Original Research Article Do Prescription Monitoring Programs Impact State Trends in Opioid Abuse/Misuse?pme_

Transcription

1 Pain Medicine 2012; 13: Wiley Periodicals, Inc. OPIOIDS, SUBSTANCE ABUSE & ADDICTIONS SECTION Original Research Article Do Prescription Monitoring Programs Impact State Trends in Opioid Abuse/Misuse?pme_ Liza M. Reifler, MPH,* Danna Droz, RPh, JD, J. Elise Bailey, MSPH,* Sidney H. Schnoll, MD, PhD, Reginald Fant, PhD, Richard C. Dart, MD, PhD,* and Becki Bucher Bartelson, PhD* *Rocky Mountain Poison and Drug Center, Denver, Colorado; Ohio State Board of Pharmacy, Columbus, Ohio; Pinney Associates, Baltimore, Maryland; University of Colorado, School of Medicine, Aurora, Colorado, USA Reprint requests to: Becki Bucher Bartelson, PhD, Rocky Mountain Poison and Drug Center, 777 Bannock St., MC 0180, Denver, CO 80204, USA. Tel: ; Fax: ; Disclosure: This study is unfunded; none of the authors received direct compensation for their role in this study. At the time this article was written, LMR, DD, JEB, RCD, and BBB were employees or advisory board members for the DHHA/RMPDC RADARS System. Abstract Objective. Prescription monitoring programs (PMPs) are statewide databases containing prescriber and patient-level prescription data on select drugs of abuse. These databases are used by medical professionals or law enforcement officials to identify patients with prescription drug use patterns indicative of abuse or providers engaging in illegal activities. Most states have implemented PMPs in an attempt to curb prescription drug abuse and diversion. However, assessment of their impact on drug abuse is only beginning. This study aimed to evaluate the relationship between PMPs and opioid misuse over time in two drug abuse surveillance data sources. Methods. Data from the RADARS System Poison Center and Opioid Treatment surveillance databases were used to obtain measures of abuse and misuse of opioids. Repeated measures negative binomial regression was applied to quarterly surveillance data (from 2003 to mid-2009) to estimate and compare opioid abuse and misuse trends. PMP presence was modeled as a time varying covariate for each state. Results. Results support an association between PMPs and mitigated opioid abuse and misuse trends. Without a PMP in place, Poison Center intentional exposures increased, on average, 1.9% per quarter, whereas opioid intentional exposures increase 0.2% (P = 0.036) per quarter with a PMP in place. Opioid treatment admissions increase, on average, 4.9% per quarter in states without a PMP vs 2.6% (P = 0.058) in states with a PMP. In addition to the time trend, population and a measure of drug availability were also significant predictors. A secondary analysis that classified PMP based upon ideal characteristic showed consistent though not significant results. Conclusions. Two observational data sources offer preliminary support that PMPs are effective. Future efforts should evaluate what PMP characteristics are most effective and which opioids are most impacted. Key Words. Prescription Monitoring Programs; Opioid Abuse; Opioid Misuse Introduction Prescription opioid abuse poses a substantial public health burden. Opioid abuse is the intentional, nonmedical use of opioids to obtain a euphoric or psychotropic effect, and according to the 2008 National Survey on Drug Use and Health (NSDUH), an estimated 34.8 million 434

2 Americans used prescription pain relievers nonmedically at least once in their lifetime [1]. In 2009, an estimated 5.3 million persons age 12 or older reported using pain relievers nonmedically in the past month [2], making pain relievers the second most commonly abused drug [3]. Prescription pain killer abuse is not only substantial in terms of lifetime and recent use, but 17.1% of NSDUH respondents reported using prescription pain relievers as their first drug [2]. Furthermore, the nonmedical use of prescription opioid pain relievers has increased over the past decade. According to 2008 NSDUH results, persons age 12 or older reporting lifetime non-medical pain reliever use increased from 12.6% in 2002 to 14.0% in 2008 [1]. Although federal regulations are in place to limit the distribution of controlled substances, many states have taken additional actions in an effort to curb increasing prescription drug abuse. The United States Drug Enforcement Administration (DEA) categorizes all substances with abuse potential into five schedules, which define a range of regulatory activities restricting substance distribution. Drugs controlled by DEA Schedules II-V comprise 12.5% of the top 400 retail prescription drugs [4]. Given common legitimate medical use and resulting availability of these controlled substances, many states have implemented additional legislation in an effort to better monitor and control the prescribing and dispensing of these substances. State prescription monitoring programs (PMPs) are an example of legislative efforts to curb prescription drug abuse. PMPs gather patient-level prescription data on select controlled substances. Most states provide this prescription information to practitioners and pharmacists [5] to aid in the provision of appropriate medical care, including identification of potential drug abusers in order to limit drug access (prevention) and to hasten treatment counseling. Some states also provide data to drug diversion law enforcement officials. As of third quarter of 2009, 34 states have implemented PMPs, and most collect prescription data electronically. A few states limit their data collection to Schedule II controlled substances, but many states have expanded PMPs to monitor other drugs with abuse potential. While some variations across PMPs exist, there are many similarities across programs. All the PMPs collect a standard set of information on patient identity, drug and quantity dispensed, dispense date, dispenser identity, and prescriber identity. Additional fields allow each state to meet other mandates determined by its legislature. While no state has published results on PMP effectiveness, some states have used PMP data to explore drug abuse behavior. In a 2006 survey of 18 state PMP administrators, Oklahoma and Maine reported conducting an evaluation of PMP public health impact, and eight others reported that future investigations with this focus are very important in their state [6]. State PMP data have been used to characterize patient and prescriber behavior, Opioid Abuse and Prescription Monitoring Programs yielding both an empirical definition for drug-seeking behavior known as doctor shopping [7] and support that doctors adjust prescriptions after accessing a patient s information in a PMP database [8]. While these studies suggest possible avenues through which PMPs may affect drug abuse, there are no state level published studies on the impact of PMPs on prescription drug abuse rates. PMPs effectiveness in reducing drug abuse has not been well studied on the national level [9]. Ideally, an evaluation of PMP effectiveness should establish a relationship between the supply of opioids and abuse. An observational study on health insurance claims found that the presence of a state PMP was associated with lower county opioid claims rates [10]. Likewise, an analysis of the DEA ARCOS data from 1997 to 2003 supported that proactively monitoring PMP states have smaller increases over time in the availability of prescription opioids [11]. However, the same authors analyzed prescription opioid abuse treatment admissions using Treatment Episode Data Set (TEDS) data and failed to find differences in states based on PMP presence. Using National Center for Health Statistics state mortality data, a recent study failed to find an association between PMPs and lower rates of accidental opioid deaths [12]. Given the lack of nationwide PMP studies, additional research is needed to assess how state prescription opioid abuse rates relate to state PMP implementation. As more states consider PMP legislation, an evaluation of the effectiveness of PMPs in curbing drug abuse is needed. If PMPs effectively limit opioid supplies, abuse will decline over time. We hypothesize that PMP implementation occurs in states which have higher overall prescription opioid abuse, misuse, and diversion. Furthermore, even though national abuse is increasing over time, we hypothesize that state PMPs are associated with a reduction in this increasing time trend. Methods Data Sources The Researched, Abuse, Diversion and Addiction-Related Surveillance (RADARS ) System is a prescription drug abuse, misuse, and diversion surveillance system that offers multiple perspectives along the drug abuse pathway through the use of six unique programs. These programs collect prescription opioid and stimulant drug data from different target populations and report data on a quarterly basis with geographic specificity at the three-digit ZIP code (3DZ) level. For the purposes of this analysis, we focused on data from two RADARS System Programs, Poison Center and Opioid Treatment, due to the populations each represent, geographic coverage, and duration of surveillance. Data from all available quarters were included in the analyses: Poison Center (from first quarter of 2003 to third quarter of 2009) and Opioid Treatment (from first quarter of 2005 to third quarter of 2009). 435

3 Reifler et al. RADARS System Poison Centers are geographically dispersed and increased in number from 11 centers in the first quarter of 2003 to 48 (of 60 total centers) in the third quarter of The total population served by these centers ranged from 65,983,332 in the first quarter of 2003 to 237,688,161 in the third quarter of 2009 (United States Census Bureau). Each center obtained institutional review board approval for their participation in the RADARS System. Centers use trained health care professionals to collect call information and case notes, on a standard computerized data collection form and submit these data weekly to the RADARS System. Exposures reported to the RADARS System contain more complete information than those reported to the National Poison Data System (American Association of Poison Control Centers) and undergo a rigorous quality control process. Case mentions are defined as intentional exposures managed by participating poison centers for the drugs of interest. Intentional exposure reasons include suspected suicide, withdrawal, abuse, intentional misuse (incorrect usage of a drug without the intent to obtain a psychotropic effect), and intentional unknown cases. A 2006 study showed that the agreement between addiction experts and poison center specialists for coding of the intentional exposure subclassifications is moderate. Thus, intentional exposures are used in the analysis as a surrogate for misuse and abuse [13]. Cases are assigned to the reported 3DZ of the exposed individual s residence. Trends in intentional exposures have been shown to track the trends in NSDUH data. Seventy-four opioid treatment centers from 33 states, all members of the American Association for the Treatment of Opioid Dependence, participate in the RADARS System Opioid Treatment Program. These treatment centers are federally approved opioid agonist treatment programs and follow federal methadone treatment protocols requiring an opioid-dependence diagnosis and an addiction history of at least 1 year. The research protocol was approved by the Institutional Review Board of the National Development and Research Institutes, Inc., and oral informed consent is obtained from program enrollees to complete an anonymous questionnaire. Questionnaire items focus on the patient s opioid of choice, opioid use in the past month, and lifetime opioid abuse. Cases are defined as selfreported use of a prescription or illicit opioid to get high in the past 30 days. The questionnaire has been validated against unlabeled drug photo selections [14]. Cases are assigned to the reported 3DZ of the patient s residence. Information on prescription monitoring program enactment and implementation in each state was compiled using resources available through the National Alliance for Model State Drug Laws. Date of PMP enactment was determined through publicly available information on each state s Website. Date of implementation was considered as the date when the PMP began collecting data. If this information was not available through the above resources, the state PMP administrator was contacted to obtain the information. Data Analysis In order to get an overall picture of prescription opioid abuse and misuse prevalence over time in states with and without active PMPs, data used in this analysis were limited to five Schedule II opioid classes covered by all 34 PMPs that are similar in availability and prescribed use: fentanyl, hydromorphone, methadone, morphine, and oxycodone. RADARS System cases are typically summarized as prevalence per 100,000 persons and per 1,000 unique recipients of dispensed drug (URDD) at the 3DZ level. The population of each 3DZ was determined using United States Census data. URDD is an estimate of the number of individuals filling a prescription for a given opioid, and it is based on sampling data supplied by pharmacy chains, independent pharmacies, mass merchandisers, and food stores using 814 projection regions throughout the United States. Unique recipient of dispensed drug data for each RADARS System opioid of interest and each 3DZ are purchased from SDI Health (Plymouth Meeting, PA, USA) on a quarterly basis. In comparison to opioid misuse prevalence in the population, opioid misuse prevalence per URDD describes prescription opioid abuse and misuse in the context of opioid availability within a community. Three-digit ZIP code data (number of mentions, URDD, and population) were summed into state level, quarterly data for each of the five opioids of interest. In order to visually display the association between PMPs and abuse and misuse rates over time, we plotted oxycodone abuse and misuse rates per 1,000 URDD. States rates were plotted in three groups based on timing of PMP implementation. States were grouped according to whether they had 1) no active PMP before or during the surveillance period; 2) a PMP implemented before the surveillance period; or 3) a PMP implemented during the surveillance period. Implementation is defined as the point in time when a PMP begins to collect prescription data. These groupings were only used for graphical representation of the data. To statistically evaluate the effect of PMPs on abuse and misuse counts, a generalized estimating equation was fit individually to data from Poison Center Program and Opioid Treatment Program. Repeated measures regression methods were utilized, where state and drug class were specified as random effects. The covariance structure within state and drug classes, which is important for correct inferences of parameters, was selected as firstorder autoregressive process over time. As the outcome of interest (drug abuse and misuse) is in the form of counts, negative binomial distribution was used to accommodate for the large variability of the outcome and the restriction to non-negative values. The five predictor variables evaluated in the model were 1) the natural log of population; 2) the natural log of URDD; 3) time-point measured in year quarters; 4) duration of the PMP program; and 5) the presence of a PMP at each quarter (within state). Thus, PMP presence was treated as a time varying covariate to allow for inclusion of states implementing 436

4 a PMP during the surveillance period. An interaction between time and PMP status was included to model the difference in opioid misuse over time with and without a PMP. An alpha of 0.05 was used to determine significance of coefficients in the models. The model can be interpreted as fitting two summary regression lines to the data, one in the absence of a PMP and one in the presence of a PMP. The difference in the intercepts address the difference in average initial opioid misuse, while the difference in slopes represents the different trends in opioid misuse with and without a PMP. All slope and intercept parameters are adjusted for the covariates mentioned earlier and are on the log scale due to the negative binomial distribution. An additional analysis was conducted to further explore the relation between state PMP program features and opioid misuse over time. A subset of superior PMPs was identified based on programs that have been in effect for a long duration (at least by the start of 2002), provide PMP data reports directly to health care workers, and capture data on drugs at least through Schedule IV. Five states met these criteria: Idaho, Kentucky, Utah, Nevada, and Michigan. These states were compared in the model to standard PMPs and no PMPs, using a categorical variable. For the Poison Center Program, Utah and Nevada do not participate and thus only three states were included in the group of superior PMPs. All analyses were performed in SAS, version 9.2, for Windows (Cary, NC, USA). Results Figure 1 summarizes the progression of states PMP implementation during the Poison Center Program study Figure 1 Prescription monitoring program (PMP) status summary of states reporting to the RADARS System Poison Center Program, first quarter of 2003 through third quarter of Opioid Abuse and Prescription Monitoring Programs period. Of 44 reporting states from the first quarter of 2003 through the third quarter of 2009, 17 states (38.6%) initiated a PMP before the study period, and 12 states (15.6%) implemented a PMP during the study period. Timing of PMP implementation for each state is listed in the Appendix (Table A1). Figure 2 displays the oxycodone intentional exposure case mentions per 1,000 URDD over time for each state reporting to the Poison Center Program. These plots, grouped by timing of PMP implementation, demonstrate that the rates in each state fluctuate over time but also follow a time trend. To evaluate whether PMPs mitigate increasing intentional exposure mentions over time, multivariate negative binomial regression models were run (Table 1). The variable for duration of a PMP is not shown in the final results because it failed to reach significance in the Poison Center and Opioid Treatment Programs and was thus dropped from the models. Poison Center model results support that URDD, population, time, and the interaction of time and PMP status (P = 0.040) were significant predictors of intentional exposure mentions. The test of the interaction indicates that the slope of the regression lines with and without PMPs differs significantly. The Poison Center results show that initially increasing counts are mitigated by a PMP. The lack of significance of PMP status indicates that the intercept of the regression lines in the presence or absence of a PMP did not differ. As the negative binomial results in Table 1 are displayed on the natural log scale, the magnitude of the parameter estimates is somewhat difficult to interpret directly. As indicated by the positive sign of parameter estimates for population, URDD, and quarter, intentional exposure mentions increase with population, drug availability, and time. 437

5 Reifler et al. (a) (b) (c) Figure 2 Poison Center Program natural log of oxycodone unique recipients of dispensed drug (URDD) rates by state, first quarter of 2003 through third quarter of Plots of the number of cases per 1,000 URDD over time by state, for three groups of states: (a) those with no active prescription monitoring program (PMP) during the study period, (b) those with a PMP active prior to the start of the study period, and (c) those with a PMP enacted during the study period. Each line represents a separate state. Note that data are shown on a natural log scale to reflect the use of negative binomial regression and to focus on the low URDD rates over time. The first panel shows overall increasing abuse rates in the absence of a PMP, while the second panel of those states with a PMP in place prior to the study period shows a diminished increase in abuse rates. Also striking is that more variability in abuse rates over time is seen for smaller states, e.g., NH or HI than for others, e.g., MA. 438

6 Opioid Abuse and Prescription Monitoring Programs Table 1 Negative binomial model results in predicting the natural log of abuse and misuse mentions Poison Center Opioid Treatment Estimate SE Z Test Statistic P Value Estimate SE Z Test Statistic P Value Intercept < <0.001 ln population < <0.001 ln URDD < <0.001 PMP status Quarter < <0.001 PMP*Quarter Regression coefficients, standard errors (SEs), and P values from the negative binomial regression model for natural log (ln) of intentional exposures. The quarter variable represents the slope of the regression line in the absence of a PMP, while the PMP*Quarter interaction gives the deviation of the slope in the presence of a PMP from the slope without a PMP. Exponentiation of the slopes results in the relative risk for a one unit change in quarter (see Table 2). URDD = unique recipients of dispensed drug. The effect of primary interest is the slope of opioid misuse over time in states with and without PMPs, represented by quarter and the PMP status by quarter interaction. Table 2 displays these factors in a more directly interpretable manner. In general, the mean increasing rate in Poison Center intentional exposure mentions is times per quarter in states without a PMP, while case mentions increase by only times per quarter in states with a PMP. Over a 1-year period, this translates to an expected increase in intentional exposure mentions of 8% in the absence of a PMP but only a 0.8% increase in the presence of PMP. The Opioid Treatment Program (Table 1) results indicate that URDD, population, and quarter were significant indicators of seeking treatment at opioid treatment centers. The main effect of PMP status was not significant; however, the interaction of PMP status and time was marginally significant (P = 0.058). The slope of opioid misuse may differ in states with PMPs, but we are unable to determine if these results could be due to chance. The lack of significance of the PMP status indicates that the intercept of the regression lines in the presence or absence of a PMP did not differ. Table 2 Negative binomial model results relative risk* over time (in quarters) of prescription opioid abuse and misuse in nonprescription monitoring program (non-pmp) and PMP states Table 2 also displays the parameter estimates for the Opioid Treatment Program in a more directly interpretable manner. Opioid Treatment Program results show that initially increasing counts are mitigated by a PMP. For opioid treatment, the predicted increase over time is per quarter in states without a PMP vs per quarter in states with a PMP. If annual increases are computed over the study period, then opioid abusers reporting to treatment centers were increasing without a PMP by 20% per year but only increasing 11% with a PMP. Results from the sensitivity analysis are consistent with the main analysis, even though not all effects of interest reached statistical significance (Table 3). The slopes over time based on the presence of a standard vs superior PMP were not significantly different in the Poison Center Program (P = ) but were significant for the Opioid Treatment Program (P = 0.027). States with standard PMPs were expected to have a slower increase in opioid misuse over time than states without PMPs and states with superior PMPs were expected to have even smaller increases over time when compared to states without PMPs. Table 4 represents these model results in terms of quarterly increases. Increases per quarter in states with no PMP, standard PMP, and superior PMP diminish as expected across these groups; however, the superior PMP increase did not differ significantly from the standard PMP increase. The increase in opioid misuse over time for the Opioid Treatment Program also follows expected diminishing pattern across the three PMP groups, but only the states with superior PMPs were significantly different than states without PMPs. RR over Time Poison Center Opioid Treatment Without PMP (1.008, 1.030) (1.036, 1.063) With PMP (0.992, 1.012) (1.009, 1.044) * Relative risks (RRs) were calculated through exponentiation of parameter estimates involving quarter. Relative risks for a one unit change in quarter with and without a PMP. Discussion Ultimately, model results from the RADARS System Poison Center and Opioid Treatment Programs support that PMPs are associated with a mitigation of increasing opioid abuse and misuse over time in both the general population as well as within the population seeking treat- 439

7 Reifler et al. Table 3 Negative binomial model results for sensitivity analysis Poison Center Opioid Treatment Estimate SE Z Test Statistics P Value Estimate SE Z Test Statistics P Value Intercept < <0.001 ln population < <0.001 ln URDD < <0.001 Standard PMP Superior PMP Quarter < <0.001 Quarter*Standard PMP Quarter*Superior PMP Regression coefficients, standard errors (SEs), and P values from the negative binomial regression model for ln (intentional exposures). The quarter variable represents the slope of the regression line in the absence of a prescription monitoring program (PMP), while the PMP*Standard*Quarter interaction gives the deviation of the slope in the presence of a standard PMP from the slope without a PMP. The Quarter*Superior*PMP variable gives the deviation of the slope in the presence of a superior PMP from the slope without a PMP. Exponentiation of the slope coefficients results in the relative risk for a one unit change in quarter (see Table 2). URDD = unique recipients of dispensed drug. ment at Opioid Treatment Programs. This analysis provides the first available support that PMPs are associated with smaller increases in drug abuse/misuse over time. A similar analysis using TEDS data failed to find any significant difference among states with PMPs, although results did suggest PMPs lower prevalence in pain reliever supply and abuse [11], and another study failed to find an association between PMPs and lower rates of unintentional opioid overdose deaths [12]. The data sources and analyses used are subject to limitations. Each RADARS System Program captures a different segment of drug misuse and abuse cases. Additionally, these observational data are subject to selfreport and selection bias. Poison Center specialists record information self-reported by callers into a standardized database. Specialists may make errors during entry or callers may be motivated to report socially desirable details. Opioid treatment data are collected through an anonymous self-report questionnaire and are also subject to self-report bias and data entry error. All Poison Center and Opioid Treatment data are quality checked by the RADARS System, and invalid or missing data on key fields Table 4 analysis Relative risk* model results for sensitivity RR over Time Poison Center Opioid Treatment Without PMP (1.009, 1.030) (1.036, 1.063) With standard PMP (0.994, 1.015) (1.007, 1.049) With superior PMP (0.984, 1.019) (0.990, 1.043) * Relative risks (RRs) were calculated through exponentiation of parameter estimates for the variables involving Quarter. Relative risks for a one unit change in quarter with and without a prescription monitoring program (PMP). are queried with sites. Unresolved cases are removed from the data, which may result in more conservative estimates of opioid misuse. Finally, self-report observational data is subject to selection bias; data only capture individuals contacting poison centers or entering into treatment programs, and thus, the estimate of abuse is likely an underestimate. Also of note, some states do not report to every surveillance program. Utah and Nevada are considered superior PMPs but they do not participate in the Poison Center Program and this limited the power of the secondary analysis of states based on the presence of superior, standard, or no PMPs. Due to these limitations, generalizing the analysis results from one program to the whole population is cautioned against. A second limitation of this study is that the variations in quality and coverage of PMPs were not included in the main analysis. As PMPs are created and implemented by each state independently, the use and maintenance of each program may be unique in a variety of ways (e.g., the presence of Web access to patient data, frequency of upload of new information, turnaround time for patient report, extent of use, and availability to physicians during patient encounters). For the Poison Center Program data, intentional exposure increases over time were lowest for superior PMPs but failed to differ significantly from the increase over time for standard PMPs. This analysis, however, is limited in that only three states classified as having superior PMPs participate in the RADARS System. A significant dose or strength of PMP effect was seen in the Opioid Treatment Program when PMPs were stratified by quality indicators. Further analyses of the individual elements of PMPs that are effective are beyond the scope of this article, but are warranted in future studies. It is possible that the results of this analysis reflect that states with PMPs may be inherently different from states 440

8 that do not have prescription surveillance legislation and programs. States likely to have active PMPs may have different opioid misuse and abuse trends. As this analysis is observational in nature, results cannot support a definitive PMP effect. Nonetheless, a few features of the data modeled support that the decrease in the trend over time is due to a real association with PMPs. Many states enacted PMPs during the surveillance period, and data captured time trends before and after the programs became active; states with PMPs are diverse in demographics and region and many other factors that could influence supply and misuse of opioids. Finally, the programs themselves are diverse due to specific legislative, financial, and administrative environment of each state. These features suggest that despite the many inherent difference across all states, active PMPs may be associated with a relative decrease over time in opioid misuse. References 1 SAMHSA. Results from the 2008 national survey on drug use and health: Detailed tables; Available at: tabs/lotsect8pe.htm (accessed November 20, 2010). 2 SAMHSA. Results from the 2009 national survey on drug use and health: Volume I. Summary of national findings; Available at: gov/nsduh/2k9nsduh/2k9resultsp.pdf (accessed November 20, 2010). 3 SAMHSA. Results from the 2009 national survey on drug use and health: Detailed tables; Available at: LOTSect1pe.htm (accessed November 20, 2010). 4 Code of Federal Regulations 21: Parts 1300, 1304, 1306, Electronic prescriptions for controlled substances; final rule pp The Alliance of States with Prescription Monitoring Programs. Who is authorized to request patient prescription data; Available at: pmpalliance.org/content/who-authorized-requestpatient-prescription-data (accessed September 6, 2011). Opioid Abuse and Prescription Monitoring Programs 6 Katz N, Houle B, Fernandez KC, et al. Update on prescription monitoring in clinical practice: A survey study of prescription monitoring program administrators. Pain Med. 2008;9(5): Katz N, Panas L, Kim M, et al. Usefulness of prescription monitoring programs for surveillance Analysis of Schedule II opioid prescription data in Massachusetts, Pharmacoepidemiol Drug Saf 2010;19(2): Baehren DF, Marco CA, Droz DE, et al. A statewide prescription monitoring program affects emergency department prescribing behaviors. Ann Emerg Med 2010;56(1):19 23.e Cooper JR, Czechowicz DJ, Molinari SP, Petersen RC. Impact of prescription drug diverson control systems on medical practice and patient care. National Institute on Drug Abuse Research Monograph Series. 1993: Curtis LH, Stoddard J, Radeva JI, et al. Geographic variation in the prescription of Schedule II opioid analgesics among outpatients in the United States. Health Serv Res 2006;41(3 Pt 1): Simeone R, Holland L. An evaluation of prescription drug monitoring programs. Simeone Associates, Inc.; Available at: com/docs/2008/03/federal_prescription_monitoring_ report.pdf (accessed November 20, 2010). 12 Paulozzi LJ, Kilbourne EM, Desai HA. Prescription drug monitoring programs and death rates from drug overdose. Pain Med 2011;12(5): Smith MY, Dart R, Hughes A, et al. Clinician validation of Poison Control Center (PCC) intentional exposure cases involving prescription opioids. Am J Drug Alcohol Abuse 2006;32(3): Smith M, Rosenblum A, Parrino M, Fong C, Colucci S. Validity of self-reported misuse of prescription opioid analgesics. Subst Use Misuse 2010;45(10):

9 Reifler et al. Appendix Table A1 Number of three-digit ZIP codes per state reporting to the RADARS System programs by year and quarter and by prescription monitoring program (PMP) status RADARS System Poison Center RADARS System Opioid Treatment State Year Quarter of First PMP Data Collection State Total # 3-Digit ZIP Codes (3DZ) Study Period: First Last YearQuarter 3DZ in Program: Avg (min, max) Study Period: First Last YearQuarter 3DZ in Program: Avg (min, max) AK No active PMP q3 2009q3 5 (5, 5) 2005q2 2009q3 2 (1, 3) AL 2006q q2 2009q3 19 (19, 19) 2005q1 2009q3 9 (4, 12) AR No active PMP 14 Not reporting 2005q1 2009q1 2 (1, 5) AZ 2008q q4 2009q3 5 (3, 11) 2005q1 2009q3 2 (1, 3) CA 1998q q1 2009q3 60 (60, 60) 2005q1 2009q3 15 (10, 27) CO 2007q q1 2009q3 17 (17, 17) 2005q1 2009q3 3 (1, 5) CT 2008q q4 2009q3 10 (10, 10) 2005q1 2009q3 5 (4, 7) DC No active PMP 6 Not reporting 2005q1 2005q1 2 (2, 2) DE No active PMP 3 Not reporting 2005q2 2005q4 1 (1, 1) FL No active PMP q1 2009q3 22 (6, 26) 2005q1 2009q3 13 (7, 17) GA No active PMP q3 2009q3 22 (22, 22) 2005q1 2009q3 8 (2, 14) HI 1992q q1 2009q3 2 (2, 2) 2008q1 2009q1 1 (1, 1) IA 2008q q1 2009q3 26 (26, 26) 2005q1 2009q3 2 (1, 4) ID 1997q q1 2009q3 7 (7, 7) 2006q1 2007q4 1 (1, 1) IL 1985q q1 2009q3 29 (29, 29) 2005q2 2009q3 3 (1, 6) IN 1994q q4 2009q3 20 (20, 20) 2005q1 2009q3 6 (4, 12) KS No active PMP q1 2009q3 19 (19, 19) 2005q2 2006q2 1 (1, 2) KY 1999q q1 2009q3 28 (28, 28) 2005q1 2009q3 10 (2, 20) LA 2008q q4 2009q3 13 (13, 13) 2005q1 2009q3 3 (1, 6) MA 1994q q4 2009q3 19 (19, 19) 2005q1 2009q3 5 (1, 10) MD No active PMP q4 2009q3 12 (10, 13) 2005q1 2008q2 1 (1, 4) ME 2004q q3 11 (11, 11) 2005q1 2009q3 5 (3, 8) MI 1988q q1 2009q3 20 (20, 20) 2005q1 2009q3 4 (1, 8) MN 2010q q1 2009q3 17 (17, 17) 2005q1 2009q3 1 (1, 1) MO No active PMP q4 2009q3 26 (26, 26) 2005q1 2009q3 3 (1, 5) MS 2005q2 12 Not reporting 2005q1 2009q3 3 (1, 5) MT No active PMP q1 2009q3 10 (10, 10) 2005q2 2005q3 1 (1, 1) NC 2007q3 20 Not reporting 2005q1 2009q3 5 (1, 11) ND 2007q q1 2009q3 9 (9, 9) 2005q4 2005q4 1 (1, 1) NE No active PMP q1 2009q3 13 (13, 13) 2005q1 2005q2 1 (1, 1) NH No active PMP q1 2009q3 9 (9, 9) 2005q1 2009q3 3 (1, 4) NJ No active PMP q2 2009q3 20 (20, 20) 2005q1 2009q3 8 (5, 11) NM 2005q3 13 Not reporting 2005q1 2009q3 2 (1, 4) NV 1997q q1 2009q3 7 (6, 8) 2005q1 2009q3 5 (1, 6) NY 1972q q1 2009q3 22 (10, 23) 2005q1 2009q3 8 (3, 22) OH 2006q q1 2009q3 26 (18, 29) 2005q1 2009q3 8 (2, 12) OK 1990q q1 2009q3 17 (17, 17) 2005q2 2009q1 1 (1, 1) OR No active PMP q3 2009q3 10 (10, 10) 2005q2 2009q3 2 (1, 3) PA 2002q q3 28 (25, 29) 2005q1 2009q3 5 (1, 10) RI 1997q q4 2009q3 2 (2, 2) 2005q2 2009q2 2 (1, 2) SC 2008q q4 2009q3 10 (10, 10) 2005q1 2009q2 3 (1, 8) SD No active PMP q1 2009q3 8 (8, 8) 2005q2 2005q2 3 (3, 3) TN 2007q q4 2009q3 16 (16, 16) 2005q1 2009q3 8 (6, 11) TX 1982q q1 2009q3 41 (33, 46) 2005q1 2009q3 4 (1, 8) UT 1996q3 8 Not reporting 2005q2 2009q3 2 (1, 3) VA Pilot in 8 3DZ q1 2009q3 28 (28, 28) 2005q1 2009q3 10 (3, 19) VT No active PMP q1 2009q3 9 (9, 9) 2005q1 2009q3 2 (1, 3) WA 1984q3* q1 2009q3 14 (14, 14) 2005q2 2009q3 3 (1, 5) WI No active PMP q4 2009q3 18 (18, 18) 2005q2 2005q3 2 (1, 2) WV 2003q q4 2009q3 22 (22, 22) 2005q1 2009q3 7 (1, 15) WY 2004q q1 2009q3 12 (12, 12) 2006q2 2009q3 1 (1, 1) * WA has a PMP that is disciplinary only. Given the restrictive scope, it was not counted as a state with an active PMP in analyses. 442