State Policies for Prescription Drug Monitoring Programs and Adverse Opioid-Related Hospital Events

Abstract

Background:

State policies to optimize prescriber use of Prescription Drug Monitoring Programs (PDMPs) have proliferated in recent years. Prominent policies include comprehensive mandates for prescriber use of PDMP, laws allowing delegation of PDMP access to office staff, and interstate PDMP data sharing. Evidence is limited regarding the effects of these policies on adverse opioid-related hospital events.

Objective:

To assess the effects of three PDMP policies on adverse opioid-related hospital events among patients with prescription opioid use.

Research Design:

We examined 2011–2015 data from a large national commercial insurance database of privately insured and Medicare Advantage patients from 28 states with fully operating PDMPs by the end of 2010. We used a difference-in-differences framework to assess probabilities of opioid-related hospital events and association with the implementation of PDMP policies. Analysis was conducted for adult patients with any prescription opioid use, a subsample of patients with long-term prescription opioid use, and stratified by older (65+) versus younger patients.

Results:

Comprehensive use mandates were associated with a relative reduction in the probability of opioid-related hospital events by 28% among patients with any opioid and 21% among patients with long-term opioid use. Such reduction was greater (in relative terms) among older patients despite the lower rate of these events among older than younger patients. Delegate laws and interstate data sharing were associated with limited change in the outcome.

Conclusion:

Comprehensive PDMP use mandates was associated with meaningful reductions in opioid-related hospital events among privately insured and Medicare Advantage adults with prescription opioid use.

Introduction

Between 1999 and 2015, drug overdose deaths involving prescription or illicit opioids increased nearly six-fold.¹ The Centers for Disease Control and Prevention estimated annual health and social costs of prescription opioid misuse at $55 billion, of which $20 billion was spent on emergency department and inpatient care for opioid poisonings.² Despite the general perception that the opioid crisis has shifted from prescription opioids to heroin and other synthetic opioids (e.g. fentanyl), in 2017, prescription opioids still accounted for over 40% of opioid overdose deaths, and 11.4 million Americans reportedly misused prescription opioids.³

Prescription Drug Monitoring Programs (PDMPs) are a prominent strategy taken by states to address the opioid crisis. PDMPs are state-wide electronic databases of controlled substances dispensed at retail pharmacies. PDMPs provide information to prescribers about controlled substances received by a patient, and thus can assist prescribers with identifying possible misuse among patients while ensuring legitimate use for pain management. In recent years, states have implemented policies to address low prescriber participation in PDMPs.^4,5 One prominent type of policy is legislative mandates that prescribers use the PDMP at the point of care. Evidence is accumulating that comprehensive use mandates that apply to all prescribers in all settings and do not rely on prescriber discretion were associated with reductions in opioid prescriptions presenting high risk of misuse and overdose among the privately insured populations^6,7 and Medicare.⁸ Other PDMP policies have been implemented to lower prescriber burden when using the PDMP (e.g., legislations allowing prescriber delegation of PDMP use to office staff) or to make information more complete or useful (e.g., enabling interstate PDMP data sharing). There is some evidence that PDMP delegation laws were associated with reductions in high-risk opioid prescriptions.⁶

Policies designed to increase prescriber PDMP use may lead to decreases in opioid-related hospital events as a result of reductions in high-risk prescription opioid use. On the other hand, one major unintended consequence of these policies (and of comprehensive use mandates in particular) may be reduced or discontinued opioid prescribing regardless of patient need or risk. Concerns are mounting that patients whose opioid therapy gets terminated abruptly without effective alternative pain management strategies are being “pushed” to illicit opioids, which, in turn, may lead to adverse events associated with illicit opioid use and overdose.^9,10 Because claims-based diagnoses may offer low specificity to differentiate prescription opioid-related events from illicit opioid-related events, our analysis assessed the net effects of PDMP policies on adverse opioid-related hospital events.

Our study makes several contributions to the literature. First, we examined the effects of three prominent policies designed to improve prescriber take-up and use of PDMP on adverse opioid-related hospital events including emergency department (ED) visits and inpatient admissions. Earlier studies largely focused on the implementation of PDMPs and opioid overdose deaths using vital statistics; they generated mixed findings regarding PDMP implementation and fatal or non-fatal drug overdoses.^11–20 Other studies examining features of PDMPs (e.g. prescriber mandates, interstate data sharing) found that prescriber mandates were associated with decreased fatal drug overdoses.^18,20 Second, we estimated such effects for younger versus older patients. National statistics between 2002 and 2014 suggest differentially evolving epidemics by age; although the rate of opioid misuse remained much lower among people ages 50+ compared to the younger population, the rate almost doubled among those 50+ in contrast with a decline among those ages 18–25 and relative stability among those ages 26–49.²¹ Given that PDMP policies are intended to reduce high-risk prescription opioid use, it is important to examine how PDMP policies have impacted younger and older patients differently. Third, we focused on implications of PDMP policies for privately insured and Medicare Advantage adults, two populations that have been less studied despite bearing large absolute burden of pain and of prescription opioid misuse.²²

Methods

Data

We used 2011–2015 data from the Health Care Cost Institute (HCCI), a large commercial insurance claims database that covers about 50 million individuals per year enrolled in a health insurance plan offered or administered (i.e. self-insured plans) by Aetna, Humana, or UnitedHealthcare, including employer-sponsored, individual market, and Medicare Advantage plans.²³ The data contain beneficiary enrollment information, inpatient facility, outpatient, physician, and pharmacy claims. We ended the study period on December 31, 2015 due to data availability.

Subjects

Our study population was adults (ages 18+) with private or Medicare Advantage insurance who had filled at least one opioid prescription during 2011–2015, a target population of PDMP policies. The HCCI data capture prescriptions dispensed by pharmacies and covered by the patient’s pharmacy or prescription drug benefit. We further considered a sub-population of patients with at least one episode of long-term opioid use (thus at heightened risk of opioid misuse or overdose) during the study years.²⁴ Long-term opioid use was defined as continuous use of prescription opioids for 90 days or longer; a gap of 30 days or more with no opioids determined the end of an episode.^24–29 Our unit of analysis was patient half-year. We thus required patients to be continuously enrolled in a half-year (January-June, or, July-December, during 2011–2015) to be included in the analysis. Additionally, we excluded patient half-years in which the patient had a diagnosis of cancer or sickle cell disease to focus on patients receiving opioids for non-cancer and non-sickle cell disease related pain.

Measures

We focused on three types of policies increasingly implemented by states during our study years to enhance prescriber PDMP use. Comprehensive use mandates are legislations that require all prescribers to use the PDMP at the point of care when prescribing opioids/controlled substances for the first time and at least annually thereafter. Delegate laws allow prescriber delegation of PDMP access to office staff. Interstate data sharing enables prescriber access to PDMP information from other states. These policies were selected because of their potential to increase prescriber use of PDMPs and change prescribing behaviors, implementation by a sufficient number of states during our study years, relative homogeneity of policies implemented across states, and the availability of reliable data on implementation dates.

Implementation of PDMP comprehensive use mandates and of delegate laws was determined based on the effective date of the legislation. The National Alliance for Model State Drug Laws³⁰ provided effective dates of pertinent state legislations. During our study period, the most robust way of enabling interstate data sharing was through state participation in PMP InterConnect®, provided by the National Association of Boards of Pharmacy (NABP).³¹ Participating states sign a memo of understanding and develop an interface to connect their PDMP with PMP InterConnect. Each state controls access to its data via a dashboard within PMP InterConnect, allowing selected states to share the data. The “go-live” date provided by NABP for each state defined the implementation of interstate data sharing.

We restricted our analysis to patients in states with a fully operating PDMP by the end of 2010. By the end of our study period (2015), seven states had implemented comprehensive use mandates, 23 had allowed prescribers to delegate PDMP use to an office staff member, and 22 participated in interstate data sharing via PMP InterConnect (Table 1). For each of the three policies, we set the policy indicator to one for each full half-year post the effective date of the policy in a given state, and zero otherwise.

Table 1.

PDMP policy implementation dates

State	PDMP accessible	Comprehensive Use Mandate	Delegate Access	Interstate Data Sharing
Alabama	2007		2013
Arizona	2008		2014	2012
California	2009
Colorado	2008		2014	2013
Connecticut	2008	2015	2015	2012
Idaho	1998			2014
Illinois	2008		2015	2013
Indiana	2007		2007	2011
Iowa	2009		2012	2015
Kentucky	1999	2012	2012	2013
Louisiana	2009		2013	2013
Maine	2005		2011
Michigan	2003			2012
Minnesota	2010		2010	2013
Mississippi	2005		2005	2013
Nebraska	2009
Nevada	1997	2015		2014
New Mexico	2005		2005	2012
North Carolina	2007		2013
North Dakota	2007		2008	2012
Ohio	2006	2015	2011	2011
Oklahoma	2006	2015	2015	2015
South Carolina	2008		2014	2012
Tennessee	2007	2013	2012	2013
Utah	1997		2012	2014
Vermont	2009		2013
Virginia	2006		2009	2011
West Virginia	2004	2012	2011	2014

Notes: The PDMP implementation dates were collected from the National Alliance for Model State Drug Laws (NAMSDL), correspondence with program administrators, and additional searches of state legislation websites. Comprehensive use mandate effective dates were collected from the Pew Charitable Trusts, NAMSDL, and additional searches of state legislation websites. Delegation law effective dates were collected from NAMSDL and additional searches of state legislation websites. Inter-state data-sharing effective dates were provided by the National Association of Boards of Pharmacy.

Our primary outcome of interest was an indicator of having any opioid-related hospital event (ED visit or inpatient admission), defined as those with a diagnosis of opioid dependence, opioid abuse, or opioid poisoning in any of the observed (primary and secondary) diagnostic codes (see Appendix Table 1, Supplemental Digital Content or SDC1, http://links.lww.com/MLR/B996). We focused on the dichotomous outcome of having at least one such event in a half-year rather than the count of events because the majority of patients (77%) who ever experienced any event had only one event. We took this inclusive approach in defining our outcome given the serious under-coding of opioid poisonings in claims data.^32,33 In several sensitivity analyses, we adopted more restrictive definitions including using primary diagnoses only and using poisoning codes only.

Analyses

We exploited staggered implementation of PDMP policies across states to estimate difference-in-differences models, using patients in states that had not implemented the policies as controls. We conducted an event study analysis to assess the assumption that the outcome in states implementing a given policy followed a parallel trend with what was seen in states that did not implement the policy. Violation of the parallel trend assumption could be suggestive of policy endogeneity. To produce an event study, we replaced the binary policy indicators with categorical indicators capturing 6-month intervals (0–6, 7–12, 13–18, 19+ months) before and after policy implementation. The reference point was the half-year before the policy implementation. The event study produces a visual representation of the differences between implementing and non-implementing states in time intervals before (as a test of the parallel trend assumption) and after (as an estimate of time-variant policy effect) policy implementation. We restricted our analysis to the twenty-eight states with a fully operating PDMP (with user access) by the end of 2010. States that launched a PDMP more recently were more likely to have implemented PDMP-enhancing policies at the same time as they launched PDMPs, making it challenging to isolate the effects of the PDMP policies from those of launching a PDMP.

Additionally, for both the population of all opioid users and sub-population of patients with long-term opioid use, we estimated the models separately for patients ages 18 to 64 and patients aged 65 or older. While prescription opioid misuse, illicit opioid use, and opioid-related adverse events were lower among older compared with younger adults,^21,34 older adults suffer from more pain conditions and were more likely to experience long-term opioid use (Appendix Table 3, SDC1, http://links.lww.com/MLR/B996).

We estimated linear probability models of the probability of having at least one opioid-related hospital event during a defined half-year, among patients who were ever dispensed an opioid during our study period. Each model included the three dichotomous PDMP policy indicators, dichotomous state indicators to control for differences among states that did not change over time, and dichotomous indicators of half-years to control for nationwide trends in the outcome. Each model controlled for patient sex and age, indicators for pain-related diagnoses (back, neck, arthritis related, or other), an indicator for any mental health condition, and indicators of alcohol use disorder, drug use disorder, and tobacco use, based on claims-based diagnostic codes in a given half-year (Appendix Table 2, SDC1, http://links.lww.com/MLR/B996). The clinical conditions were identified if any claim (inpatient, outpatient, or physician) during the half-year had any diagnosis (primary and secondary) suggesting the condition. Because patients may contribute multiple half-years to the analysis, we derived robust standard errors clustered at the patient-level.

Results

Patients with any prescription opioid use

Our sample included 31,482,222 half-years from 6,423,416 unique patients who had filled at least one opioid prescription during the study years. Of all patient half-years, 0.10% experienced at least one opioid-related hospital event, 0.04% experienced at least one opioid-related ED visit, and 0.07% experienced at least one opioid-related inpatient admission (Table 2).

Table 2.

Summary statistics of study outcomes and sample characteristics

	Full sample	Long-term use sample
	Mean	Mean
Opioid-related events (%)
Any ED or inpatient	0.099	0.373
ED visit	0.040	0.145
Inpatient admission	0.067	0.255
Opioid poisoning events (%)
Any ED or inpatient	0.027	0.102
ED visit	0.014	0.046
Inpatient admission	0.014	0.059
Characteristics (%)
Insurance
Employer based	69.64	43.00
Individual market	3.75	1.93
Medicare Advantage	26.62	55.08
Age group
18–24 years	9.09	1.27
25–34 years	14.04	5.03
35–44 years	16.35	9.97
45–54 years	18.98	18.30
55–64 years	17.56	23.74
65–74 years	14.62	23.42
75–84 years	7.08	13.39
85+ years	2.28	4.86
Female	54.69	58.64
Any mental health condition	13.83	24.63
Alcohol use disorder	0.62	1.19
Drug use disorder	0.88	2.78
Tobacco use	2.63	4.81
Back pain	15.60	34.95
Neck pain	6.13	12.39
Arthritis pain	28.61	51.42
Other pain	13.81	25.11
Number of observations	31,482,222	2,088,000
Number of unique patients	6,423,416	358,940

Notes: Full sample are adults (ages 18+) with private insurance or Medicare Advantage who had filled at least one opioid prescription during study years, did not have a diagnosis of cancer or sickle cell, and were living in states that had an operational PDMP by December 2010. Long-term use sample are patients with at least one long-term episode during study years. A long-term opioid episode was defined as continuous use of prescription opioids for 90 days or longer; a gap of 30 days or more with no opioid use was used to determine the end of a long-term opioid use episode. The unit of analysis was patient half-year.

The event study analysis indicated that implementing and non-implementing states largely followed parallel trends in our primary outcome in the four half-years leading to policy implementation (Appendix Figures 1 and 2, SDC1, http://links.lww.com/MLR/B996).

Implementation of comprehensive use mandates was associated with a reduction in the probability of any opioid-related hospital event from 0.101% (95% confidence interval or CI: 0.100, 0.103) to 0.073% (95% CI: 0.067, 0.080), amounting to a relative reduction of 27.7%. Delegate laws were associated with a 5.3% reduction, from 0.102% (95% CI: 0.099, 0.104) to 0.096% (95% CI: 0.094, 0.099). Interstate data sharing was associated with a 6.2% reduction, from 0.102% (95% CI: 0.100, 0.104) to 0.096% (95% CI: 0.093, 0.098) (Figure 1).

Figure 1.

Changes in the probability of any opioid-related hospital event among patients with any opioid use, 2011–2015

Source: Author’s analysis of data for 2011–15 from the Health Care Cost Institute’s insurance claims database.

Notes: The exhibit shows the predicted changes in the probabilities of outcomes associated with implementation of PDMP policies among privately insured or Medicare Advantage adults who were ages 18+, had at least one opioid prescription in the study period, did not have a diagnosis of cancer or sickle cell, and lived in the twenty-eight states that had an operating program by December 2010. The unit of analysis was patient half-year. The whiskers indicate 95% confidence intervals. The percentages (relative effects) indicate the difference between probabilities with and without the policies.

Patients with long-term opioid use

Our sample included 2,088,000 half-years from 358,940 unique patients with at least one episode of long-term opioid use. Of these half-years, 0.37% experienced an opioid-related hospital event, 0.15% experienced an opioid-related ED visit, and 0.26% experienced an opioid-related inpatient admission (Table 1).

In this sub-sample, comprehensive use mandates were associated with a 20.6% reduction in the probability of any opioid-related hospital event, from 0.380% without a mandate (95% CI: 0.370, 0.390) to 0.302% with a mandate (95% CI: 0.261, 0.342). Neither delegate laws nor interstate data sharing were associated with a statistically significant change in the probability of such an event (Figure 2).

Figure 2.

Changes in the probability of any opioid-related hospital event among patients with long-term opioid use, 2011–2015

Source: Author’s analysis of data for 2011–15 from the Health Care Cost Institute’s insurance claims database.

Notes: The exhibit shows the predicted changes in the probabilities of outcomes associated with implementation of PDMP policies among privately insured or Medicare Advantage adults who were ages 18+, had at least one opioid prescription in the study period, did not have a diagnosis of cancer or sickle cell, and lived in the twenty-eight states that had an operating program by December 2010. Long-term use sample are patients with at least one long-term episode during study years. A long-term opioid episode was defined as continuous use of prescription opioids for 90 days or longer; a gap of 30 days or more with no opioid use was used to determine the end of a long-term opioid use episode. The unit of analysis was patient half-year. The whiskers indicate 95% confidence intervals. The percentages (relative effects) indicate the difference between probabilities with and without the policies.

Sensitivity analyses

Our results are robust across multiple sensitivity analyses. First, we ended our study period at June 30, 2015 to avoid any implication of the transition from ICD-9 to ICD-10 codes starting in October 2015 (Appendix Figures 3A and 3B, SDC1, http://links.lww.com/MLR/B996). Second, we defined our outcome of opioid-related hospital events based on the primary diagnosis only (rather than using primary and secondary diagnoses) (Appendix Figures 4A and 4B, SDC1, http://links.lww.com/MLR/B996). Results based on these two analyses (in terms of relative changes in outcomes) were similar to those of the main analysis. Lastly, we defined our outcome as hospital events involving opioid poisonings only (Appendix Figures 5A and 5B, SDC1, http://links.lww.com/MLR/B996). Poisoning events, however, were rare and were known to be seriously underreported. While comprehensive use mandates were associated with reduced opioid poisoning events among patients with any opioid use, the policy was not associated with significant changes among patients with long-term opioid use.

Policy effects for older versus younger patients

Among patients with any opioid use, the rate of opioid-related events among patients 65 or older was about one-third of that of younger patients (0.0004 vs 0.0012). Among patients with long-term opioid use, the rate of opioid-related events for older patients was about one-fifth of that of younger patients (0.001 vs 0.005) (Appendix Table 3, SDC1, http://links.lww.com/MLR/B996).

As shown in Figure 3, although the rate of opioid-related hospital events was much lower among older patients, the relative reduction associated with a comprehensive use mandate was substantially higher (46–52%) among older patients compared to younger patients (10–20%). Delegate laws were associated with reductions in opioid-related hospital events for patients of all ages (5–13%); the estimates did not achieve statistical significance among patients with long-term opioid use. Interstate data sharing was associated with reductions in opioid-related hospital events for patients ages 65 or older (6–22%).

Figure 3.

Changes in the probability of any opioid-related hospital event associated with a state’s implementation of comprehensive PDMP use mandates, 2011–2015

Source: Author’s analysis of data for 2011–15 from the Health Care Cost Institute’s insurance claims database.

Notes: The exhibit shows the predicted changes in the probabilities of outcomes associated with implementation of comprehensive use mandates among privately insured or Medicare Advantage adults who were ages 18+, had at least one opioid prescription in the study period, did not have a diagnosis of cancer or sickle cell, and lived in the twenty-eight states that had an operating program by December 2010. Long-term use sample are patients with at least one long-term episode during study years. A long-term opioid episode was defined as continuous use of prescription opioids for 90 days or longer; a gap of 30 days or more with no opioid use was used to determine the end of a long-term opioid use episode. The unit of analysis was patient half-year. The whiskers indicate 95% confidence intervals. The percentages (relative effects) indicate the difference between probabilities with and without a mandate.

Discussion

Using a large national commercial insurance database, we found that state policies designed to enhance prescriber PDMP use and, in particular, comprehensive use mandates for prescriber use of PDMP at the point of care, were associated with as much as 28% reduction in the probability of a hospital event related to opioid dependence, abuse, or overdose over a half-year. This finding suggests that 27,486 fewer individuals with private insurance or Medicare would have experienced opioid-related hospital events in the second half of 2015 alone if comprehensive use mandates had been implemented in every state (Appendix Table 4, SDC1, http://links.lww.com/MLR/B996). Delegate laws and interstate data sharing were associated with limited reduction (5–6%) in the probability of an opioid-related hospital event among patients receiving at least one opioid prescription and no significant change among patients with long-term prescription opioid use. Despite a much lower rate of such events, patients 65 or older saw a much larger relative reduction in the probability of having these events in response to a comprehensive use mandate compared to younger patients. Our findings provide strong evidence that comprehensive use mandates contributed to reductions in prescription opioid misuse and overdose.

One mechanism by which PDMP policies might affect opioid-related hospital events is through reduced prescription opioid use that puts patients at high risk of opioid misuse or overdose. Our analysis of HCCI data pertaining to the same study population indicated that comprehensive use mandates were associated with a 10–11% reduction in the probability of having overlapping opioid prescriptions and a more modest reduction in the probability of having opioid prescriptions from three or more prescribers, two prominent measures of high-risk opioid prescriptions (Appendix Figure 6, SDC1, http://links.lww.com/MLR/B996), thus providing support for this mechanism.

Our findings are also consistent with those of a recently published study that found comprehensive use mandates were associated with a 4.2% reduction in opioid-related inpatient discharges and 17.8% reduction in opioid-related ED discharges among Medicaid patients.³⁵ Our estimated effect on the composite outcome of opioid-related inpatient or ED events was slightly larger (28% overall and 20% and 46% reduction for younger and older patients, respectively). This likely reflects differences in the study populations; while people with private insurance or Medicare Advantage – the focus of our study – had a lower likelihood of opioid misuse and overdose compared to Medicaid enrollees,³⁶ we restricted our population to patients with at least one opioid prescription during the study years (and thus a heightened risk of opioid misuse or overdose) compared to the general Medicaid population in the other study.

Our (and other recent) findings may alleviate concerns regarding potential unintended consequences of PDMP policies, for example, by increasing harmful illicit opioid use among patients whose use of prescription opioids were curtailed or discontinued abruptly. Increased illicit opioid use associated with a higher and rapidly increasing rate of overdose may counteract the (intended) effects of the PDMP policies on adverse events related to prescription opioid misuse and overdose. Although we were not able to separately assess the effects of PDMP policies for prescription opioid- and illicit opioid-related events, the sizable reduction in all opioid-related hospital events suggests that it was unlikely that comprehensive use mandates were associated with a large unintended shift to harmful illicit opioid use in the study population.

Our findings of large relative effects for older patients are consistent with age differences in the evolving epidemic of prescription opioid misuse over 2002–14.²¹ Specifically, the rate of prescription opioid misuse doubled among people 50 or older compared with a decline among individuals 18–24 and stability among those 25–49. In contrast, the rate of illicit opioid use (and illicit opioid-related overdoses³⁷) increased much more rapidly among younger than older adults. While comprehensive use mandates seemed associated with similar or smaller relative reduction in high-risk opioid prescriptions among older versus younger patients (Appendix Figure 6, SDC1, http://links.lww.com/MLR/B996), these policies, by addressing prescription opioid misuse, may have been more effective in reducing opioid-related hospital events among older patients since prescription opioid misuse (rather than illicit opioid use) may have more dominantly accounted for hospital events among older patients.

Our study had several limitations. First, whether a state implemented a PDMP policy and the timing of implementation may be correlated with development of the opioid crisis in the state, leading to potential biases in our estimates. Our event study analysis, however, indicated that implementing and non-implementing states exhibited parallel trends in study outcomes prior to policy implementation. Second, in addition to implementing PDMP policies, states may have concurrently engaged in other actions to publicize and prioritize strategies to address the opioid epidemic. We cannot disentangle the effects of PDMP policies of interest from these other activities. Third, many states implemented multiple PDMP enhancing policies. In particular, six of the seven states with comprehensive use mandates in our analysis had implemented delegate laws before or at the same time as their comprehensive use mandates took effect. The estimated effects pertaining to comprehensive use mandates thus more closely reflect the combined effects of comprehensive use mandates and delegate laws. Meanwhile, 17 of the 23 states with delegate laws did not implement comprehensive use mandates. The estimated effects of delegate laws thus captured their effects independent of the effects of comprehensive use mandates. Fourth, we did not include non-PDMP policies that might bear implications for prescription opioid use. Previous studies did not find state laws governing pain clinics to be associated with reduction in opioid prescribing or opioid overdose death rates.¹⁸ Although several studies had found medical marijuana legalizations (MMLs) to be associated with reductions in population rates of prescription opioid overdose deaths or opioid-related hospitalizations^38–40, these MMLs largely took effect before 2011 or after 2015,⁴¹ and, thus had little overlap with the implementation of PDMP policies examined in this study. Importantly, of the seven states that implemented comprehensive PDMP use mandates in our study, three (KY, OK, and TN) never had MML, one (NV) had MML in 2001, and two (OH, WV) did not have MML effective until 2016 and 2017, respectively; in only one state (CT), MML took effect during our study years (2012), three years before their comprehensive use mandate took effect.⁴² It is thus unlikely that the association we found between comprehensive use mandates and opioid-related hospital events was confounded by the MML. Fifth, although the HCCI data cover approximately one-third of those with private insurance and one-half of those with Medicare Advantage nationwide,⁴³ the generalizability of our findings to the privately insured and Medicare Advantage populations is unknown and may vary across states.

Our analysis of national data of privately insured and Medicare Advantage adults provides evidence that state implementation of comprehensive mandates for prescriber use of PDMP was associated with large relative reductions in hospital events related to opioid dependence, abuse, and overdose. In relative terms, such mandates seemed more beneficial for older adults who used prescription opioids.

Supplementary Material

Supplemental Digital Content 1

• Appendix Table 1. ICD-9 diagnosis codes for adverse opioid-related events

• Appendix Table 2. ICD-9 diagnosis codes for pain, mental health, and substance use conditions

• Appendix Table 3. Summary statistics of outcomes and characteristics by age group

• Appendix Table 4. Estimated reduction in the number of adults (18+) with private insurance or Medicare who experience opioid-related hospital events in a half-year associated with implementation of comprehensive PDMP use mandates

• Appendix Figure 1. Results of event study analysis to assess the parallel trend assumption in the difference-in-differences approach: Patients with any opioid use

• Appendix Figure 2. Results of event study analysis to assess the parallel trend assumption in the difference-in-differences approach: Patients with any long-term opioid use

• Appendix Figures 3A. Changes in the probability of any opioid-related hospital event among patients with any opioid use, January 2011–June 2015

• Appendix Figures 3B. Changes in the probability of any opioid-related hospital event among patients with long-term opioid use, January 2011–June 2015

• Appendix Figures 4A. Changes in the probability of any opioid-related hospital event (primary diagnosis only) among patients with any opioid use, January 2011–December 2015>

• Appendix Figures 4B. Changes in the probability of any opioid-related hospital event (primary diagnosis only) among patients with long-term opioid use, January 2011–December 2015

• Appendix Figures 5A. Changes in the probability of any opioid poisoning-related hospital event among patients with any opioid use, January 2011–December 2015

• Appendix Figures 5B. Changes in the probability of any opioid poisoning-related hospital event among patients with long-term opioid use, January 2011–December 2015

• Appendix Figure 6. Changes in the probability of high-risk opioid prescriptions associated with a state’s implementation of comprehensive PDMP use mandates, January 2011–December 2015