Vol. 32 No. 4 October 2006 <strong>Addiction</strong> <strong>Behaviors</strong> <strong>Checklist</strong>345is using medications appropriately?’’) assessingeither appropriate or inappropriate opioidmedication use over the past month. It shouldbe noted that the rating clinicians were familiarwith both the patients (for the most partpatients were routinely seen by only one ortwo clinicians over the course <strong>of</strong> the study)and with current concepts <strong>of</strong> addiction, pseudoaddiction,physical dependence, and tolerance.Participants were assessed on a monthlybasis using the global clinical judgment.Opioid Medication Discontinuation for Noncompliance.An additional outcome measure wasthe occurrence <strong>of</strong> major medication contractviolations leading to opioid discontinuationover the course <strong>of</strong> the one-year study. Discontinuationdecisions were made as a group bythe multidisciplinary pain clinic treatmentteam based on results <strong>of</strong> urine toxicologyscreens, medication contract violations, andother factors, such as alcohol intoxication orinappropriate behaviors in clinic. <strong>The</strong> primaryreasons for discontinuing opioid medicationswere as follows: 1) alcohol or illicit substanceabuse, 2) noncompliance with prescribed medications(e.g., repeatedly running out <strong>of</strong> medicationsearly, seeking opioid medications fromother sources), and 3) noncompliance withclinic procedures (e.g., refusing to submiturine toxicology screens, repeatedly missingappointments, and expecting medication refillsvia phone request). Opioid medicationdiscontinuation decisions were not based specificallyon scores <strong>of</strong> the ABC, PDUQ, orGlobal Clinical Judgment evaluations.Visual Analog Scales (Pain Scores). At eachmonthly visit, patients received an assessment<strong>of</strong> both their usual level <strong>of</strong> pain and their worstlevel <strong>of</strong> pain since the last visit. This was measuredusing a 10 cm visual analog scale, rangingfrom 0 (no pain) to 10 (highest level <strong>of</strong>pain imaginable).Interrater Reliability ProceduresInterrater reliability estimates for the ABCwere determined for two separate samples <strong>of</strong>patients during the study (Sample 1, n ¼ 23;Sample 2, n ¼ 19). For each sample, the samepair <strong>of</strong> raters completed the ABC for thesame individuals during their visit, but eachrater was blind to the other’s scoring. Visitsto be rated were based on consecutive clinicappointments without preselection <strong>of</strong> patientsto be rated and regardless <strong>of</strong> position withinthe subject’s study year.ResultsPatient DemographicsMean age was 53 years (range 25e65 years).<strong>The</strong> study sample included 8 female and 128male participants. <strong>The</strong> high male to female ratio<strong>of</strong> study participants is typical <strong>of</strong> VA populations.Based on initial general categorization,105 participants had a primary musculoskeletalpain problem, 26 participants had a primaryneuropathic pain problem, and the remaining5 participants had multicategory pain problemsor it was unclear what category the primarypain problems fit into. Regardingmarital status, 53 participants (39%) were marriedor living with their partner, 19 participants(14%) were never married, 55 participants(41%) were divorced or separated, and 8 participants(6%) were widowed. Concerning employment,17 participants (13%) were workingfull time, 8 (6%) were working part time, andthe remainder (107 participants, approximately81%) were not working. Of the 107 participantswho were not working, approximately46% (48 participants) indicated that they wereunemployed or unable to work due to pain.ReliabilityInterrater Reliability. ABC scores for Sample 1(Raters 1 and 2) ranged from 0 to 10(mean ¼ 2.74, SD ¼ 2.53), and for Sample 2(Raters 2 and 3) ranged from 0 to 9(mean ¼ 1.93, SD ¼ 2.13). <strong>The</strong> Pearson correlationcoefficient between Raters 1 and 2 was0.94 (n ¼ 23, P < 0.01), and between Raters 2and 3 was 0.95 (n ¼ 19, P < 0.01). <strong>The</strong>se resultsshow strong interrater reliability for the ABC.ValidityIndividual Item Validity. Table 1 displays correlationsbetween item responses and providerglobal clinical judgment <strong>of</strong> problematic opioiduse. <strong>The</strong>se correlations were based on ABC assessmentstaken following the first four to fivemonthly visits, providing enough time for providersto more accurately make determinations
346 Wu et al.Vol. 32 No. 4 October 2006Table 1Correlation Between ABC Individual Itemsand Concurrent Global Clinical Judgmentat Approximately 4 MonthsAfter Start <strong>of</strong> Study (n ¼ 101)ABC ItemPearson Correlation Coefficient1 0.182 0.30 a3 0.54 a4 0.44 a5 0.64 a6 0.31 a7 0.0689 0.0210 0.0711 0.52 a12 0.25 b13 0.35 a14 0.25 b15 0.54 a16 0.21 b17 0.25 b18 0.51 a19 0.58 a20 0.04a P < 0.01.b P < 0.05.<strong>of</strong> patients’ opioid use patterns. As noted, Items2e6 and 11e19 displayed significant correlationswith the interviewer’s global clinical judgment.It should be briefly noted that Table 1does not include information on Variable 8,‘‘patient bought meds on the street,’’ as thisitem was not endorsed across all subjects duringthis initial period <strong>of</strong> follow-up visits. Over theyear follow-up in general, this item shows highlyinfrequent endorsement (see Discussion).Concurrent ValiditydGlobal Clinical Judgment.Using the provider’s global clinical judgmentas a criterion variable to operationalize appropriatemedication use, initial t-test resultsindicated that the mean ABC scores can significantlydistinguish between patients whodisplay appropriate use <strong>of</strong> opioid medicationsand patients who display inappropriate use <strong>of</strong>opioid medications. Mean ABC scores werehigher for inappropriate users (mean ¼ 5.31,SD ¼ 2.96) than appropriate users (mean ¼ 1.00,SD ¼ 1.27, t(16) ¼ 5.75, P < 0.001).Concurrent ValiditydPDUQ. Correlations weredetermined between mean ABC scores andPDUQ total scores. Mean ABC scores are theaverage ABC score across the initial 4e5months. <strong>The</strong> Pearson correlation coefficientbetween mean ABC scores and the PDUQ totalscore at this time point was 0.40 (P < 0.01).Sensitivity, Specificity, and Determination<strong>of</strong> Cut-Off ScoreUsing the global clinical judgment as the criterionmeasure for inappropriate opioid use,measurements <strong>of</strong> sensitivity and specificitywere determined comparing the cross-tabulationmeasurements <strong>of</strong> ABC total score vs. globaljudgment rating. Fig. 1 shows a graphical representation<strong>of</strong> sensitivity [true positives/(true positivesþ false negatives)] and specificity [truenegatives/(true negatives þ false positives)]measurements for different ABC total scores usingglobal clinical judgment as the criterionmeasurement <strong>of</strong> inappropriate opioid use. Because<strong>of</strong> a relatively small number <strong>of</strong> endorsedopioid misuse behaviors at any specific visit,ABC monthly data were combined across allvisits (n ¼ 998) in order to maximize calculation<strong>of</strong> sensitivity and specificity. <strong>The</strong>se results indicatethat a cut-<strong>of</strong>f score <strong>of</strong> 3 or greater on theABC shows optimal sensitivity and specificity indetermining whether a patient is displaying inappropriateopioid use. To evaluate the impact<strong>of</strong> combining data, calculations <strong>of</strong> sensitivityand specificity at the initial visit time pointwere also performed and reproduced the samefindings <strong>of</strong> optimal sensitivity/specificity ata cut-<strong>of</strong>f score <strong>of</strong> 3 or greater (e.g., using ABCdata from initial visit only, sensitivity ¼ 87.50%and specificity ¼ 86.14%).To further evaluate the utility <strong>of</strong> this cut-<strong>of</strong>fscore, t-tests were performed comparingPercent120100806040200SensitivitySpecificity0 1 2 3 4 5 6 7 8 9ABC Cut<strong>of</strong>f ScoreFig. 1. Sensitivity vs. specificity based on varyingABC total score. Criterion variable ¼ Interviewer’sGlobal Judgment (n ¼ 998). ABC data collectedacross all visits.