Volume represents cumulative monthly morphine milligram equivalent (MME) dose. See the Statistical Analysis subsection of the Methods section for additional details. Source: IMS Health LifeLink LRx Database (2010-2012) (IMS Health Incorporated).
eAppendix 1. Average Monthly Opioid Volume, Morphine-Milliequivalent (MME), and Days’ Supply in Florida and Georgia
eAppendix 2. Impact of Florida’s Policies on Aggregate Opioid Utilization Based on Varying Time Windows
eAppendix 3. Overall Impact of Policies on Monthly Prescription Opioid Utilization, Open Cohort
Customize your JAMA Network experience by selecting one or more topics from the list below.
Rutkow L, Chang H, Daubresse M, Webster DW, Stuart EA, Alexander GC. Effect of Florida’s Prescription Drug Monitoring Program and Pill Mill Laws on Opioid Prescribing and Use. JAMA Intern Med. 2015;175(10):1642–1649. doi:10.1001/jamainternmed.2015.3931
Prescription Drug Monitoring Program (PDMP) and pill mill laws are among the principal means states use to reduce prescription drug abuse and diversion, yet little high-quality evidence exists regarding their effect.
To quantify the effect of Florida’s PDMP and pill mill laws on overall and high-risk opioid prescribing and use.
Design, Setting, and Participants
We applied comparative interrupted time-series analyses to IMS Health LifeLink LRx data to characterize the effect of PDMP and pill mill law implementation on a closed cohort of prescribers, retail pharmacies, and patients from July 2010 through September 2012 in Florida (intervention state) compared with Georgia (control state). We conducted sensitivity analyses, including varying length of observation and modifying requirements for continuous observation of individuals throughout the study period.
Main Outcomes and Measures
Total opioid volume, mean morphine milligram equivalent (MME) per transaction, mean days’ supply per transaction, and total number of opioid prescriptions dispensed. Analyses were conducted per prescriber and per patient, in aggregate and after stratifying by volume of baseline opioid prescribing for prescribers and use for patients.
From July 2010 through September 2012, a cohort of 2.6 million patients, 431 890 prescribers, and 2829 pharmacies was associated with approximately 480 million prescriptions in Florida and Georgia, 7.7% of which were for opioids. Total monthly opioid volume, MME per transaction, days’ supply, and prescriptions dispensed were higher in Florida than Georgia before implementation. Florida’s laws were associated with statistically significant declines in opioid volume (2.5 kg/mo, P < .05; equivalent to approximately 500 000 5-mg tablets of hydrocodone bitartrate per month) and MME per transaction (0.45 mg/mo, P < .05), without any change in days’ supply. Twelve months after implementation, the policies were associated with approximately a 1.4% decrease in opioid prescriptions, 2.5% decrease in opioid volume, and 5.6% decrease in MME per transaction. Reductions were limited to prescribers and patients with the highest baseline opioid prescribing and use. Sensitivity analyses, varying time windows, and enrollment criteria supported the main results.
Conclusions and Relevance
Florida’s PDMP and pill mill laws were associated with modest decreases in opioid prescribing and use. Decreases were greatest among prescribers and patients with the highest baseline opioid prescribing and use.
Prescription opioids provide necessary analgesia to millions of Americans, yet the country faces soaring rates of opioid diversion, addiction, and overdose deaths.1-3 In the mid-2000s, Florida emerged as the epicenter of this epidemic. From 2003 to 2009, prescription drug overdose deaths in Florida increased more than 80%.4 In 2010, among the 100 US physicians purchasing the greatest amounts of oxycodone, 90 were in Florida.5 As a direct response, in 2010, Florida’s legislature addressed pill mills, or rogue pain management clinics where prescription drugs are inappropriately prescribed and dispensed.6 Florida’s pill mill law required these clinics to register with the state and have a physician-owner, created inspection requirements, and established prescribing and dispensing requirements and prohibitions for physicians at these clinics. The law’s implementation began in 2010, with additional elements becoming effective in July 2011 that prohibited prescriber dispensing of certain drugs.7 In September 2011, Florida’s Prescription Drug Monitoring Program (PDMP) became operational.8 Florida’s PDMP uses an electronic database to collect information about prescription drugs dispensed within the state. Florida-based prescribers and dispensers may voluntarily access the PDMP’s information to review individuals’ history to identify and address problematic practices such as physician shopping.9 Within the first 3 months of operation, more than 8000 prescribers registered, and the PDMP received almost 340 000 queries. After 1 year, in September 2012, the PDMP had received more than 2.3 million queries from more than 18 000 registered prescribers.10
Recent studies have identified promising findings after Florida’s legislative actions. Johnson and colleagues11 determined that Florida’s prescription drug–attributable mortality rate decreased by 23% from 2010 to 2012 and found declines in the prescribing rates of drugs often associated with overdose deaths. The findings of a recently published quasi-experimental study12 suggest that oxycodone-caused mortality declined 25% after PDMP implementation. A study by Surratt and colleagues13 found that diversion rates for prescription opioids in Florida were significantly reduced during a similar period. While the results of these studies suggest that Florida’s legislative initiatives may be having their desired effect, little is known about how these laws have influenced prescribing. Such information is important because it provides evidence of the practical effects of these laws on prescriber and patient behaviors,14 which greatly contribute to the amount of prescription opioids in circulation. We used a comparative interrupted time-series framework to quantify the degree to which Florida’s recent legislative actions influenced prescription opioid prescribing and use within the state compared with these practices in Georgia over the same period.
The study did not require institutional review board approval because it involved deidentified secondary data. We used IMS Health LifeLink LRx (IMS Incorporated) data,15 consisting of anonymized, individual-level prescription claims derived from tens of thousands of retail, food store, independent, and mass merchandiser pharmacies. They represent approximately 65% of retail prescriptions dispensed in the United States, including claims paid by Medicare, Medicaid, commercial insurance, and cash. Each prescription contains information about the retail transaction, the patient, and the prescriber. Transaction data include National Drug Code–level product information, quantity dispensed, days’ supply, source of payment, and 5-digit zip code of the dispensing pharmacy. Patient information includes sex, year of birth, a mail-order flag, and date of the first appearance in the data. Prescriber information is derived from the American Medical Association Physician Masterfile and includes specialty and 5-digit zip code.
We divided our study period into the following 3 segments: (1) a 12-month preintervention period (July 2010 through June 2011) preceding the policy changes; (2) a 3-month implementation period (July through September 2011), when the pill mill and PDMP laws were implemented; and (3) a 12-month postperiod after the policy changes (October 2011 through September 2012). Georgia served as a comparison state because it had not implemented a pill mill or PDMP law during our analysis period, had comparable trends in the outcomes of interest during the preintervention period, and is located in the same US region as Florida.
We identified approximately 12 million individuals who filled at least 1 prescription for any drug in Florida or Georgia from July 2010 through September 2012. We assigned each individual a state of residence based on the modal zip code reflected in their prescription claims. In our primary analyses, we used a 2-step process to derive a closed cohort of individuals to minimize bias from individuals entering or leaving the study population. First, we excluded 3.6 million patients (approximately 28%) who filled at least 1 prescription from stores that did not consistently report data to IMS Health throughout the study period. Second, we excluded 4.3 million individuals (approximately 36%) who did not fill claims for any drug within 3 months of the first and last months of the study period. We excluded approximately 2% of transactions with erroneous or extreme values (eg, negative quantities dispensed or transactions with morphine milligram equivalents [MMEs] >360 mg per transaction).
We examined 4 outcomes, derived on a monthly basis and examined at prescriber and transaction levels. First, we quantified total opioid volume prescribed using MME doses, which standardizes opioid prescriptions and accounts for differences in molecules and quantity and strength of doses dispensed.16 Second, we examined mean MME per transaction, which provides a sense of the magnitude of opioid use within individual transactions. Risk of opioid-related morbidity and mortality increases as MME increases,17 and experts have argued that clinicians should not exceed an MME of 80 to 100 mg daily across all prescribed opioids.16-18 Third, we examined mean days’ supply per transaction because greater days’ supply increases opportunities for abuse, diversion, and overdose. Fourth, we quantified total number of opioid prescriptions dispensed.
We applied a comparative interrupted time-series analysis to evaluate 2 related Florida laws on these outcomes, taking into account autocorrelation across time.19 Although we derived our outcomes as monthly measures, we averaged the 3 months when these 2 laws were initially implemented (ie, implementation period), giving us 25 observations per state (12 monthly preimplementation observations, a 3-month implementation period, and 12 monthly postimplementation observations). We used linear regression to quantify the policy changes’ effect on each outcome, and a linear trend was found to fit the data well. Two interaction terms—one with a state indicator (Florida or Georgia) and a period indicator and another with a state indicator (Florida or Georgia) and a postimplementation monthly indicator—were our main focus, which represented the difference in change of level and prescription rate (trend) from the preimplementation to postimplementation periods between the states. We performed additional analyses stratifying prescribers and patients into groups based on total opioid volume prescribed or used during the preimplementation period.
To account for clustering of observations across time within each state, we adjusted for autocorrelation when constructing models using the generalized Durbin-Watson test. The R2 of all models was higher than 0.95, reflecting large sample sizes and little variation on the outcomes of interest over time. All analyses were performed using statistical software (SAS, version 9.4 [proc autoreg command with nlag function]; SAS Institute Inc).
We performed sensitivity analyses to examine whether our results were robust according to varied assumptions. First, we varied length of observation in the preimplementation and postimplementation periods using 6-month and 18-month intervals. Second, to mitigate the potential for selection bias from analyzing only those patients with claims at the study period’s beginning and end, we repeated our analyses using an open cohort in which we permitted patients to drop in and out. Third, given the reformulation of oxycodone in August 2010, we repeated our analyses with the exclusion of extended-release oxycodone.
Our final cohort consisted of 2.6 million patients, 431 890 prescribers, and 2829 pharmacies. From July 2010 through September 2012, the cohort filled approximately 480 million prescriptions, of which 7.7% were for opioids. Eligible prescription opioids accounted for 7.5% of captured prescriptions in Florida and 7.8% of captured prescriptions in Georgia. Most prescriptions (77.4%) were filled in chain stores, with fewer filled by independent retailers (9.9%), food stores (9.0%), and mass merchandisers (3.7%).
Total opioid volume (327.2 vs 118.3 kg), mean MME per transaction (54.88 vs 46.55 mg), and mean days’ supply per transaction (18.74 vs 16.23 days) were higher in Florida than Georgia during the preimplementation period (eAppendix 1 in the Supplement). Total opioid volume in Florida decreased approximately 4% (from 327.2 to 313.9 kg) from the preimplementation to postimplementation periods, whereas mean MME per transaction decreased 5.7% (from 54.88 to 51.74 mg), and mean days’ supply per transaction increased 3.8% (from 18.74 to 19.46 days) over the same period. In Georgia, overall total opioid volume decreased 2.3%, mean MME per transaction decreased 4.7%, and mean days’ supply per transaction increased 5.7% from preimplementation to postimplementation.
The Figure shows trends in observed and predicted total opioid volumes for Florida and Georgia from July 2010 through September 2012. From July 2010 through June 2011, monthly total MME per transaction in Florida was consistently 3 times higher than that in Georgia. This difference begins to gradually decrease when Florida’s law prohibiting prescriber dispensing of opioids was implemented in July 2011. The Figure stratifies the same outcome by patients in the top 10th, 5th, 3rd, and 1st percentiles of opioid use at baseline in Florida and Georgia. Monthly total MME per transaction among patients with high opioid use in Florida increased from July 2010 through June 2011. However, during the postintervention period, from October 2011 through September 2012, total monthly MME per transaction decreased by approximately 36%. Comparatively, decreases in Georgia’s monthly total MME per transaction during this period were negligible.
Table 1 summarizes the policies’ overall changes in prescription opioid sales in Florida compared with Georgia. Although there was no statistically significant change in levels of the outcomes at the time of policy implementation, the policies were associated with statistically significant reductions in trends in total opioid volume and mean MME per transaction. For example, the policies resulted in a statistically significant relative reduction of approximately 2.5 kg/mo in total opioid volume in Florida compared with Georgia from the preimplementation to postimplementation periods, a decrease equivalent to a reduction approximately equal to half a million 5-mg tablets of hydrocodone bitartrate per month. The policies were associated with a statistically significant 0.45 mg/mo relative reduction in mean MME across all transactions in Florida compared with Georgia. The policies had no apparent effect on days’ supply per transaction or on total number of opioid prescriptions dispensed.
Table 2 summarizes differences between monthly actual and predicted values of total opioid volume, mean MME per transaction, mean days’ supply, and total number of opioid prescriptions in Florida had the policies not been implemented. For example, during the first 6 months after implementation, there was a 0.59% difference between total opioid volume dispensed in Florida and total opioid volume expected had the PDMP and pill mill laws not been implemented. One year after these changes, the policies were associated with a 2.52% reduction in total opioid volume, 5.64% reduction in mean MME per transaction, no change in days’ supply per transaction, and 1.35% reduction in total number of opioid prescriptions dispensed.
There were modest and statistically significant decreases in total opioid volume among patients whose baseline opioid use was greatest (Table 3). For example, among patients at the 90th percentile of baseline use, the policies were associated with a statistically significant relative reduction of 5.1 kg/mo in total opioid volume. Significant decreases in MME per transaction attributable to the laws were limited to those with the highest levels of opioid use at baseline (the 90th, 95th, 97th and 99th percentiles) and were of a similar magnitude at approximately 1-mg/mo decline per transaction. There were statistically significant relative reductions in total number of opioid prescriptions dispensed to patients at the 90th, 95th, 97th, and 99th percentiles. For example, among patients at the 95th percentile of baseline use (40 694 patients in Florida and 19 647 patients in Georgia), the policies were associated with a reduction of approximately 740 opioid prescriptions dispensed per month.
Table 4 summarizes changes at the prescriber level. For example, among prescribers at the 99th percentile of total opioid volume at baseline, the policy change was associated with a statistically significant relative reduction of approximately 3.0 kg/mo in total opioid volume, or the equivalent of 600 000 5-mg hydrocodone bitartrate tablets per month. The strongest changes were on trends in total opioid volume and mean MME per transaction among those with the highest baseline prescription volume, although there were small, statistically significant relative increases in mean days’ supply per transaction among these subpopulations of prescribers.
In analyses using 18-month and 6-month (rather than 12-month) windows, the results’ direction and statistical significance were similar, although the effects’ magnitude varied (eAppendix 2 in the Supplement). Analyses using an open cohort showed similar results: the magnitude and statistical significance of the relative change in trends across outcomes were usually greater, but general trends remained the same (eAppendix 3 in the Supplement). The results were similar after the exclusion of extended-release oxycodone from our analyses.
State-based PDMP and pill mill laws have become prominent policy mechanisms to address prescription drug abuse and diversion.20,21 We used comparative interrupted time-series analyses to characterize changes associated with these laws in opioid prescribing and use in Florida, a state with high rates of opioid-related injuries and deaths. We found that jointly the policies were associated with modest reductions in total opioid volume, mean MME per transaction, and total number of opioid prescriptions dispensed, with no apparent effect on duration of treatment. These reductions were generally limited to patients and prescribers with the highest baseline opioid use and prescribing. Our results are important given soaring rates of prescription opioid abuse, as well as the prominent role that laws have in shaping states’ responses to the epidemic.
Our findings highlight the need for more evidence demonstrating the effect of PDMP and pill mill laws. A recently published ecological study22 using data from the Automation of Reports and Consolidated Orders System (ARCOS)23 from 1999 to 2008 found that PDMPs had no overall influence on dispensing of MMEs per capita and noted that the effect varied dramatically between states, which is likely explained by large differences among states’ PDMPs. Our study included Florida and Georgia as comparison states. The results from another ecological study24 using ARCOS data from 1997 to 2003 suggested that PDMPs were associated with declines in quantity of oxycodone shipments. However, these studies did not consider PDMP utilization itself.
Our study adds to a growing evidence base evaluating state policies designed to curb epidemic rates of opioid prescribing. Differences in outcome measurements, exposures, data sources, and analytic approaches have led to mixed conclusions about PMDPs’ influences on opioid prescribing and make direct comparison of our results difficult. Few, if any, studies have evaluated pill mill laws exclusively, and only a handful have considered these laws within a suite of policy interventions.11,13 Our findings suggest that PDMP and pill mill law implementation jointly was associated with reductions in mean MME per transaction among patients and prescribers with the highest baseline use in Florida relative to Georgia. However, given wide variability in PDMP functioning, the generalizability of these results is likely limited to states with similarly designed PDMPs, pill mill laws, and sociodemographic profiles.
Most prescribers support policies such as those considered by our group.14 Given this support and reductions in total opioid volume and mean MME per transaction among high-volume prescribers that we observed after implementation of Florida’s policies, other states may want to consider similarly comprehensive regulatory approaches. This initiative might require prescribers to register with their state’s PDMP and routinely query its data,25 although such measures must be balanced by concerns regarding usage mandates.26 To ensure that high-volume prescribers are aware of these policies, states should engage in targeted outreach campaigns, particularly among subspecialties known to most commonly prescribe opioids.27 In addition, states should consider drug treatment services because recent findings have confirmed that, as the prescription opioid supply decreases or is reformulated, individuals who misused these drugs turn to heroin.28,29
Our study has several limitations. First, although more than 85% of prescription opioids are dispensed through retail channels,23 our analyses excluded other distribution channels, although this exclusion would likely lead us to underestimate the effects of the policies of interest. Second, our data provided an incomplete picture of the retail market, and patients may enter and leave the database we used for various reasons. To account for this possibility, we derived a closed cohort for our primary analysis and required patients to have filled at least 1 prescription for any drug within 3 months of the study period’s beginning and end. Third, our sensitivity analyses yielded substantial differences in the magnitude of the policy effects, although direction, statistical significance, and substantive interpretation did not differ. To determine sustained effect of these policies, longer-term trends should be examined. Fourth, we focused on opioid prescribing and use rather than opioid-related injuries or deaths. However, sales of opioids are highly correlated with rates of injuries and death from their use.30,31 Fifth, our analyses did not account for possible spillover effects from Florida’s laws that may have influenced opioid prescribing and use in Georgia, leading to a possible overestimation of the effects of Florida’s laws. Sixth, our analyses did not allow us to determine the individual effect of Florida’s PDMP and pill mill laws because these policies were implemented at essentially the same time. Therefore, we evaluated these policies together, consistent with Florida’s framing of its multifaceted approach to addressing prescription drug abuse and diversion.7 However, our findings regarding high-use patients and prescribers suggested that Florida’s pill mill law may have been the primary law of influence. This possibility could be further studied in states that have enacted a pill mill law but have lower levels of opioid prescribing and use.
To curb epidemic rates of prescribing, morbidity, and mortality associated with opioid misuse and diversion, states have spent millions of dollars implementing policies designed to reduce excessive dispensing of these products. Paramount to these efforts are studies empirically testing these policies’ effectiveness and a growing evidence base informing policy makers of the benefits and harms that may result. Our study adds to this evidence base and using pharmacy claims data shows that implementation of Florida’s PDMP and pill mill law was associated with modest decreases in opioid use and prescribing among patients and providers with high levels of opioid use at baseline relative to Georgia, a comparison state.
Accepted for Publication: June 17, 2015.
Corresponding Author: G. Caleb Alexander, MD, MS, Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, 615 N Wolfe St, Room W6035, Baltimore, MD 21205 (email@example.com).
Published Online: August 17, 2015. doi:10.1001/jamainternmed.2015.3931.
Author Contributions: Dr Alexander had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: All authors.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Rutkow, Chang, Daubresse, Alexander.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Chang, Daubresse, Webster, Stuart.
Obtained funding: Alexander.
Administrative, technical, or material support: Rutkow, Chang, Daubresse, Alexander.
Study supervision: Rutkow, Alexander.
Conflict of Interest Disclosures: Dr Alexander reported being the chair of the US Food and Drug Administration’s peripheral and central nervous system advisory committee, reported serving as a paid consultant to a mobile start-up (PainNavigator) and to IMS Health Incorporated, and reported being a member of an IMS Health scientific advisory board. This arrangement has been reviewed and approved by The Johns Hopkins University in accord with its conflict of interest policies. No other disclosures were reported.
Funding/Support: This work was funded by the Robert Wood Johnson Foundation Public Health Law Research program and by the Centers for Disease Control and Prevention.
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; analysis or interpretation of the data; and preparation or final approval of the manuscript before publication.
Disclaimer: The statements, findings, conclusions, views, and opinions contained and expressed in this article are based in part on data obtained under license from the following IMS Health Incorporated information services: IMS Health LifeLink LRx Database (2010-2012), IMS Health Incorporated. All rights reserved. The statements, findings, conclusions, views, and opinions contained and expressed herein are not necessarily those of IMS Health Incorporated or any of its affiliated or subsidiary entities.
Create a personal account or sign in to: