREC Reference: #HC12149). Additional information pertaining to the recruitment procedures and related methods, as well as the POINT study sample, has been published elsewhere [8,36].

Measures

Potential items for the screening tool were identified from the two-year questionnaire. Individual items were identified from the following scales and measures:

Opioid-Related Behaviors in Treatment Scale

The ORBIT is a 10-item measure of aberrant behaviors such as doctor shopping, diversion, and other examples of unsanctioned use of medications in the past three months. Items are scored on a five-point scale from never (0) to very often (4), where item scores are summed, with higher scores indicating greater or more frequent aberrant behaviors [22].

Prescribed Opioid Difficulties Scale

The 16-item PODS examines patient concerns and difficulties associated with their opioid medication, including side effects and concerns about use. Items are scored on a five-point scale from 0 to 4, where higher scores indicate that patients attribute more problems to their use of opioid medications [33].

Severity of Dependence Scale

The SDS assesses the severity of dependence to a range of substances using a four-point scale from 0 to 3. Item scores are summed, and higher scores indicate higher levels of dependence [25]. The scale was initially developed to assess severity of illicit drug dependence; however, for the purposes of the current study, items were adapted to assess dependence to prescription opioids.

Composite International Diagnostic Interview 3.0

The Composite International Diagnostic Interview 3.0 (CIDI) was used to assess prescription opioid dependence [40]. All participants who had used prescription opioids in the preceding 12 months completed these questions. The CIDI has been extensively used in epidemiological studies in many countries [41–44] and has been shown to have excellent inter-rater reliability [40], test–retest reliability [40], and agreement with clinician diagnoses [45].

Current and Recent Prescribed Medication and Substance Use

Questions were developed for the longitudinal cohort study to explore if participants had self-regulated their opioid use (e.g., increased or decreased dose) and if participants had used a range of other substances in the past 12 months including nicotine and cannabis [36].

Opioid Dose

Participants completed a seven-day medication diary for all pain, psychiatric, and sleep medication, including over-the-counter medications that they had used, including the dose taken for each day. Daily oral morphine equivalent (OME) opioid doses taken by the cohort were estimated using conversion units established through review and synthesis of a range of clinical guidelines [46].

Adverse Events

Participants were asked about possible side effects they may have experienced when taking their opioid medication (e.g., nausea, vomiting, drowsiness, and fatigue).

Pain Self-Efficacy Questionnaire

Pain self-efficacy relates to a person’s beliefs about the extent to which they can carry out daily activities in the presence of pain. The Pain Self-Efficacy Questionnaire (PSEQ) is a 10-item scale to assess pain self-efficacy at the present point [47, 48].

Alcohol Use Disorders Identification Test

The 10-item Alcohol Use Disorders Identification Test (AUDIT) is a screening tool developed by the World Health Organization (WHO) to assess alcohol consumption, drinking behaviors, and alcohol-related problems [49, 50].

Statistical Analysis

One researcher (SN) identified variables that were conceptually and broadly related to substance use or dependence. Univariate logistic regression identified 64 individual items that were significantly associated with meeting ICD-11 criteria for prescription opioid dependence (dichotomous variable). Dependence criteria were initially considered, as they are narrower than the broader DSM-5 opioid use disorder criteria.

Exploratory and Confirmatory Factor Analysis

Exploratory factor analysis (EFA) was performed in MPLUS, version 7.4, to explore the underlying structure of the 64 candidate items, with the aim of identifying a small number of discriminative items that loaded on a single factor so that a total score would be appropriate (Figure 1). After each round, items were removed based on statistical parameters (e.g., poor fit) and redundancy, rarity, or wording that was either obviously related to dependence or deemed inappropriate for a primary care setting, which was done via consultation between authors (SN and RB), guided by a previous discussion with the study advisory committee. Subsequently, confirmatory factor analysis (CFA) was used to determine the fit of the single-factor structure for the reduced set of items using polychoric correlations with weighted least squares with the mean- and variance-adjusted chi-square (WLSMV) estimator. Goodness of fit was determined by three criteria: Comparative Fit Index (CFI) >0.95 [51], root mean square error approximation (RMSEA) ≤0.06 [52], and Tucker-Lewis Index (TLI) >0.90 [53].

Figure 1.

Stages of item reduction from the initial 575 items.

Cutoff Scores

Receiver operating characteristics (ROC) curves were used to determine the best cutoff for the screening tool against commonly used diagnostic criteria (ICD-10 and -11 prescription opioid dependence, ICD-10 any prescription opioid use disorder, and DSM-5 any prescription opioid use disorder). The area under the ROC curve values were interpreted and used to define the accuracy of the cutoff point according to the following thresholds: 0.5–0.7 as “poor,” 0.7–0.8 as “fair,” 0.8–0.9 as “good,” and >0.9 as “excellent” [54].

Results

Univariate Analysis

Candidate items were subjected to univariate logistic regression analyses to confirm they were significantly associated with meeting ICD-11 criteria for prescription opioid dependence (dichotomous variable). Results revealed that 64 variables were significantly associated, and these were included in further analyses (Supplementary Figure 1).

Exploratory Factor Analysis

In total, four rounds of EFA were performed, with the aim of identifying a small set of items that were discriminative and shared a simple underlying structure, while also identifying the best combination of items for a brief tool to measure prescription opioid dependence. The first round included the 64 items that were significant at the univariate stage and revealed six items for removal due to either perfect correlation between items demonstrating redundancy between items (N = 1) or items that were too obviously related to symptoms of dependence (N = 5). The latter items were seen as a barrier to routine screening, based on feedback from the project advisory committee, and were therefore removed. The subsequent round of EFA, examining factor structures of one to five factors, included 58 items and identified a further 26 items for removal due to high uniqueness or substantial cross-factor loading, in addition to items from an existing scale (PSEQ) all grouping together, separating itself from the other items. The remaining 32 items were subjected to a third round of EFA, where four additional items were removed due to uniqueness (demonstrated by low loading on the single-factor structure, as well as low loadings on the two- and three-factor structures). One additional item was removed due to extremely low base rate (leading to a Heywood case in some models) and concerns that the content was too obviously related to dependence to use in the scale. The fourth and final round of EFA resulted in a further 22 items being removed, leaving five items. These 22 items were removed, as the aim was to identify an item set with shared loadings on a single underlying construct. To this end, items that had poor loadings on the primary factor or readily split into multiple factors were removed. In addition, a final careful review of item content (by SN and RB) was used to remove further items due to the following reasons: clearly related to dependence and therefore a barrier to routine use in the primary care setting (e.g., “have friends or family thought that you were addicted?”); considered less salient in a pain population (e.g., “I wished that I could stop using medication”); conceptual overlap and hence were redundant (e.g., “that medications caused mental cloudiness vs that I had problems thinking clearly”; or “used more pain medication than prescribed vs needed more medications than prescribed”) (Supplementary Figure 2).

Confirmatory Factor Analysis

CFA was then performed on the remaining five items; however, this revealed that these five items did not load well onto a single-factor structure, but rather identified two possible four-item combinations (Option A and Option B) to have the best fit, where each option shared three of the same items but differed by one additional item. The Supplementary Data show the two possible combinations of these four items and the proportion of the sample that responded positively to each item. Both Options A and B were tested, and Option B was chosen for two reasons: One item in Option A (“In the past three months, I have used my opioid medicines for other purposes, for example, to help me sleep or to help with stress or worry”) had low endorsement (4.1%) and was viewed as less useful to include in the screening tool, and the results of goodness of-fit indices also indicated that Option B was optimal (CFI = 0.997, RMSEA = 0.021, TLI = 0.990).

Screening Efficiency of the Scale

To further test the predictive ability to identify prescription opioid dependence, the sensitivity and specificity of the sum score for these final four items (i.e., Option B) were determined for identifying cases of CIDI-defined ICD-11 dependence using ROC analysis. To facilitate ease of scoring in the intended settings for use (i.e., primary care and pharmacy), dichotomous scoring was used. “Not at all” was scored as 0, whereas the remaining three response options endorsing each of the domains to different degrees were scored as 1. The area under the curve was 0.78 (95% confidence interval = 0.74–0.80), against the ICD-11 dependence criteria, which suggests that the scale has a “good” ability to discriminate between dependence cases and noncases. A cutoff point of 2 was chosen as a pragmatic proposed cutoff, as it had adequate sensitivity (62%), good specificity (80%), and was correctly classified in more than 75% of cases. Using the same cutoff, the sensitivity and specificity against the ICD-10 dependence criteria were 77.17% and 76.57%, respectively (Table 1).

Table 1.

Diagnostic sensitivity of Option A and Option B of the OWLS screening tool against different diagnostic criteria using ROC analysis

	ICD-10 Any Disorder		ICD-10 Dependence		ICD-11 Dependence		DSM-5 Any Disorder
Option A	Sensitivity	Specificity	Sensitivity	Specificity	Sensitivity	Specificity	Sensitivity	Specificity
≥1, %	80.08	57.41	86.67	52.94	77.73	55.57	80.41	55.72
≥2, %	51.87	87.76	65.71	83.77	50.24	86.02	53.61	86.32
≥3, %	21.16	96.24	35.24	95.33	26.07	96.82	25.77	96.64
≥4, %	4.56	99.76	8.57	99.59	5.21	99.77	5.15	99.78
AUC	0.75 (0.73–0.78)		0.80 (0.77–0.82)		0.73 (0.71–0.76)		0.75 (0.73–0.78)
Option B
≥1, %	89.14	46.92	96.74	42.57	88.83	45.39	90.40	45.43
≥2, %	63.35	81.54	77.17	76.57	61.70	79.34	64.97	79.78
≥3, %	29.41	93.72	46.74	92.19	36.17	94.34	36.16	94.15
≥4, %	11.31	98.72	21.74	98.35	15.43	99.26	14.12	99.03
AUC	0.78 (0.75–0.80)		0.83 (0.81–0.86)		0.77 (0.74–0.80)		0.78 (0.76–0.81)

Option A results are included for comparisons purposes.

AUC = area under the curve; DSM-5 = Diagnostic and Statistical Manual of Mental Disorders; ICD-10 = International Classification of Diseases; OWLS = Overuse (taking more than prescribed), Worrying about use of opioid medicines, Losing interest in usual activities due to opioid medicines, and feeling Slowed down, sluggish, or sedated due to opioid medicine; ROC = receiver operating characteristics.

To facilitate administration as a single scale, minor modifications to these items were subsequently made to ensure consistency across time frames (past three months), consistency across response options (“not at all,” “a little,” “quite a lot,” and “a great deal”), and to use the same language to refer to prescription opioids in each to the questions, so they are consistently referred to as “opioid medicines” (see Figure 2 for finalized wording and the Supplementary Data for original wording).

Figure 2.

OWLS screening tool for prescription opioid dependence.

Discussion

The current study has utilized data from a prospective cohort of people with CNCP who were prescribed strong opioids in order to develop a brief patient-administered screening tool, which can be used in primary care settings to identify patients who would benefit from a detailed assessment for prescription opioid dependence. Existing tools that measure opioid dependence are limited by their wording, length, and emphasis on illicit opioid use. In addition, few tools have been validated for use in patients using prescription opioids for CNCP, validated against a gold standard such as diagnostic interviews for opioid dependence, or developed with language that is intended to be acceptable for this population. Accordingly, attention was given to ensure the wording of specific items focused on opioids used as prescribed medicines, as opposed to an illicit substance, so as to lend itself to use within a primary care setting, ultimately addressing an existing gap.

This newly developed tool captures four aspects of opioid use, Overuse (taking more than prescribed), Worrying about use of opioid medicines, Losing interest in usual activities due to opioid medicines, and feeling Slowed down, sluggish, or sedated due to opioid medicine, abbreviated to “OWLS.” On their own, each domain is relevant to clinical care, in addition to identifying patients who would benefit from further assessment for opioid dependence. The OWLS is designed to be a quick and effective method to aid clinical decision-making and is not designed as a diagnostic measure. It is a patient-administered screening tool, which has additional advantages, including being a fast and efficient way to collect patient information. Self-complete also assists confidentiality, as this method removes the need for patients to disclose sensitive information verbally, addressing consumers’ expressed preferences for privacy and confidentiality in settings such as community pharmacy [55]. A range of brief (two to four questions) screening tools have been used as prescreens before more detailed assessment in the context of Screening, Brief Intervention and Referral to Treatment (SBIRT) programs, which are commonly used in the United States for drug and alcohol use [56]. This highlights the role for brief screening tools to identify where more detailed assessment is warranted. As SBIRT services offer a mechanism for remuneration in some countries, this may support the feasibility of implementing such screening more broadly.

The OWLS screening tool uses a time frame of the past three months, with response options comprising “not at all,” “a little,” “quite a lot,” and “a great deal,” consistent with the response options used with the SDS [25]. The OWLS has been shown to have optimal goodness-of-fit indices in addition to an acceptable predictive ability, when compared with ICD-11 dependence, ICD-10 dependence, ICD-10 any disorder, and DSM-5 any disorder diagnostic criteria. The tool displayed greater specificity than sensitivity across all diagnostic criteria, with the exception of ICD-10 dependence, where the sensitivity (77.17%) and specificity (76.57%) were almost the same.

A cutoff of 2 offers a pragmatic indication of the need for a detailed assessment of prescription opioid dependence; however, an endorsement of any item could be considered clinically relevant and lead to a discussion of that symptom. It is important to acknowledge that the sensitivity and specificity of the OWLS screening tool with a cutoff of 2 were 62% and 80%, respectively (against the ICD-11 dependence criteria). As such, it is possible that the tool may miss some positive cases and will identify some positive cases that are in fact false positives (in other words, identify people for further assessment who are not dependent on prescription opioids). Accordingly, as with all screening tools, clinicians are reminded to exercise a level of caution when interpreting the scores [57]. Where patients are endorsing even a single item on the scale, this can be used as a trigger for primary care clinicians to have a conversation with their patients in order to better understand the circumstances behind the patient’s answer. This would be especially relevant where patients are reporting dose escalation or concerns about their own opioid use. Further discussion in these cases would represent good clinical care and enhance the sensitivity of the scale, though use of a lower cutoff (i.e., one) would result in more false-positive cases, meaning that clinicians would be spending time doing detailed assessments unnecessarily. These needs are particularly important to balance in a primary care setting where there are multiple competing demands on health care professionals’ time. Where a clinician has a preference to avoid missing any cases, a conservative cutoff of 1 could be utilized.

A systematic review of principles for population-based screening decisions identified a number of principles for screening, including the need to understand the condition and target population, the importance of the condition, and ensuring that appropriate treatment is available and that screening is fit for the purpose, acceptable, and cost-effective [58]. These principles support the implementation of the routine use of the OWLS screening tool for prescription opioid dependence within primary care settings. Given the increasing prevalence of prescription opioid dependence [59], coupled with the availability of effective treatment [60], use of screening tools designed specifically for use within primary care settings may enable a more proactive response and reduce harms from opioid dependence arising from therapeutic use. Further research may consider if there are potential applications outside primary care settings also. Efforts across multiple settings to identify and treat prescription opioid dependence early are likely to reduce related harms such as overdose, psychosocial deterioration, and medical complications [61].

The following limitations should be considered when interpreting the findings. The existing data set comprised a sample of CNCP patients who were prescribed strong opioids. As such, it is not clear how findings relate to those who have been prescribed weaker opioids, those who have been taking opioids for shorter periods of time, or those using opioids for acute pain. Therefore, these findings may not be generalizable to all people who take prescription opioids. It is also important to consider the subjective nature of the screening tool and acknowledge that accurate and truthful reporting may not always occur, especially where patients may be concerned about the repercussions of reporting overuse or side effects with medications. It is possible that DSM-5 and ICD-10/11 criteria considering symptoms over the past 12 months and our questions addressing the past three months would affect the sensitivity of our measure, meaning that our results would be more conservative and may not identify dependence in the absence of recent symptoms. This in unlikely to affect its clinical utility, as recent symptoms may be most relevant but could lead to differences in screening results compared with those of a diagnostic assessment over a 12-month time period. Finally, we took the deliberate approach of choosing questions that measured dependence indirectly, a strategy chosen to facilitate health care professionals’ comfort with screening. This trade-off was considered acceptable given the barriers to asking such questions, but it has the disadvantage of potentially offering less specificity than a tool that probes about symptoms of dependence more directly. Future research should validate the tool for use in clinical settings, when patient-administered, and among a broader population of people who are prescribed both strong and weak opioids, while also exploring the impact of different cutoff scores and stability over time.

These limitations notwithstanding, the current study included a large national sample, which was used to develop a brief screening tool for prescription opioid dependence intended for routine use in primary health care settings. Given the common use of prescription opioids among CNCP patients [59, 62], in conjunction with the potential harms that are associated with their long-term use [2–6], regular screening to monitor for emerging opioid dependence is recommended [63]. Addressing prescription opioid dependence is an important part of the global public health challenge surrounding rising opioid overdoses, and this study addresses an important initial requisite step in the form of a brief and appropriate screening tool. Further research will establish how well this tool may perform in clinical use.

Funding sources: Funding was received to support the screening tool development from the Central and Eastern Sydney Primary Health Network. The POINT cohort study received funding from the Australian National Health and Medical Research Council (NHMRC; #1022522). G. Campbell (#1119992), L. Degenhardt (#1135991, #1041742), and S. Nielsen (#1132433, #1163961) are supported by NHMRC research fellowships. L. Degenhardt is supported by National Institute on Drug Abuse R01DA1104470. The National Drug and Alcohol Research Centre at UNSW Sydney is supported by funding from the Australian Government Department of Health under the Drug and Alcohol Program.

Disclosure: Suzanne Nielsen reports grants from Indivior and Seqirus unrelated to the current work. SN has delivered training on opioid dependence for Indivior for which honoraria were paid to her institution. SN has participated in an advisory board meeting Mundipharma relating to intranasal naloxone (sitting fee not taken). Louisa Picco: no conflict. Gabrielle Campbell: no conflict. Nicholas Lintzeris has received research grant funding from Indivior and Braeburn and consultancies or advisory board participation from Indivior and Mundipharma, all outside the submitted work. Briony Larance reports grants from Indivior for studies of buprenorphine depot and an untied educational grant from Seqirus for studies of tapentadol, all outside the submitted work. Michael Farrell received grants from Indivior for studies of buprenorphine depot and an untied educational grant from Seqirus for studies of tapentadol, all outside the submitted work. Louisa Degenhardt received grants from Indivior for studies of buprenorphine-naloxone and buprenorphine depot and an untied educational grant from Seqirus for studies of tapentadol, all outside the submitted work. Raimondo Bruno reports grants from Indivior outside the submitted work.