Addiction-Related Assessment Tools and Pain Management: Instruments for Screening, Treatment Planning, and Monitoring Compliance

Transcription

1 PAIN MEDICINE Volume 9 Number S ORIGINAL ARTICLE Addiction-Related Assessment Tools and Pain Management: Instruments for Screening, Treatment Planning, and Monitoring Compliance Steven D. Passik, PhD,* Kenneth L. Kirsh, PhD, and David Casper, BA* *Memorial Sloan-Kettering Cancer Center, and Cornell University Weill College of Medicine, New York, New York; Department of Pharmacy Practice and Science, College of Pharmacy, University of Kentucky, Lexington, Kentucky, USA ABSTRACT Objective. To review and critique the various assessment tools currently available to pain clinicians for assessing opioid use and abuse in patients with chronic noncancer pain to allow pain clinicians to make informed selections for their practices. Methods. A literature search on PubMed was conducted in June 2006 using the search terms opioid plus screening or assessment with or without the additional term risk, and opioid-related disorders/ prevention and control in order to identify clinical studies published in English over the previous 10 years. Additional studies were identified using the PubMed link feature and Google. When abstracts described or referred to a tool for opioid abuse screening, the corresponding publication was acquired and reviewed for relevance to the pain treatment setting. Results. Forty-three publications were selected for review from the abstracts identified, and 19 were rejected because they did not describe a specific tool or provide adequate information regarding the screening tool used. The remaining 24 publications described relevant screening tools for opioid abuse risk and were reviewed. Conclusions. A variety of self-administered and physician-administered tools differing in their psychometrics and intended uses have been developed, but not all have been validated for use in chronic pain patients seen in a clinical practice setting. Some tools assess abuse potential in patients being considered for opioid therapy, whereas other tools screen for the presence of substance abuse. By recognizing the psychometrics of each tool, clinicians can select the ones most appropriate for their patient population and screening needs. Key Words. Analgesics; Opioid; Pain Measurement; Substance Abuse Detection Reprint requests to: Steven D. Passik, PhD, Memorial Sloan- Kettering Cancer Center, 641 Lexington Avenue, New York, NY 10022, USA. Tel: ; Fax: ; passiks@mskcc.org. Introduction In the early days of the movement that has broadened the use of, and indications for, opioid therapy, it was common to trivialize the potential for substance abuse when these medications were being used for pain. This trivialization sometimes took the form of rhetoric, suggesting that the risk in the population of potential patients to be exposed is low or nonexistent [1]. Further, there were miscalculations made about how much control over addiction or abuse the use of certain delivery systems might bestow. Specifically, there was an overestimation of the protective effects of long-acting delivery systems. This overestimation of the protective effects of long-acting American Academy of Pain Medicine /08/$15.00/S145 S145 S166 doi: /j x

2 S146 delivery systems took the form of rhetoric, suggesting that these systems could lower or eliminate risk [1]. It has become clear that if a broader swath of the population of those with chronic pain is to be exposed to opioids, the risk of addiction, abuse, and aberrant behaviors is at least as high as it is in the general population. Risk is higher still if the pain is related to disease states wherein substance abuse is a common comorbidity [2]. In fact, the risk of addiction and abuse is related to the interaction between the rewarding properties of these drugs (which also vary according to their differing ability to deliver the drug rapidly) and the personal vulnerabilities in individual patients, and the degree to which this interaction can be contained within certain treatment settings. There is now an evolving consensus that safe opioid prescribing hinges on risk stratification and the accommodation of that risk into a treatment plan. Pain clinicians need to become proficient in performing and documenting a risk assessment. The use of a screening or another assessment tool fulfills the growing requirement for due diligence in the area of screening for the individual patient s vulnerabilities and risk, and incorporating the results of the assessment in treatment planning (drug selection and the degree of safeguards to be built into the setting). Additionally, the use of validated tools not only helps guide the assessment, but when incorporated into the medical record, also upgrades the clinician s documentation of this assessment. However, until recently, there was a lack of empirical studies on addiction-related aspects of chronic pain management in general, and even fewer devoted to creating validated tools for assessing addiction-related risk and outcomes in pain patients being considered for, or being managed on, opioid therapy. Over the last few years, however, there has been a significant increase in the number and types of addiction-related tools that are now available to pain clinicians. The existing tools vary in length, focus, and psychometric sophistication. They also vary with regard to the subgroup of pain patients in which they were validated for example, a pain clinic subgroup vs a primary-care-based pain subgroup vs populations being screened for addiction in the absence of pain. Although no standard of care exists for how addiction-related assessments should be performed (i.e., structured vs unstructured; measurement tool vs clinical interview), it is clear that the standard of care is fast becoming one in which some form of risk stratification of patients being considered for Passik et al. opioid therapy must be carried out and documented, especially in light of the growing problem of prescription drug abuse. Many pain clinicians are struggling to find addiction-related assessment tools that they can incorporate into their practice settings and medical record keeping. As these tools continue to be developed, clinicians may find it difficult to choose for their pain patients the best tool that fits into the treatment setting. The challenge relates to the now large number of measures and the many differences among them. Dimensions in which screening tools differ include their mode or ease of administration, their psychometrics, the population best suited for each tool, and the aspects of addiction each tool is meant to monitor or predict. In this supplement, we intend to clarify these aspects for each of a number of assessment tools (Table 1). After reading this article, pain clinicians will be able to make better-informed decisions on the best-suited assessments to use, which in turn should enable treatment more tailored to each patient s pain/risk level. Because this risk is higher in noncancer pain populations, and because cancer pain patients represent a unique population, the focus of this review is noncancer pain [3 5]. Table 1 Assessment tools reviewed (page # for textual discussion of tools) 1. ADDIS: alcohol and drug diagnostic instrument (S163) 2. ASI: addiction severity index (S161) 3. Atluri six-point screening tool (S155) 4. CAGE/CAGE-AID: cut down, annoyed, guilty, eye-opener/ adjusted to include drugs (S159) 5. CCI: chemical coping inventory (S156) 6. Chabal five-point prescription opiate abuse checklist (S161) 7. COMM: current opioid misuse measure (S162) 8. CUAD: chemical use, abuse, and dependence scale (S159) 9. DAPA-PC: drug abuse problem assessment for primary care (S156) 10. DAST: drug abuse screening test (S158) 11. DIRE: diagnosis, intractability, risk, efficacy (S154) 12. KMSK scale: Kreek-McHugh-Schluger-Kellogg scale (S157) 13. ORT: opioid risk tool (S154) 14. PDUQ: prescription drug use questionnaire (S162) 15. PMQ: pain medication questionnaire (S161) 16. POSIT: problem-oriented screening instrument for teenagers (S161) 17. RAFFT: relax, alone, friends, family, trouble (S160) 18. SASSI: substance abuse subtle screening inventory (S158) 19. SCID-P: alcohol and drug sections of the DSM-III-R (S163) 20. SISAP: screening instrument for substance abuse potential (S155) 21. SMAST-AID: short Michigan alcoholism screening test/adapted to include drugs (S160) 22. SOAPP: screener and opioid assessment for patients with pain (S148) 23. STAR: screening tool for addiction risk (S155) 24. SUDDS: substance use disorder diagnostic schedule (S163) 25. TICS: two-item conjoint screen (S156)

3 Screening Tools of Opioid Abuse The introduction of addiction-related assessment tools into a pain management setting requires skill and sensitivity on the part of the clinician. On the one hand, many patients who are truly at low risk for abuse are, ironically, tremendously fearful of becoming addicted to opioids [6]. Detailed questioning about addiction via the utilization of these tools could heighten anxiety in such patients. In such instances, clinicians can use the screening process as a way of reassuring patients that they indeed do not have a history suggestive of risk. On the other hand, many patients will be fearful that admissions of substance use or abuse might lead to their exclusion from pain treatment with opioids. Given that almost all these tools are susceptible to deception, the subject s incentive to lie may be quite high. The clinician will need to introduce the screening process to patients as a way of helping to plan for safe opioid treatment, assuring them that no set of answers will automatically lead to the avoidance of appropriate pain management. Patients with a history of addiction can be reassured that safeguards will be applied to their treatment that may help them to avoid becoming out of control in their opioid use. To complicate this dynamic, there is a problematic middle ground of patients sometimes referred to as pseudoaddicts. Pseudoaddiction (the problem of potentially difficult behavior set in motion by poorly treated pain) is not easily dealt with in the adoption of these tools in pain management settings; this is particularly true of those instruments developed mainly for use in the treatment for drug addiction [7]. Certain behaviors and even illicit drug use are not unusual in the face of uncontrolled pain in some patient populations. Screening and assessment tools are better at recording behaviors than assessing the driving forces behind them. None of the existing tools have a pseudoaddiction correction. As will often be observed in this article, a validation trial for a given measure in a pain population is needed. In the absence of a correction or an assessment of the patient s intent when he or she is engaging in certain problematic behaviors, the best that clinicians can hope for is an estimate of how a given measure performs with pain patients. Methods Several methods were used to identify publications describing clinical tools used to screen patients for risk of opioid abuse. A literature search of the PubMed National Center for Biotechnology Information database was conducted on June 16, 2006, and was limited to human studies published in English. Three independent searches were performed for clinical trials published in the last 10 years using the search terms 1) opioid and (screening or assessment), 2) opioid and (screening or assessment) and risk, and 3) opioid-related disorders/prevention and control. In addition, the PubMed related articles search link for three primary studies was used to supplement the articles identified in the primary searches. Search results were printed as abstracts and were reviewed for relevance to the identified topic. In order to identify publications not indexed in PubMed, online searches were also performed using the Google search engine for the same terms as those utilized in the PubMed searches. Additionally, anecdotal information received from expert colleagues in the field of pain management was used to identify late-breaking publications regarding opioid abuse risk assessment tools. Regardless of the method used to identify the publications, upon review, if an abstract described a tool or made reference to a tool for opioid abuse screening, the corresponding publication was acquired and reviewed in full for relevance. Results S147 The three quantitative PubMed database searches identified 722, 83, and 15 publications for the opioid and (screening or assessment), opioid and (screening or assessment) and risk, and opioid-related disorders/prevention and control searches, respectively. The majority of the hits identified by this method was excluded after the abstract review, because they described no specific tool or provided inadequate information regarding the screening tool used. The number of publications identified through all search methods and subsequently reviewed during this process is shown in Table 2. After all the abstracts identified through electronic and hand methods were reviewed, 43 publications were selected for review of the full-length manuscript. Nineteen of these 43 were later rejected because they either did not describe a specific tool or did not provide adequate information regarding the screening tool used. In all, 24 publications described relevant screening tools for opioid abuse risk in patients, and these are reviewed in detail here. Table 3 lists the citations included in this review with a description of each. Table 4 provides descriptions and commentaries on each tool based on these studies.

4 S148 Passik et al. Table 2 Publication selection process Search Strategy* Measures of Tool Effectiveness The effectiveness of opioid assessment tools is characterized by their sensitivity, specificity, positive predictive value, and negative predictive value. The sensitivity of a tool reflects the proportion of persons with a condition (e.g., a substance abuse disorder) who test positive, whereas the specificity represents the proportion without the condition who test negative. The positive predictive value of a tool reflects the proportion of persons who test positive who have the condition, whereas the negative predictive value represents the proportion with negative test results who do not have the condition. A c statistic is sometimes used to assess the predictive ability of a tool by taking both sensitivity and specificity into consideration (c = 0.5 indicates no discrimination, whereas c 0.8 shows excellent discrimination). Correlation coefficients show the strength of a relationship between two variables, with a coefficient of 1.0 reflecting a perfect correlation. The Pearson r coefficient is often used to show the reliability between initial testing and retesting; Cronbach s alpha coefficient is used to estimate the internal consistency or reliability of a test; and kappa coefficients are used to show agreement between categorical variables. Risk-Assessment Instruments for Pain Populations The Screener and Opioid Assessment for Patients with Pain (SOAPP) The SOAPP is a 14-item, self-report measure that is designed to assess the appropriateness of longterm opioid therapy for chronic pain patients [8]. Each item is measured on a five-point scale (0 = never, 4 = very often), with a total score of Abstracts Retrieved Rejected after Review PubMed NCBI database searches Search 1: opioid and (screening or assessment) Search 2: opioid and (screening or assessment) and risk Search 3: opioid-related disorders/prevention and control Related Article link search from Adams et al. [33] Related Article link search from Dunbar and Katz [40] Related Article link search from Webster and Webster [11] Other publication identification methods Manual search of reference lists in review articles Google searches Anecdotal information from peers Results after qualitative and quantitative searches Total full-length publications reviewed 43 Publications excluded after review 19 Publications identified for description 24 Selected for Detailed Review * Also included in this review are recent publications describing new assessment tools deemed by the authors to be pertinent and relevant to this review. Also included publications identified in search 1. NCBI = National Center for Biotechnology Information. greater than or equal to 8 indicating a high risk of misuse/abuse. The SOAPP created by an expert panel of 26 members was reduced from an initial 24 items to 14 items after Butler et al. [9] tested each item s reliability and validity. Akbik et al. [8] found that 355 of 396 noncancer chronic pain patients (90%) answered all 14 items. (A high rate of completion is important, as partially completed tools have limited usability.) Patients with scores of 8 or higher were younger (mean age 40.7 vs 45.5 years, P < 0.05), more likely to have had a urine screen (46.4% vs 31.1%, P < 0.01), and more likely to have had abnormal urine screen results (33.7% vs 27.5%, P < 0.05) than those with scores below 8. The authors contend that the SOAPP employs a low cutoff because individuals who believe that their responses may determine their opioid treatment may underreport their behavior, and because some patients fear that their answers may be misconstrued. Whereas the SOAPP is an accurate tool for assessing abuse potential in patients being considered for opioid therapy, it remains problematic in a few areas, most notably in that the data are correlational and not causal. Another problem is that very few demographic and medical data were recorded in the validation of SOAPP, so it is difficult to know whether the validation cohort had an unusually high or low baseline risk. The SOAPP has undergone a number of iterations. It is presently briefer and perhaps less susceptible to deception than some of the more face-valid tools. The assessment is also more accurate at predicting

5 Screening Tools of Opioid Abuse S149 Table 3 Study descriptions Citation Tool Name Patient Population Study Objectives ary Akbik et al. [8] SOAPP 396 chronic noncancer pain patients prescribed opioids Belgrade et al. [10] DIRE 61 chronic noncancer pain patients treated with opioids at outpatient pain clinic Holmes et al. [32] PMQ 271 chronic noncancer pain patients newly evaluated for treatment in an interdisciplinary pain management program Schieffer et al. [41] PDUQ 288 chronic noncancer pain patients with history of current or past opiate use Webster and Webster [11] ORT 185 chronic noncancer pain patients referred to single pain clinic Adams et al. [33] PMQ 184 chronic noncancer pain patients at an interdisciplinary pain treatment center Atluri and Sudarshan [12] Six-point screening tool 210 chronic noncancer pain patients at a community pain clinic Butler et al. [9] SOAPP chronic noncancer pain patients in a hospital-based pain management center Nemes et al. [16] DAPA-PC 327 chronic noncancer pain patients presenting for care at a university health clinic Friedman et al. [14] STAR 48 chronic noncancer pain patients in a large inner-city hospital Kellogg et al. [18] KMSK Scale 100 healthy volunteers, 46 of whom met the DSM-IV criteria for alcohol, cocaine, or opiate dependence Bastiaens et al. [26] RAFFT 215 patients coming to a psychiatric emergency and evaluation center during a 1-month period Brown et al. [17] TICS 1,136 randomly selected patients from three practices of a family medicine department of a university medical school Further validate 14-item version of SOAPP in patients with chronic noncancer pain Compare patient groups at high and low abuse risk (according to SOAPP scores) Test validity and reliability of DIRE score in randomly selected cases from a database Validate PMQ scores, and assess the relationship between PMQ scores and treatment outcomes Assess the role of opiate medication beliefs in chronic pain patients response to prescribed opiate medications Introduce brief screening tool for office use to predict likelihood of opioid abuse Develop self-reporting screening tool to assess risk for aberrant behaviors in opioid medication use among pain patients Develop screening tool for distinguishing patients at risk for inappropriate prescription opioid use Develop and validate self-administered brief screening tool predictive of aberrant drug-related behavior in chronic pain patients being considered for long-term opioid therapy Examine possible differences in responses of older adults and younger adults to the DAPA-PC Evaluate STAR as a screening tool for chronic pain patients at risk of substance abuse Demonstrate that KMSK scale is a valid method for drug-dependence screening in the clinic Compare performance of RAFFT and CAGE in adults with addictive disorders Determine criterion validity of TICS for alcohol and other drug abuse or dependence for a split sample of primary care patients. Preliminary support for internal reliability and predictive validity of SOAPP Reliability analysis: Cronbach s alpha = 0.75 Reliability analysis: Cronbach s alpha = 0.80 Sensitivity and specificity for predicting patient compliance: 94% and 87%, respectively PMQ reliably predicted history of substance abuse and treatment effect No formal reliability studies Follow-up study to Adams et al. [33] Not designed to test the reliability or validity of PDUQ ORT considered predictive of probability for abusing opioids, but no formal validation tests were done Reliability: Cronbach s alpha = 0.73 Correlation found between high PMQ scores and history of substance abuse, psychosocial distress, and poor functioning Good correlation with patients at high risk for abuse Reliability and validity not certain: prospective validation studies are required ROC curve for SOAPP prediction score sensitivity and specificity had AUCs from to Formal reliability and validity testing not performed Follow-up study published: Akbik et al. [8] No formal tests of reliability or validity Similar responses by younger and older adults to DAPA-PC, except for two questions Reliability and validity testing not performed Only two questions related to smoking and one question related to prior treatment for substance abuse distinguished pain patients with and without a substance abuse history ROC curves for KMSK prediction score sensitivity and specificity had AUCs of 0.64 Formal reliability and validity testing not performed RAFFT and CAGE had similar sensitivity and specificity (good validity) No formal reliability testing reported Sensitivity of 79.3% and specificity of 77.9% for current SUD Formal reliability studies not performed

6 S150 Passik et al. Table 3 Continued Citation Tool Name Patient Population Study Objectives ary Compton et al. [35] PDUQ 52 chronic noncancer pain patients referred from a university-based multidisciplinary pain clinic for psychiatric evaluation for problematic narcotic use or drug-seeking behaviors Dyson et al. [23] CAGE; CAGE-AID; SMAST; DAST; ASI; CUAD; SCID-P 100 adult psychiatric inpatients recently admitted to a public psychiatric hospital Jonasson et al. [37] ADDIS/SUDDS 265 orthopedic and chronic noncancer pain patients treated at a Swedish rehabilitation clinic Lazowski et al. [20] SASSI 772 patients with DSM-III-R diagnosis of substance abuse problems Chabal et al. [34] 5-point prescription opiate abuse checklist 403 chronic noncancer pain patients referred to a pain clinic Latimer et al. [31] POSIT 342 healthy adolescents, aged years, from school, clinical, and correctional settings Coambs et al. [13] SISAP 9,915 Canadian adults aged 15 and older participating in the National Alcohol and Drug Survey (NADS) Brown and Rounds [21] CAGE; CAGE-AID; SMAST-AID 124 healthy adult patients in an academic, community-based family medicine clinic Beresford et al. [42] CAGE 915 adult patients in a general hospital, selected at random from daily admissions from January 1987 to April 1989 Gavin et al. [38] DAST 501 patients seeking treatment for alcohol and/or drug dependence at an alcohol and drug treatment center Test reliability of PDUQ as an adjunct to ASAM and DSM-IV for assessing of presence of addiction in the context of chronic pain Compare several alcohol and drug abuse scales to determine their efficiency and validity for psychiatric patients Document the prevalence of past and present analgesic use disorders, and compare the prevalence rates based on the DSM-III-R and DSM-IV criteria Reliability: Cronbach s alpha = 0.79 A larger study was published later: Schieffer et al. [41] Reliability: Interrater, kappa: CAGE, 1; CAGE-AID, 1; SMAST, 0.99; DAST, 0.96; CUAD, 1; ASI, ; SCID-P, 0.81 Validity: Intercorrelation arithmetic mean between 0.51 and 0.64 No attempt to look at reliability or validity of the tools; not the focus of this report Validate latest version of SASSI (SASSI-3) Sensitivity, 97%; specificity, 95% Validity: 97% accuracy Create opiate abuse criteria, test interrater reliability of the criteria, apply the criteria to a group of chronic pain patients, and correlate risk of opiate abuse with the results of alcohol and drug testing Examine ability of POSIT scale to screen for drug abuse disorders among adolescents. Validity: good correlation with MAST and DAST Reliability not formally tested Reliability: alpha = Validity: good correlation with independent measures of drug abuse diagnosis (via the Adolescent Diagnostic Interview, ADI) and drug use frequency (via the Personal Experience Inventory, PEI) Validation study Correct classification rate of 80.2%; sensitivity: 91% of actual substance abusers correctly identified; specificity: 78% of non-substance-abusers correctly classified Validity: good correlation with DAST and MAST Compare CAGE-AID and SMAST-AID as criterion measures for alcohol or drug problems Compare results with CAGE and several computer-assisted laboratory data profiles for screening of covert alcoholism Validate DAST with respect to diagnosis of drug abuse and drug dependence Sensitivity: GAGE-AID, 70%; SMAST-AID, 40% Specificity: GAGE-AID, 85%; SMAST-AID, 95% Formal reliability studies not done Positive predictive power: 87%d Sensitivity: 76% Specificity: 94% Formal reliability study not performed (41) ROC curve for sensitivity and false positive rates, AUC = Validity: good correlation with DSM-III Formal reliability studies not performed Abbreviations: ADDIS/SUDDS = alcohol and drug diagnostic instrument/substance use disorder diagnostic schedule; ASAM = American Society of Addiction Medicine; ASI = addiction severity index; AUC = area under the curve; CAGE-AID = cut down, annoyed, guilty, eye-opener/adapted to include drugs; CUAD = chemical use, abuse, and dependence scale; DAPA-PC = drug abuse problem assessment for primary care; DAST = drug abuse screening test; DIRE = diagnosis, intractability, risk, efficacy; KMSK = Kreek-McHugh- Schluger-Kellogg scale; ORT = opioid risk tool; PDUQ = prescription drug use questionnaire; PMQ = pain medication questionnaire; POSIT = problem-oriented screening instrument for teenagers; RAFFT = relax, alone, friends, family, trouble; ROC = receiver operating characteristic; SASSI = substance abuse subtle screening inventory; SCID-P = Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders, 3rd ed. (DSM-III-R), Patient Version; SISAP = screening instrument for substance abuse potential; SMAST-AID = short Michigan alcoholism screening test/adapted to include drugs; SOAPP = screener and opioid assessment for patients with pain; STAR = screening tool for addiction risk; TICS = two-item conjoint screen.

7 Screening Tools of Opioid Abuse S151 Table 4 Tool descriptions Tool Name Format/Use s Citations ADDIS Interviewer-administered, structured series of questions Swedish version of SUDDS ASI Semistructured clinical interview Severity of alcohol and drug abuse rated on a 10-point scale Atluri tool Physician-administered Measures six clinical criteria: 1 (yes) or 0 (no) Score ranges from 0 and 6; score of 4 is red flag for risk of opioid abuse CAGE Physician-administered, four-item tool Screens for covert alcoholism in adults CAGE-AID Four-item CAGE adapted to assess alcohol and drug abuse CCI Designed to capture personality traits and attitudes that could lead to problematic drug use, failure to progress toward functional goals, and overreliance on medication for coping with chronic pain Chabal five-point prescription opiate abuse checklist Physician-administered checklist of five abuse criteria Relies on easily observable behaviors instead of patient questioning COMM 17-item scale Designed to assess current aberrant behavior Elicits information for use in diagnosing SUDs in accordance with DSM-III-R criteria Structured interview is time-consuming Uses both objective and subjective information sources to rate severity of problem Gold standard for classifying abuse and dependence and quantifying severity May be too cumbersome for typical clinic setting More suited for research clinics Short, easy-to-use tool Total scores correlate with likelihood of inappropriate opioid use Evaluated in retrospective case-control study only; needs further validation in prospective clinical studies Widely used and well validated Sensitive and specific for recognition of alcohol dependence Quick and easy to administer Screens for a wider range of substance abuse disorders than those screened by CAGE Familiar format may make it useful as a quick screen that can be followed up by a more detailed assessment Not validated in pain patients Currently in development Validation trial underway Could potentially be useful in treatment planning Further study needed Based on observed behaviors; not influenced by patient responses to questions Patients with three abuse criteria are considered opiate abusers Valuable tool for gauging adherence once patients are on opioid therapy Currently in development Useful in assessing adherence Acceptable psychometrics May prove useful in the reassessment of opioid therapy Further study needed CUAD Brief, semistructured interview Provides both DSM-III-R diagnoses and severity scores for alcohol and drug use Takes longer to administer than CAGE-AID and DAST, but performs better and provides a DSM-III-R diagnosis Validated only in severely mentally ill inpatient population Broad applicability unclear DAPA-PC Self-administered, Internet-based tool Used to screen alcohol/drug abuse problems in adult populations Tool completed in privacy; may be more likely to elicit honest responses; may be particularly useful for screening older adults who often find the alcohol/drug problems shameful Requires dedicated computer with Internet access Features automatic scoring, generation of a patient profile for medical reference, and presentation of unique motivational messages and advice to patient Tool has two levels: a very brief risk and trauma assessment (Health and Safety Screen) and a more detailed assessment (Drug and Alcohol Problem Screen) Easy to use; does not require a trained professional for administration [37,43,44] [23,27] [12] [42,45 48] [21] [15] [34] [36] [23] [16,49]

8 S152 Passik et al. Table 4 Continued Tool Name Format/Use s Citations DAST Self-administered, 28-item tool Several abbreviated versions available including versions specific for adolescents DIRE Physician-administered, seven-item tool, with total score ranging from 7 to 21 DIRE scores 13 indicate that patient is not a suitable candidate for long-term opioid analgesia KMSK Physician-administered tool with sections on alcohol, cocaine, heroin/opioids, and tobacco Each section assesses frequency, amount, and duration of use during period of greatest consumption Used to screen for lifetime diagnosis of alcohol, opiate, or cocaine dependence ORT Self-administered, five-item quantitative screening tool Screens for risk of aberrant behaviors associated with substance abuse in chronic pain patients PDUQ Physician-administered, 42-item tool with yes-or-no answers Designed for use in chronic pain patients with suspected addiction to pain medication Scores > 15 associated with substance abuse disorder PMQ Self-administered, 26-item tool Used to evaluate risk of opioid misuse in chronic pain patients POSIT Self-administered, 139-item tool Abbreviated versions with 11 and 17 items are available with yes-or-no answers RAFFT Self-administered, five-item tool concerning drug or alcohol use Excellent psychometrics Cutoff of 6 indicates drug abuse or dependence problems Low cutoffs should be used when screening known drug abusers, and higher cutoffs for screening non-substance-abusers No major validation trials in pain patients Does not predict aberrant behavior during pain treatment Easy to use; takes <2 minutes on average Designed for primary care physician Score correlates with patient compliance and efficacy of opioid analgesia Tool validated by six experts evaluating patient case vignettes Prospective validation needed Easy to use; takes only 5 minutes Results correlate with well-validated SCID-I in general population (but not in opiate-dependent subpopulation) when opiate use is measured Measures use patterns during period of greatest consumption, but also asks whether subject is using drug currently and whether drug is a drug of choice Needs to be tested in pain patients who are being prescribed opioids Brief, easy-to-score tool validated in pain population Provides excellent discrimination for patients with low-risk scores (<3) vs high-risk scores ( 8) One question limited by patients knowledge of family history of substance abuse Susceptible to deception Tool takes about 20 minutes to administer Designed for use with supporting information sources, including referring physician, family members, and past medical history Subset of three items may also predict presence of substance abuse disorder Useful and accurate in retrospectively evaluating nonadherence and the degree to which it corresponds to DSM-IV diagnoses of SUD Useful in gauging progress of pain patients already taking opioids Needs further validation in other patient populations May prove useful in evaluating patients who are in or plan to be in a multidisciplinary pain clinic Uses low cutoff values to identify adolescents who require further assessments in 10 functional areas Long Not validated in pain population Fast and easy to use Cutoff of 3 positive answers has sensitivity of 84% and specificity of 67% for SUD diagnosis Designed for adolescents Not validated in pain population Unclear utility in patient management [23] [10] [18] [11] [35,41] [32,33] [31,50] [26]

9 Screening Tools of Opioid Abuse S153 SASSI-3 93 items, mostly true-or-false questions (10 questions are for research purposes only and not included in the actual scoring) Useful as screening tool for human service practitioners in a variety of settings Excellent psychometrics Includes both face-valid items and subtle items that bear no apparent relation to substance misuse Questions do not address substance misuse directly or obviously; decreased susceptibility to deception Identifies individuals with high probability of a diagnosable SUD Not prospectively validated in pain patients SCID-P Structured interview SCID is the gold standard for validating other tools SCID-P is a patient-directed version that is perhaps less stigmatizing than face-to-face interviews SISAP Physician-administered, five-item tool Screens for potential opioid abusers in patients with chronic noncancer pain Developed from a large database of pain patients (approximately 5,000) Brief Designed to minimize misrepresentation or falsification of substance abuse Prospective validation needed SMAST Physician-administered 13-item tool Three or more positive responses suggests alcohol abuse [21] [21] SMAST-AID 13-item SMAST adapted to screen for both alcohol and drug abuse SOAPP-24 Self-administered, 24-item tool Screens for opioid misuse in chronic pain patients SOAPP-14 Self-administered; 14 items most predictive of aberrant drug behavior taken from original 24-item SOAPP-24 Each item rated from 0 to 4 Cutoff score of 8 indicative of high risk of opioid abuse STAR Self-administered, 14-item tool Assesses substance addiction risk in chronic pain patients Low sensitivity for screening alcohol and drug abuse in a primary care setting Other, better screening tools exist Limited value in pain management setting Best psychometrics of any measure designed to predict aberrant behavior before therapy is begun Briefer and perhaps less susceptible to deception than more face-valid tools Briefer tools like ORT may be preferable in low-risk populations, but the longer SOAPP may be preferable in high-risk populations Easy to use; provides numeric score for patient evaluation purposes Low cutoff allows recognition of addiction even if patient purposefully underreports aberrant behavior SOAPP-14 has not been validated over a set period or in specific subpopulations for which different cutoff scores may be necessary Brief and simple; no correction for lying Prospective validation needed SUDDS Structured interview tool Suitable for assessing dependence or abuse in chronic pain patients who are receiving opioid analgesics TICS Physician-administered, two-item screen Designed to be memorized and incorporated into clinical interview Structured interview may be time-consuming Quick and easy to use; suitable for routine clinical practice Needs to be validated in pain populations [20] [13] [9] [8] [14] [37] [17] Abbreviations: ADDIS = alcohol and drug diagnostic instrument; ASI, addiction severity index; CAGE-AID = cut down, annoyed, guilty, eye-opener/adapted to include drugs; CCI = chemical coping inventory; COMM = current opioid misuse measure; CUAD = chemical use, abuse, and dependence scale; DAPA-PC = drug abuse problem assessment for primary care; DAST = drug abuse screening test; DIRE = diagnosis, intractability, risk, efficacy; DSM-III-R = Diagnostic and Statistical Manual of Mental Disorders, 3rd ed.; KMSK = Kreek-McHugh-Schluger-Kellogg scale; ORT = opioid risk tool; PDUQ = prescription drug use questionnaire; PMQ = pain medication questionnaire; POSIT = problem-oriented screening instrument for teenagers; RAFFT = relax, alone, friends, family, trouble; SASSI-3 = substance abuse subtle screening inventory; SCID-P = Structured Clinical Interview for DSM-III-R, Patient Version; SISAP = screening instrument for substance abuse potential; SOAPP = screener and opioid assessment for patients with pain; SMAST-AID = short Michigan alcoholism screening test/adapted to include drugs; SUD = substance use disorder; SUDDS = substance use disorder diagnostic schedule; STAR = screening tool for addiction risk; TICS = two-item conjoint screen.

10 S154 problematic behavior in people prescribed with opioids for pain than expert clinician predictions (T. Jones, personal communication). In high-risk pain populations, these advantages may make its use desirable despite its greater length and scoring requirements as compared with other similar measures, most notably the Opiate Risk Tool (ORT), discussed further. The SOAPP probably has the best psychometrics of any of the measures designed to predict aberrant drug-taking behavior prior to the initiation of opioid therapy. Passik et al. Diagnosis, Intractability, Risk, and Efficacy Score (DIRE) The DIRE is a clinician-rated scale designed to predict the analgesic efficacy of, and patient compliance to, long-term opioid treatment in the primary care setting. The scale is intended for use in patients who have chronic noncancer pain and who are currently being treated with opioids or are being considered for opioid treatment. The DIRE includes four categories: diagnosis, intractability, risk, and efficacy. The risk category is further divided into four subcategories: psychological, chemical health, reliability, and social support. Each factor is rated from 1 to 3, with higher scores indicating a more persuasive case for opioid therapy in terms of treatment efficacy and compliance. Patients with scores of 14 and above are considered good candidates for longterm opioid treatment, whereas those with lower scores are not considered good candidates, thus identifying the subset of patients who can be deemed at risk. Belgrade and colleagues [10] performed a retrospective analysis of the DIRE score in 61 patients who had been treated with opioids for chronic noncancer pain at an outpatient pain management center. Most patients had chronic musculoskeletal back and neck pain (41%), abdominal pain (15%), or neuropathic pain (13%), and were treated with opioids for a median duration of 37.5 months. In this cohort, the DIRE score exhibited a high internal consistency, with a Cronbach s alpha coefficient of All factors besides diagnosis were significantly related to treatment compliance (P < 0.001), and all except intractability were significantly associated with efficacy (P < 0.05). This was to be expected because, by definition, efficacy is hard to achieve in an intractable condition. Although the diagnosis subscore was not correlated with outcome, it is included in order to avoid treating with opioids patients who do not have a diagnosis or condition that is associated with moderate or severe pain. At a cutoff point of 13, the sensitivity and specificity of the DIRE score for predicting compliance in the study cohort were 94% and 87%, respectively, and for predicting efficacy, 81% and 76%, respectively. Interclass correlation for interrater reliability and intrarater reliability was 0.94 and 0.95, respectively. The DIRE score performed well in identifying suitable candidates with chronic noncancer pain for long-term opioid therapy, but the retrospective nature of the study raises several limitations, most notably that investigators scored patients according to case history. Moreover, the study population was relatively small and included a variety of chronic pain etiologies. Prospective analyses in more homogeneous chronic pain populations are still needed for confirming the utility of the DIRE score. However, for pain clinicians who prefer an observer-based, clinician-rated assessment strategy, the DIRE has tremendous potential. Using the DIRE is actually a process of systematizing the clinical judgments that pain clinicians typically make, and quantifying them. This process is comfortable for, and familiar to, most pain clinicians and avoids the use of paper-and-pencil measures, where these may be less a part of particular clinics routines. ORT The ORT is a five-item yes-or-no self-report that is designed to predict the probability of a patient s displaying aberrant behavior when prescribed opioids for chronic pain. It consists of items on family history of substance abuse, personal history of substance abuse, age, history of preadolescent sexual abuse, and psychological disease. The items on substance abuse contain three subsections covering alcohol, illegal drugs, and prescription drugs, and the item on psychological disease has two subsections that distinguish depression from other disorders. Each positive response is given a score based on patient gender, and then the scores are summed to derive the probability of opioidrelated aberrant behavior. Scores of 0 3 are associated with low risk, 4 7 with moderate risk, and 8 and over with high risk. Webster and Webster [11] evaluated the ORT in 185 consecutive new patients at a pain clinic. Seventeen of 18 patients (94.4%) in the low-risk category did not display aberrant behavior. In contrast, 40 of 44 patients (90.9%) in the high-risk category, and 35 of 123

11 Screening Tools of Opioid Abuse patients (28.5%) in the moderate-risk category did display aberrant behaviors. The most common aberrant behaviors were solicitation of opioids from other providers, unauthorized escalation of opioid dose, abnormal urine or blood screening, and use of more opioids than those prescribed. The ORT displayed an excellent discriminatory ability in both men and women, with observed c statistic values of 0.82 and 0.85, respectively. Because of its brevity and ease of scoring, the ORT has tremendous clinician appeal and is clearly the easiest way to perform a risk assessment with a tool validated in pain patients and specifically designed to predict problematic behavior in people prescribed opioids for pain. Its lone drawback is its susceptibility to deception. Clinicians will have to decide if guarding against deception is important enough to use a longer and more cumbersome tool or if the documentation of risk assessment (not to mention a clear evidence of deception, should it occur) satisfies their requirements. Atluri Screening Tool Atluri and Sudarshan [12] developed a clinicianrated screening tool to detect the risk of inappropriate prescription opioid use in patients with chronic pain. Using a case-control design, the investigators retrospectively identified 107 patients who were dismissed from the pain clinic for inappropriate prescription opioid use (i.e., selfdose escalation, questionable urine screens, etc.), and compared them with 103 randomly chosen chronic pain patients who did not have evidence of inappropriate prescription opioid use. On multivariate analysis, six clinical criteria were significantly associated with opioid abuse; these included focus on opioids, opioid overuse, other substance abuse, low functional status, unclear pain etiology, and exaggeration of pain (i.e., pain scores always at most severe at all times). The investigators identified a checklist of questions for each of these six criteria. The screening tool is based on the number of positive criteria, ranging from 0 to 6. Most patients (77%) in the inappropriate use group scored above the cutoff of 3, whereas most (84%) in the control group scored below this cutoff level. Notably, patients with scores greater than 3 had an odds ratio of 16.6 (95% confidence interval [CI]: ; P 0.001) for opioid abuse, compared with the odds ratio for those with scores below this cutoff. S155 These preliminary results are promising, but it is important to recognize that the study was retrospective in design, included only patients being treated with opioids for chronic pain, and excluded those with cancer pain or acute pain. Screening Instrument for Substance Abuse Potential (SISAP) The SISAP is a physician-administered screening tool designed to identify chronic noncancer pain patients who may be at risk of abusing opioids if prescribed. The instrument is easy to use and takes only a few minutes to administer. The SISAP was developed and validated using data from the National Alcohol and Drug Survey (NADS), conducted in Canada in 1989 [13]. The five questions elicit information about the number of drinks in a typical day and typical week, use of marijuana in the past year, history of cigarette smoking, and age. In the development cohort of 4,948 NADS respondents, SISAP correctly identified 91% of substance abusers and 77% of those who did not have alcohol or drug abuse problems. SISAP was validated in the other half of the subject pool from NADS and showed a comparable performance by correctly classifying 91% of the actual substance abusers and 78% of the nonabusers. Overall, SISAP exhibited an accuracy of 80%, with a sensitivity of 91% and a specificity of 78%. Thus, SISAP can stratify chronic pain patients seen in a primary care setting, thereby allowing increased opioid availability to those who are not at risk of opioid abuse and providing improved monitoring or referral to those who are at risk. The SISAP was developed on perhaps the largest database of pain patients of any of the screening tools included in this review. It is unclear why, in the several years since its development, the tool has not received further validation in prospective trials. Perhaps the requirement that clinicians ask a set of pointed questions about alcohol and drug use has delayed the tool s adoption by pain clinicians. Its brief format, though, would lend itself to use in pain clinics, and a prospective trial of the tool s ability to predict aberrant drug-taking behaviors is needed. Screening Tool for Addiction Risk (STAR) The STAR, a screening tool for addiction risk, consists of 14 yes-or-no questions relating to cigarette, alcohol, and drug use; family or household members with drug or alcohol abuse; visits to pain

12 S156 clinics and emergency rooms; and feelings of depression, anxiety, and altered mood. Friedman and coworkers [14] evaluated the STAR in a sample of 48 chronic pain patients including 14 with a history of substance abuse. Individual screening questions related to tobacco abuse, prior treatment in a drug or alcohol rehabilitation facility, or treatment at another pain clinic were more likely to be positive in patients with current substance abuse (P < 0.05). On logistic regression, a history of treatment in a drug or alcohol rehabilitation facility was a significant predictor of addiction (positive predictive value, 93%; negative predictive value, 5.9%). The STAR is brief, has been used in chronic pain patients, and has potential as an aid to screening and treatment planning. Larger prospective studies that examine the tool s ability to predict aberrant drug-taking behaviors are needed. Chemical Coping Inventory (CCI) The CCI is a tool in development meant to capture personality traits and attitudes that could lead to problematic drug use, failure to progress toward functional goals, and an overreliance on medication as a sole way of coping with chronic pain [15]. Inventory items are designed to assess somatization, sensation seeking, alexithymia, and overcentrality of drug taking. Kirsh and colleagues [15] contend that there is a vast middle ground of chronic pain patients who have some of the aforementioned personality traits and who are at risk for problematic drug use unrelated to substance use disorder (SUD). In essence, the tool is designed to identify patients who engage in problematic behavior but do not have the same compulsive use despite harm and other facets indicative of true addiction problems. Initial instrument development work has been promising, and a large validation trial is under way. The CCI will add a great deal to pain treatment planning (i.e., to bring in psychosocial treatments early and to utilize drug regimens that are unlikely to become problematic for psychological reasons) should its psychometrics prove to be acceptable. Screening Tools in Nonpain Populations Drug Abuse Problem Assessment for Primary Care (DAPA-PC) The DAPA-PC is a self-administered, Internetbased screening tool that was developed under a Passik et al. contract from the National Institute on Drug Abuse [16]. Patients initially answer a very brief risk and trauma assessment called the Health and Safety Screen, which does not overtly address substance abuse but rather explores related issues such as depression and physical/emotional abuse over the previous 5 years. The patient s score on this initial screen determines whether he or she will be moved to a second screen, the Drug and Alcohol Problem Screen, which focuses directly on drug and alcohol problems through a series of 12 questions. Once this second screen is completed, the system posts information indicating whether the patient s health is at risk from alcohol or drug use. If such risk exists, the system provides motivational messages and advice to the patient, including health links for the patient to explore. The clinician can then access a summary of the patent s results as well as useful links of interest to health care professionals. The DAPA-PC can be easily implemented in a primary care setting with the availability of a dedicated computer. It offers the advantage that a trained professional is not needed to administer the questions or score the results. Moreover, because it is completed in private, this tool may be more likely than others to elicit honest answers to questions about alcohol and drug use. The DAPA-PC is an intriguing tool that has potential utility for some pain clinics, especially those with electronic medical records in which such an assessment could be incorporated, even before a first clinic visit. This system requires validation in chronic pain samples. Two-Item Conjoint Screen (TICS) The TICS is a two-item self-report questionnaire designed to detect current substance abuse [17]. The questions are scored on a four-point scale (0 = never, 4 = often), but a response of rarely, sometimes, or often is interpreted as a positive reply. This approach is taken to allow patients to minimize their responses while still answering in a positive manner. Each question in TICS is conjoined, in that it asks about alcohol and drug abuse simultaneously. There are three reasons for this: first, people who have problems with a variety of substances may be more likely to give a positive response. Second, people may be more likely to be honest and give positive responses when a question encompasses many drugs, because the question diminishes the fear of possible legal ramifications, the stigma of a specific drug, or

13 Screening Tools of Opioid Abuse other side effects. Finally, clinicians can screen for both alcohol and drug problems in the same amount of time it takes to screen for alcohol problems alone. Nonetheless, there are still problems with this method. Notably, patients who only use alcohol may be hesitant to give positive responses for fear that they may be considered drug users as well. Brown and coworkers tested the TICS in two phases in adults aged years [17]. Five questions were administered to a screening cohort of 434 participants in the first phase, and then based on the results, only two questions were retained in the second phase and administered to a validity cohort of 702 subjects. Including the other three questions did not yield any substantial improvements in identifying substance disorders. Overall, the two-item TICS had a sensitivity of 79.3% and a specificity of 77.9% when there was a positive response to at least one item. The TICS is more sensitive to dependence than abuse, and particularly well suited for those who have disorders involving marijuana or cocaine. Like all tools, TICS has its downsides; although the negative predictive value was 92.7%, the positive predictive value was only 51.8%. Among respondents with only one positive response, the prevalence of substance abuse was 36.5%, and for those with positive responses to both questions, the prevalence was 72.4%. Due to the high false-positive rates, it would be best to administer a more specific test to any patient who has a positive response to either question. The TICS is a fairly sensitive and specific and certainly very rapid tool. However, to be of greater relevance to pain management, this tool would have to demonstrate consistency and predictive validity in pain patients, who may interpret the questions in an idiosyncratic fashion as compared with the normative sample. Kreek-McHugh-Schluger-Kellogg Scale (KMSK Scale) The KMSK scale is an eight-item tool that measures self-exposure to opiates, cocaine, alcohol, and tobacco during the individual s period of greatest consumption. Separate scales are available for each of these substances, which assess the frequency, the amount, the duration, the mode of use, and whether the substance is the individual s substance of choice. The frequency, duration, and S157 amount are transformed into numerical values, and then the total score is determined from the sum of these values. Kellogg et al. [18] assessed the KMSK scale in 100 subjects (mean age 36.2 years), including healthy volunteers and those with defined addictive diseases, and found different psychometrics for each type of drug. The opioid scale ranges from a score of A score of 2 or above was sufficient to lead to a dependence diagnosis, but a cutoff score of 9 was optimal, based on a chisquare analysis, having a sensitivity of 100%, a specificity of 99%, a positive predictive value of 95%, and a negative predictive value of 100%. Although the KMSK was able to identify opioid dependence, it was unable to assess the severity of the problem. Because 71% of those who were considered opioid dependent had six or seven of the seven Structured Clinical Interview for DSM- IV (Diagnostic and Statistical Manual of Mental Disorders, 4th ed.) Axis I Disorders (SCID-I) dependence criteria, the ability to discriminate dependence severity may have been restricted. The KMSK cocaine scale ranges from 0 to 16. Using a cutoff of 11, the KMSK cocaine subscale had a sensitivity of 97%, a specificity of 94%, a positive predictive value of 88%, and a negative predictive value of 99% in detecting cocaine dependence [18]. A high correlation between the cocaine scale and the number of DSM-IV criteria was found in the subgroup of 31 subjects with cocaine dependence. As with the opioid scale, the cocaine scale correctly assessed dependence but not severity. The range in severity was restricted in the same way as the opioid-dependent subgroup, as 71% met six or seven SCID-I dependence criteria. Finally, the KMSK alcohol scale ranges from 0 to 13. Using a cutoff of 11, the alcohol scale had a sensitivity of 90%, a specificity of 90%, a positive predictive value of 69%, and a negative predictive value of 97% [18]. A high correlation was found between the KMSK alcohol scale and the SCID-I alcohol scales, although the correlation was lower than those between the opioid and cocaine scales and their respective SCID-I dependence scores. The alcohol scale successfully predicted severity in individuals who were exclusively alcohol dependent, but not in those who had also been dependent on opioids and cocaine. The KMSK has not been validated in people with pain. The measure failed in its attempt to quantify

14 S158 the severity of drug dependence in drugs other than alcohol, and therefore adds little to other existing screening measures from the vantage point of pain clinicians. Drug Abuse Screening Test (DAST) The DAST a 28-item yes-or-no self-report questionnaire is designed as a clinical screening tool for substance abuse. A cutoff score of 6 is usually used to indicate drug abuse or dependence problems. Several abbreviated versions of DAST are also used, including DAST-20, DAST-10, and DAST-A, the latter of which is designed for adolescents [19]. The DAST-28 has high internal consistency, with Cronbach s alpha coefficients ranging from 0.92 to 0.94, and high test retest reliability with a correlation coefficient of 0.85, although the test retest was only separated by a few weeks. The DAST-28 is a highly face-valid instrument; it measures what it is designed to measure, namely, aberrant drug use. Because the tool is susceptible to deception, aberrant substance users who intentionally give false responses may not be identified. The sensitivity of the DAST-28 ranges from 81% to 96%, and the specificity ranges from 71% to 94%. Increasing the cutoff from 6 reduces the sensitivity of DAST-28, but increases its specificity. Therefore, a cutoff should be selected that best fits the objectives of the screening purposes. The DAST-10 includes 10 items from the original questionnaire, three of which have been rewritten with minor modifications [19]. It has high internal consistency, with coefficients ranging from 0.86 to In a psychiatric population of 45 patients, the DAST-10 had an acceptable test retest reliability over a 2-week period, with a coefficient of Using a cutoff score from 1/2 to 3/4, the sensitivity ranged from 95% to 41%, and the specificity from 68% to 99%. The lowest sensitivity and the highest specificity came from the same sample of psychiatric patients and was based on discharge diagnosis. If this sample is not considered, the overall predictive accuracy for diagnosis was 70% or higher. The DAST-20 uses the DAST-10 in combination with 10 other items from the original DAST-28 with minor modifications to two additional questions. The internal consistency of DAST-20 ranged from 0.74 to In the aforementioned psychiatric patient population, the test retest reliability over a 2-week period was As cutoff scores rose from 3/4 to 5/6, the sensitivity went from 89% to 74%, and the specificity increased from 68% to 83%. Passik et al. DAST-20 is highly correlated with DAST-28 (r = 0.99). In addition, DAST-10 and DAST-20 are highly correlated with each other (r = 0.97), and are also highly correlated with other drug, alcohol, and psychiatric indicators. No criterion validity for the DAST-A has been measured. Although the sensitivity of the DAST tests is high, there is still a significant variation across study groups. In order to optimally use the DAST tools, it is important to choose cutoffs that match the clinical need. It is better to use a low cutoff when screening patients with a remote history of drug abuse, but a higher cutoff when screening patients who are not drug abusers. The various forms of the DAST have excellent psychometrics and, especially with the briefer versions, are suitable for the pain treatment setting as a measure of substance abuse. The pain clinician could incorporate this scale into pretreatment assessments. The measure has yet to undergo a major validation trial in pain patients. It predicts substance abuse but not specifically aberrant behavior during pain treatment. Substance Abuse Subtle Screening Inventory (SASSI) The SASSI is an objective screening tool designed to identify patients with a high probability of having a diagnosable SUD, which can be used in a variety of clinical settings. Because some substance abusers may not be able or willing to acknowledge relevant symptoms, the SASSI was designed to include both face-valid items, which ask about lifetime frequency of specific behaviors related to substance use, as well as subtle true-or-false items that have no apparent relationship with substance abuse. The SASSI-3 is the third version of this instrument. Lazowski et al. [20] evaluated the SASSI-3 in a cohort of 1,958 patients from a variety of clinical settings including addiction treatment centers, general psychiatric hospitals, a vocational rehabilitation program, and a sexoffender treatment program. The internal consistency of the SASSI-3 was high, with an alpha coefficient of The test retest reliability over a 2-week period in a subgroup of 40 respondents had stability coefficients of When compared with clinical diagnoses of substance dependence according to the Diagnostic and Statistical Manual of Mental Disorders, 3rd ed., revised (DSM- III-R), the SASSI-3 showed an overall accuracy of 97% in a cross-validation cohort of 381 subjects,

15 Screening Tools of Opioid Abuse with a sensitivity of 97%, a specificity of 95%, a positive predictive power of 99%, and a negative predictive power of 90%. The accuracy of SASSI-3 ranged from 93% to 98% across the five clinical settings evaluated. Logistic analyses showed that classifications based on the SASSI-3 are not significantly affected by demographic variables or by the level of patient adjustment or functioning. The SASSI-3 may be particularly relevant for the early identification of people who may have substance dependence or who may not be able to acknowledge symptoms relevant to such dependence. A prospective trial of the tool s ability to correctly identify pain patients at risk of both substance abuse and aberrant drug-taking behavior would be welcome, given its excellent psychometrics and decreased susceptibility to deception. Cut Down, Annoyed, Guilty, Eye-Opener Tool, Adjusted to Include Drugs (CAGE-AID) The CAGE-AID, an adapted version of the famous four-item CAGE alcoholism screening test, is designed to assess alcohol abuse as well as substance abuse. Essentially, the CAGE-AID questionnaire was adapted by the addition of the phrase or drug use to each of the four items. Brown and Rounds [21] assessed the CAGE-AID in a group of 124 subjects from an academic, community-based family practice. The study participants had a mean age of 38.6 years and came from a mixed community (78% were employed, 48% had an associate degree or higher, and 23% had an income at or below the poverty level). Approximately half of the participants met the Diagnostic and Statistical Manual of Mental Disorders, 2nd ed. (DSM-II), criteria for dependence, with a variety of substance-related diagnoses. Using the standard criterion score of 2 or more positive answers, the sensitivity and the specificity of CAGE-AID was 70% and 85%, respectively. When only one positive answer was required, the sensitivity and the specificity was 79% and 77%, respectively. In this cohort, the CAGE-AID had higher sensitivity in subjects with lower education and income levels. The CAGE-AID was also tested in a group of 50 type 2 diabetes Northern Plains Native Americans at an Indian Health Service hospital [22]. The individual items coefficients were positively correlated with the total score, and the internal consistency reliability coefficient was S159 The CAGE-AID has not been validated in pain patients. Nevertheless, its brevity and familiarity to many clinicians make it a reasonable initial screen, after which a more detailed assessment using an aberrant-behavior predicting instrument might be useful. This would be a particularly reasonable approach in many clinics wherein a precise substance-abuse diagnosis is not the goal. Chemical Use, Abuse, and Dependence (CUAD) Scale The CUAD is a brief, semistructured interview designed to assess problems with all types of drugs. Although it is an interview, the design of the CUAD is optimized for use in a clinical setting, inasmuch as it does not require a trained interviewer and can be completed in less time than that needed for the Addiction Severity Index (ASI). The CUAD includes only two items for those who deny using both alcohol and drugs, but can reach 80 items for those who admit to concurrently using four illicit drugs. Although the CUAD is designed to assess current substance use, it can be used to gather information about typical use. Once a subject has admitted to using either drugs or alcohol, questions are asked about the frequency, amount, mode, and duration of use for each substance through a set of seven categories (17 yesor-no questions). On the basis of the responses, a total severity score, as well as a score for each substance used, is calculated. The presence or absence of a DSM-III-R substance use diagnosis is also determined for each substance. The CUAD was assessed in a severely mentally ill population, consisting of 100 consecutively admitted adults to a public psychiatric facility [23]. Most were male (72%); African American (75%); either single, divorced, or separated (95%); and unemployed (90%). Most lived with their families, although 15% were homeless. Approximately twothirds had a psychiatric diagnosis, and more than half had been admitted to a hospital in the previous year. The mean Global Assessment of Functioning scale score in this cohort was In this cohort, the CUAD exhibited a high interrater reliability (correlation coefficient of 0.98) and a high internal consistency for the assessment of alcohol abuse (Cronbach s alpha coefficient of 0.96) as well as the two most frequently abused drugs, cocaine (alpha of 0.97) and marijuana (alpha of 0.95) The test retest reliability over a mean period of 3.4 days for the absence or presence of either an alcohol or a drug use disorder was perfect, with a

16 S160 kappa coefficient of 1.0. The overall test retest reliability for the total severity score was very high, with a Pearson correlation coefficient of The CUAD was shown to be a valid measure, with a high correlation to the DAST, Short Michigan Alcohol Screening Test (SMAST), and ASI alcohol and drug scales. The high correlation to SMAST, a self-report assessment of alcohol abuse, probably reflects that 75% of the chemicals being described as substance 1 was alcohol. McGovern and Morrison [24] showed that the predictive validity of the CUAD was tested by its ability to distinguish patients at three levels of substance abuse treatment: inpatient, partial hospitalization, or outpatient. This was assessed based on the assumption that those with the highest scores were the most likely to be in inpatient facilities. In this analysis, the CUAD was able to correctly identify the treatment setting (P < 0.001). Overall, the CUAD is a very good tool for assessing SUDs in a hospital-based substance abuse treatment program and in the severely mentally ill population. Additional research is needed to ascertain the validity of the CUAD in other populations, as well as the ability of the CUAD to measure change during treatment. For pain clinicians preferring a structured interview in history taking, this measure may be preferable to others as it can be somewhat shorter than the ASI. A detailed substance abuse history might be particularly needed in high-risk patient groups. However, the CUAD is yet to be validated in a pain patient population, especially those entering treatment wherein the incentive to deceive may be particularly high. Short Michigan Alcoholism Screening Test, Adjusted to Include Drugs (SMAST-AID) The SMAST-AID is an adapted version of the self-administered, 13-item SMAST, which is designed to assess both alcohol and substance abuse. The SMAST-AID was evaluated in the same academic, community-based family practice sample as that of the CAGE-AID [21]. Using the standard criterion score of 3 or more positive answers, the sensitivity of SMAST-AID was 40% and the specificity was 95%. When the criterion was reduced to two positive answers, the sensitivity was 51% and the specificity was 92%. Based on these results, Brown and Rounds stopped examining the SMAST-AID as a tool for screening alcohol and substance abuse. Passik et al. Superior screening tools for assessing alcohol and drug abuse exist and are likely to make this instrument of limited value to pain management. Relax, Alone, Friends, Family, Trouble (RAFFT) (Adult Studies) The RAFFT a five-item yes-or-no questionnaire for SUD has been tested on adolescents who were referred to an emergency room or ambulatory evaluation clinic [25], and on adults who presented to a psychiatric emergency and evaluation center [26]. The RAFFT takes about 1 minute, and tests for alcohol and drug abuse, but not the potential of abuse of pain medications specifically. The fifth question ( Have you ever gotten into trouble by drinking/drugging? ) is related to the criteria for substance abuse as defined by the DSM-IV. (The DSM-IV criterion for substance abuse captures maladaptive patterns of substance use within the last 12 months, leading to clinically significant impairment or distress, such as failure to fulfill major obligations; recurrent substance use in hazardous situations, such as driving a car; substance-related legal problems; or persistent or recurrent social or interpersonal problems.) Bastiaens et al. [26] tested this questionnaire in a population of 215 adults, and found that the positive and negative predictive values for two positive answers were 88.8% and 78.6%, respectively. The best results with RAAFT in this adult cohort were achieved with three positive answers, which yielded an overall classification accuracy of 81%. Adults who provided two or fewer positive answers had no SUDs 79% of the time. It was also more common for adults to score false positives than it was for adolescents to do so, as shown by the lower levels of specificity in adults than adolescents for two positive answers (51% vs 69%) and three positive answers (67% vs 90%). Like the adolescent population, the adults in this study were likely to have SUD and therefore may not be representative of the overall population. The RAFFT is of unclear utility for pain management. It may prove useful in the management of younger populations with pain, particularly those treated in emergency situations, such as sickle-cell anemia patients. The questionnaire s brevity is a positive feature in such a paradigm, but the tool will require validation in adolescents with pain.

17 Screening Tools of Opioid Abuse ASI The ASI is a semistructured interview administered by a trained clinician or interviewer. The tool is designed to assess the severity of drug and alcohol abuse as well as employment and psychosocial problems. The interview generally takes about minutes, and uses both objective and subjective information sources, as well as current and lifetime experiences, to provide severity ratings on a 10-point scale in each of the six areas commonly affected by addition [23]. The goal of this assessment is to identify treatment needs in each area. Symptom distress and desire for additional treatment are also obtained. The ASI showed high reliability and validity in assessing these problems in a cohort of 524 male veterans with alcohol and drug addictions [27]. In a cohort of 100 consecutive patients admitted to a public psychiatric facility, the interclass correlation coefficient for the alcohol and drug problem area severity scores on ASI was 0.78 and 0.83, respectively [23]. In this population, the ASI showed high degrees of association with the DAST and CUAD drug scores. The ASI is viewed by many in addictions research as the gold standard for the classification of drug abuse and dependence, and the assessment of its severity. Its interview format may prove cumbersome for some clinics, but this instrument may prove useful in research clinics or whenever finer distinctions among patients with drug abuse problems are needed. Screening Tools for Adolescent Populations While the focus of this supplement is the adult pain patient, it should be noted that several tools exist specifically for adolescent populations. In brief, Table 4 includes the RAAFT scale [26], the Drug Abuse Screening Test for Adolescents (DAST-A) [28], and the Problem-Oriented Screening Instrumentation for Teenagers (POSIT) [29 31]. s for the first two tools are similar to that seen previously for the adult versions. The POSIT is intriguing, but does require further validation and study for adolescent pain patients. Instruments that Assess for Current Misuse or Aberrant Behavior Pain Medication Questionnaire (PMQ) The PMQ is a 26-item self-report questionnaire designed to assess the risk of opioid misuse in S161 chronic pain patients [32,33]. Each question is answered using a five-point Likert format ranging from disagree to agree (with somewhat disagree neutral, and somewhat agree in between), and subsequently, numerical values ranging from 0 to 4 are assigned to these responses. Adams et al. [33] found that higher PMQ scores were associated with a history of substance abuse, higher levels of psychosocial distress, and poorer functioning in a cohort of 184 patients evaluated at an interdisciplinary pain treatment center. When the test was administered to 19 patients at two time points, approximately 30 minutes apart, the test retest reliability is significant (Pearson r coefficient = 0.85). For the entire cohort, the internal consistency of the PMQ was acceptable (Cronbach s alpha coefficient of 0.73). Holmes et al. [32] successfully replicated these findings in a cohort of 271 newly evaluated chronic pain patients. Compared with patients in the lowest tertile, those in the highest tertile were 2.6 times more likely to have a history of substance abuse, 3.2 times more likely to request early refills of prescription medication, and 2.3 times more likely to drop out of treatment. Moreover, patients receiving disability payments and those who were separated or divorced were more likely to score high on the PMQ. Holmes and colleagues [32] also found that PMQ scores decreased significantly with the completion of an interdisciplinary pain management program. Although Adams et al. [33] showed a positive relationship between high PMQ scores and concurrent measures of substance abuse, this scale still requires further development; it needs to be examined for its predictive and incremental validity, and its ideal length needs to be ascertained. The PMQ may prove most useful in evaluating patients who are in, or plan to be in, a multidisciplinary pain clinic so that clinicians can maximize patient selection for such programs. The tool is otherwise most useful in gauging the progress of pain patients already taking opioids. Chabal Five-Point Prescription Opiate Abuse Checklist This physician-administered checklist evaluates a series of behaviors that suggest or are consistent with prescription opiate abuse rather than relying on answers to specific questions. Patients meeting three or more of the following criteria are considered prescription opiate abusers: 1) overwhelming

18 S162 focus on opiate issues; 2) pattern of three or more early refills or escalating drug use without acute changes in their medical condition; 3) multiple telephone calls or visits to request additional opiates or early refills; 4) pattern of prescription problems due to lost, spilled, or stolen medications; and 5) supplemental sources of opiates from other providers or illegal sources. Chabal et al. [34] used the checklist to evaluate 403 pain-clinic patients including 76 patients (19%) who were using opiates for more than 6 months. Of this latter group, 21 patients (28%) met three or more criteria for prescription opiate abuse. The interrater reliability of the five-point checklist was greater than 0.9. Patients with prescription opiate abuse did not differ from the other chronic opiate users in terms of history of drug or alcohol abuse, or in scores on MAST, DAST, or psychosocial testing tools. This five-point checklist relies on easily observable behaviors in a clinic setting and accurately describes chronic pain patients who are abusing prescription opiates. The Chabal scale is a valuable tool to be used to gauge adherence once patients are already on opioid therapy. As such, it joins the Current Opioid Misuse Measure (COMM) and PMQ, but is the only evaluative instrument that is calculated by the clinician. Such a system is valuable for the clinician who is thinking through changes in treatment plans and levels of monitoring. This series of questions might be especially valuable for the pain clinician to put to a referring physician to quantify the degree of nonadherence that has been encountered prior to a pain specialist consultation. Prescription Drug Use Questionnaire (PDUQ) The PDUQ is a 42-item yes-or-no measure designed to be used by clinicians in an interview format with chronic pain patients who may be addicted to their pain medication. Requiring approximately 20 minutes to administer, the tool evaluates the following areas: pain condition, opioid use patterns, social and family factors, family history of pain and substance abuse, patient history of substance abuse, and psychiatric history. Compton et al. [35] evaluated the PDUQ in 52 consecutive opioid-treated chronic pain patients who were referred from a university-based multidisciplinary pain clinic for problematic narcotic use or drug-seeking behaviors. The majority of subjects were white (92%), female (60%), and currently married (58%) and suffered from more than Passik et al. one painful condition (65%). Opioids had been administered for a mean of 54 months. Overall, 20 patients (38.5%) met the diagnostic criteria for substance abuse, and 14 patients (26.9%) met the diagnostic criteria for substance dependence, whereas the remaining 18 patients (34.6%) did not meet the criteria for a substance abuse disorder. Scores were calculated based upon answers to the questionnaire items. Patients with substance abuse disorders had significantly higher scores than those without such disorders; all patients who had scores above 15 also met the defined diagnostic criteria for a substance abuse disorder. After excluding 10 cases with missing data, the internal consistency measured by Cronbach s alpha was On logistic regression, responses to three specific questions (patient believes he/she is addicted, pattern of increasing analgesic frequency or dose, and preference for specific analgesic or route of administration) were shown to best predict the presence of addictive disease, correctly classifying 93% of the study subjects. This interview format tool is useful and accurate in retrospectively evaluating patients nonadherence and the degree to which it corresponds to the DSM-IV diagnoses of SUD. COMM The COMM is a new 17-item self-report measure designed to identify aberrant drug-related behavior of patients on chronic opioid therapy [36]. The measure was designed to provide a simple, practical method for the continued assessment of current opioid misuse. A 40-item alpha version of the COMM was tested in 227 patients with chronic, noncancer pain. Patients also received the Aberrant Drug Behavior Index (ADBI), which relates positively with opioid medication misuse, and the Marlowe Crowne social desirability scale. Seventeen items were selected for further study based on good test retest reliability, a good correlation with ADBI, and a relatively poor correlation with the Marlowe Crowne social desirability scale. (The latter two criteria were used to help identify items that were better at revealing true aberrant behavior than capturing socially desirable responses.) For each item, respondents were asked to rate the frequency of a thought or behavior over the last 30 days, ranging from never (scored as 0) to very often (scored as 4). Questions capture signs and symptoms of drug misuse (e.g., problems with thinking), emotional/psychiatric issues, evi-

19 Screening Tools of Opioid Abuse dence of lying, appointment patterns (e.g., doctor shopping), and medication misuse/noncompliance (e.g., borrowing pain medications from friends, taking more than the prescribed dose). In the 227 patients, the correlation of these 17 items with the ADBI was 0.51, and with the Marlowe Crowne was The 1-week test retest reliability in a subset of the study subjects (N = 55) was good (0.86). In 86 individuals tested 3 months later, a receiver operating characteristic (ROC) analysis of the COMM with ADBI yields a high area under the curve of 0.92 (95% CI: ), suggesting the COMM could accurately detect aberrant behavior relative to the ADBI. A COMM score of 9 had a sensitivity of 77% and a specificity of 73% [36]. The COMM is a new measure that is useful in assessing adherence issues in patients already on opioid therapy. Its psychometrics are acceptable and it may prove quite useful in the reassessment of opioid therapy. Further study in more patients is needed, and long-term reliability must be confirmed. Structured Interview Tools Alcohol and Drug Diagnostic Instrument (ADDIS) and Substance Use Disorder Diagnostic Schedule (SUDDS) The ADDIS is the Swedish version of the SUDDS, which consists of a structured series of questions designed to elicit information to allow a diagnosis of substance abuse dependence and abuse in accordance with the DSM-III-R criteria. The ADDIS takes approximately minutes to complete. Jonasson and colleagues [37] administered the ADDIS interview to 243 orthopedic and chronic pain patients who were referred to an orthopedic hospital ward for rehabilitation. The study cohort was mostly female (61.5%) and married or cohabiters (69%), and had been on sick leave for at least 1 month. All patients were interviewed during their first week in the hospital concerning their use of alcohol and drugs. Eighty patients (33%) were found to have some form of substance abuse disorder according to the DSM- III-R criteria including 22% with abuse or dependence on analgesics, 14% on alcohol, and 7% on sedatives. When the DSM-IV criteria were used, 64 patients (26%) were diagnosed with a substance abuse disorder including 18% with abuse or dependence on analgesics, 9% on alcohol, and 4% on sedatives. Overall, 18 patients (8%) had S163 multiple substance abuse disorders. In the cohort, patients with an analgesic use disorder according to the DSM-III-R were more likely to be receiving dextropropoxyphene than those without this disorder (47% vs 26%, P = 0.003). ADDIS and SUDDS are suitable instruments for assessing dependence or abuse in chronic pain patients who are receiving opioid analgesics, but the structured interview may be time-consuming. Structured Clinical Interview for DSM-III-R, Patient Version (SCID-P) The SCID-P contains sections assessing lifetime and current diagnoses of alcohol or drug abuse or dependence, in addition to other sections to assess psychiatric disorders. The substance abuse section is often used as the gold standard for comparing the sensitivity and specificity of various screening tools. For example, Dyson and colleagues [23] used diagnoses from SCID-P as the criterion measure for assessing the validity of several different screening tools including CAGE-AID, SMAST, and DAST. Similarly, Gavin and colleagues [38] established the diagnostic validity of DAST for classifying patients according to comparisons with SCID-P. The SCID-P has performed very well in studies of patients admitted to psychiatric facilities. In a study of 89 psychiatric patients, Albanese and colleagues [39] assessed the accuracy of the substance abuse section of SCID-P (administered by two of the authors) vs urine toxicology tests and admission/discharge evaluations performed by the hospital staff. (The hospital staff consisted of either the treating psychiatrist, treating psychologist, residents, or interns under the supervision of a senior staff). The SCID-P was more accurate than urine toxicology and admission/discharge evaluations in identifying current substance abuse [39]. One would expect a trained research staff using a systematic substance abuse tool to perform better than a clinical team composed of staff with various levels of training and experience. Perhaps the most relevant observation of this study is that current clinical practice can be improved through the routine use of systematic tools like the SCID-P, and that even objective laboratory tests, like urine toxicology screens, have limitations. The SCID-P has yet to be validated in the pain treatment setting. A laptop-administered version that scores the tool immediately is also available.

20 S164 Discussion We have seen the emergence of a great many assessment tools with an intent of helping clinicians to better determine safe or risky candidates for opioid therapy. We feel this comprehensive review of the literature is a good first step toward gathering the various efforts and helping clinicians to decide what might be appropriate in their own practices and clinics. Performing a comprehensive literature search and review such as this, however, does have some shortcomings. Limitations of this strategy include the following: no prospectively defined criteria that assessed the strength of the study designs were used; article selection was based on a subjective assessment of perceived relevance based on the authors experience; and search terms may have been too narrow to capture all relevant publications. Conclusion A variety of self-administered and physicianadministered tools differing in their psychometrics and intended uses have been developed, but not all have been validated for use in chronic pain patients seen in a clinical practice setting. Several tools may be viable for assessing abuse potential among patients being considered for long-term opioid therapy; these tools include the SOAPP, DIRE, and ORT. Of the other measures designed for use in the pain management setting, many are intended to characterize the degree of medication misuse or the aberrant behavior that characterizes the patient s opioid use once he or she has been taking opioids for some length of time (PMQ, PDUQ, and COMM). Most screening tools, however, are more suitable for assessing current alcohol and/or drug abuse rather than the potential for such abuse. These tools include the RAFFT, TICS, KMSK scale, DAST, CAGE-AID, PMQ, PDUQ, DAPA-PC, and SASSI. Some tools have been adapted for use in adolescents (with, to date, no validation in adolescents with pain); these included the POSIT and DAST-A. Finally, some tools, including the ADDIS and SUDDS, provide a structured interview to allow the diagnosis of SUD akin to that provided in the alcohol and drug sections of the SCID-P. By recognizing the psychometrics and other features of each tool, clinicians can select the ones most appropriate for screening or ongoing assessment in their particular clinical population. While most of the these tools are easily accessible via the Web (i.e., painknowledge.org, and org, interested readers would be encouraged to contact the references corresponding to each tool to see if they are readily available and whether a cost might be associated. Acknowledgment Passik et al. Financial support for editorial assistance in the development of the manuscript was provided by Cephalon, Inc. References 1 Passik SD. Responding rationally to recent report of abuse/diversion of Oxycontin. J Pain Symptom Manage 2001;21: Passik SD, Kirsh KL, Donaghy KB, Portenoy RK. Pain and aberrant drug-related behaviors in medically ill patients with and without histories of substance abuse. Clin J Pain 2006;22: Passik SD, Kirsh KL, Donaghy K, Portenoy RK. Pain and aberrant drug-related behaviors in medically ill patients with and without histories of substance abuse. Clin J Pain 2006;22(2): Passik SD, Kirsh KL, McDonald MV, et al. A pilot survey of aberrant drug-taking attitudes and behaviors in samples of cancer and AIDS patients. J Pain Symptom Manage 2000;19(4): Passik SD, Kirsh KL, Whitcomb LA, et al. Monitoring outcomes during long-term opioid therapy for non-cancer pain: Results with the pain assessment and documentation tool. J Opioid Manage 2005;1(5): Passik SD, Kirsh KL, McDonald MV, et al. A pilot survey of aberrant drug-taking attitudes and behaviors in samples of cancer and AIDS patients. J Pain Symptom Manage 2000;19: Weissman DE, Haddox JD. Opioid pseudoaddiction An iatrogenic syndrome. Pain 1989;36: Akbik H, Butler SF, Budman SH, et al. Validation and clinical application of the Screener and Opioid Assessment for Patients with Pain (SOAPP). J Pain Symptom Manage 2006;32: Butler SF, Budman SH, Fernandez K, Jamison RN. Validation of a screener and opioid assessment measure for patients with chronic pain. Pain 2004; 112: Belgrade MJ, Schamber CD, Lindgren BR. The DIRE score: Predicting outcomes of opioid prescribing for chronic pain. J Pain 2006;7: Webster LR, Webster RM. Predicting aberrant behaviors in opioid-treated patients: Preliminary