Effects of State Opioid Prescribing Laws on Use of Opioid and Other Pain Treatments Among Commercially Insured U.S. Adults

Abstract

Background:

There is concern that state laws to curb opioid prescribing may adversely affect patients with chronic noncancer pain, but the laws’ effects are unclear because of challenges in disentangling multiple laws implemented around the same time.

Objective:

To study the association between state opioid prescribing cap laws, pill mill laws, and mandatory prescription drug monitoring program query or enrollment laws and trends in opioid and guideline-concordant nonopioid pain treatment among commercially insured adults, including a subgroup with chronic noncancer pain conditions.

Design:

Thirteen treatment states that implemented a single law of interest in a 4-year period and unique groups of control states for each treatment state were identified. Augmented synthetic control analyses were used to estimate the association between each state law and outcomes.

Setting:

United States, 2008 to 2019.

Patients:

7694 514 commercially insured adults aged 18 years or older, including 1 976355 diagnosed with arthritis, low back pain, headache, fibromyalgia, and/or neuropathic pain.

Measurements:

Proportion of patients receiving any opioid prescription or guideline-concordant nonopioid pain treatment per month, and mean days’ supply and morphine milligram equivalents (MME) of prescribed opioids per day, per patient, per month.

Results:

Laws were associated with small-in-magnitude and non-statistically significant changes in outcomes, although CIs around some estimates were wide. For adults overall and those with chronic noncancer pain, the 13 state laws were each associated with a change of less than 1 percentage point in the proportion of patients receiving any opioid prescription and a change of less than 2 percentage points in the proportion receiving any guideline-concordant nonopioid treatment, per month. The laws were associated with a change of less than 1 in days’ supply of opioid prescriptions and a change of less than 4 in average monthly MME per day per patient prescribed opioids.

Limitations:

Results may not be generalizable to non-commercially insured populations and were imprecise for some estimates. Use of claims data precluded assessment of the clinical appropriateness of pain treatments.

Conclusion:

This study did not identify changes in opioid prescribing or nonopioid pain treatment attributable to state laws.

Primary Funding Source:

National Institute on Drug Abuse.

The U.S. opioid crisis has been driven by a 4-fold increase in opioid prescribing between 1999 and 2012 (1). Opioid prescribing rates peaked in 2011 but remain nearly twice as high as 1999 levels (2). Although opioid overdose deaths in the United States are now driven by nonprescription opioids, including heroin and illicitly produced fentanyl (3), a substantial proportion of people begin using prescription opioids before transitioning to other opioids (4–6).

Over the past 2 decades, clinical guidelines for management of acute and chronic noncancer pain have increasingly emphasized use of nonopioid alternatives (7–12). When opioids are deemed clinically necessary, guidelines emphasize prescribing the lowest effective dose for the shortest possible time (7–12). Over the same period, states have passed multiple types of laws designed to curb opioid prescribing. Since 2010, the opioid prescribing laws enacted by states have included 4 main types (13–16): opioid prescribing cap laws limiting the dose and/or duration of opioid prescriptions; pill mill laws regulating pain management clinics to prevent rogue clinics from issuing opioid prescriptions without medical indication; mandatory prescription drug monitoring program (PDMP) query laws requiring prescribers to check the PDMP before prescribing an opioid; and mandatory PDMP enrollment laws requiring prescribers to enroll in their state’s PDMP, thereby giving them access to–but not requiring them to check–the PDMP database. Our study team found that as of 2021, 38 states had opioid prescribing cap laws, 11 had pill mill laws, 33 had mandatory PDMP query laws, and 22 had mandatory PDMP enrollment laws (details on the legal research are provided in the “Data” section later in the article).

State laws designed to govern opioid prescribing have been opposed by some on the grounds that they restrict prescribers’ clinical judgment and impose an administrative burden on health systems, prescribers, and pharmacies (17–23). Pain experts and patient advocates have expressed concern that these laws may lead to rapid tapering, sudden discontinuation, or otherwise restricted access to prescription opioids among patients with chronic noncancer pain without promoting substitution of effective nonopioid alternatives (23, 24). This concern has been raised even in the context of opioid prescribing cap laws, which primarily target acute pain (15), due to concern that they may have spillover effects given the indistinct demarcation between acute versus chronic pain (for example, acute flares of noncancer chronic pain conditions [24, 25]). There is also concern that, by reducing access to prescription opioids, these laws may lead to increased use of the illicit opioids that currently drive the majority of U.S. drug overdose deaths (24, 26, 27). These concerns echo issues raised with regard to the 2016 Centers for Disease Control and Prevention (CDC) guideline, which informed the dose and duration limits codified in state opioid prescribing cap laws (24, 28). In the context of these concerns, in 2019, the U.S. Department of Health and Human Services issued a guide for clinicians on best practices for pain management (29), and the CDC has attempted to clarify (30) and is slated to update its guideline in 2022 (31).

The evidence on state opioid prescribing laws’ effects on opioid prescribing patterns is mixed. Although many prior studies have examined PDMP laws (32, 33), and a growing number have considered pill mill laws (34–39) and prescribing cap laws (40–44), these studies reported mixed findings. Nearly all studies of opioid prescribing laws have focused on the overall population or on prescribing for acute pain (32, 33). One cross-sectional study found no effects of a state mandatory PDMP query law on receipt of opioid prescriptions among a national sample of patients with chronic noncancer pain (45). A key challenge across the existing research is that many states implemented multiple opioid prescribing laws at or around the same time (46). This makes it very difficult, even when controlling for other types of laws in statistical models, to isolate the effects of the laws of interest on outcomes. Our study was designed to overcome this challenge.

Methods

Design

This study was designed to examine state-specific effects of opioid prescribing laws on patterns of opioid prescribing and nonopioid pain treatment in a set of states that implemented only 1 law of interest and no potentially confounding laws in a 4-year period. The same type of state law may have a different effect on outcomes in one state versus another due to differing provisions within the laws or differential implementation. This study is part of a larger project (14) that used legal research on laws’ provisions (described later) and previously published qualitative research characterizing law implementation (47) to help explain potential heterogeneity of laws’ associations with outcomes across states.

We identified “treatment states” that implemented 1 of the 4 types of state opioid prescribing laws of interest in 2010 or later and implemented no other opioid prescribing laws 2 years before or after the law of interest. Treatment states could have other opioid prescribing laws in place so long as those laws were not newly implemented during the study period. For each treatment state, we identified a set of control states that did not have the law of interest and made no changes to any other opioid prescribing laws during the same 4-year period.

Each treatment state and its control group had a distinct 4-year study period. Thirteen treatment states meeting these criteria were identified: 4 states with opioid prescribing cap laws (Delaware, Kentucky, New York, and Ohio), 3 with pill mill laws (Mississippi, Ohio, and Texas), 4 with mandatory PDMP query laws (New York, Oklahoma, Pennsylvania, and Virginia), and 2 with mandatory PDMP enrollment laws (Colorado and Idaho). Two states (New York and Ohio) were included twice because they implemented 2 different laws with sufficient time between laws to meet inclusion criteria. Study periods, control states, and law details are shown in Table 1.

Table 1.

States That Implemented Opioid Prescribing Laws and Their Control States

State Law*	Date Implemented†	Key Provisions	Study Period	Control States	Analytic Sample Size‡
Opioid prescribing cap law
Delaware	4/1/17	Limits initial opioid prescriptions (all schedules) for acute pain to a 7 days’ supply	4/1/15-3/31/19	AL, IA, KS, MT, MS, ND, NM, OR, TN, WY	All adults aged ≥18 y: n = 261 155 Adults with CNCP: n = 76 376
Kentucky	7/1/17	Limits all schedule II opioid prescriptions for acute pain to a 3 days’ supply	7/1/15-6/30/19	AL, IA, KS, MS, MT, ND, NM, OR, WY	All adults aged ≥18 y: n = 274 062 Adults with CNCP: n = 82 053
New York	7/22/16	Limits initial schedule II-IV opioid prescriptions for acute pain to a 7 days’ supply	8/1/14-7/31/18	AL, IA, KS, MS, MT, ND, OR, WY	All adults aged ≥18 y: n = 428 649 Adults with CNCP: n = 112 690
Ohio	8/31/17	Limits initial opioid prescriptions (all schedules) for acute pain to a 7 days’ supply and 30 MME/d and initial opioid prescriptions for chronic noncancer pain to a 7 days’ supply and 120 MME/d	9/1/15-8/31/19	AL, IA, KS, MS, MT, ND, NM, OR, WY	All adults aged ≥18 y: n = 373 092 Adults with CNCP: n = 106 484
Pill mill law
Mississippi	3/1/11	Requires that pain management clinics be owned by a physician with an unrestricted Mississippi license who has not been convicted of a crime related to illegal distribution of controlled substances; that clinic owners meet pain management certification requirements; and that pain clinics undergo annual certification renewal through the State Board of Medical Licensure	3/1/09-2/28/13	AL, AZ, CO, IA, ID, IL, IN, LA, MI, MO, NC, NV, NY, ND, OK, PA, RI, SC, VA, WY	All adults aged ≥18 y: n = 2 128 482 Adults with CNCP: n = 466 011
Ohio	7/1/11	Requires that pain clinic owners supervise all persons who provide chronic noncancer pain treatment at the clinic; that clinic owners meet pain management certification requirements; and that pain clinics undergo annual verification of licensure through the State Medical Board	7/1/09-6/30/13	AL, AZ, CO, ID, IN, IA, IL, LA, MA, MI, MO, NC, NV, NY, ND, OK, PA, RI, SC, VA, WY	All adults aged ≥18 y: n = 2 429 187 Adults with CNCP: n = 531 314
Texas	9/1/10	Requires that pain clinics be owned by a physician with an unrestricted Texas license who has not been convicted of a felony or misdemeanor related to distribution of prescription drugs and that pain clinics undergo biennial certificate renewal by the State Medical Board	9/1/08-8/31/12	AL, AZ, CO, CT, ID, IL, IN, LA, MA, MI, MO, NC, NV, NY, OK, PA, RI, SC, TN, VA, WV, WY	All adults aged ≥18 y: n = 2 998 858 Adults with CNCP: n = 381 050
Mandatory PDMP query law
New York	8/27/13	Prescribers must check the PDMP every time they prescribe an opioid to any patient	9/1/11-8/31/15	AK, AZ, CA, CO, IA, FL, LA, KS, MO, MI, MN, NC, ND, OR, SD, UT, WA, WY	All adults aged ≥18 y: n = 2 012 484 Adults with CNCP: n = 438 801
Oklahoma	11/1/15	Prescribers must check the PDMP for every new opioid prescription and for refills <180 days since the last check	11/1/13-10/31/17	FL, GA, IA, KS, KY, LA, MI, MO, MS, MT, ND, NE, NM, OR, SD, TN, WV, WY	All adults aged ≥18 y: n = 1 360 898 Adults with CNCP: n = 346 195
Pennsylvania	6/30/15	Prescribers must check the PDMP every time they prescribe an opioid to a new patient	7/1/13-6/30/17	FL, GA, IA, KS, KY, LA, MI, MO, MS, MT, ND, NE, NM, OR, SD, TN, WV, WY	All adults aged ≥18 y: n = 1 449 380 Adults with CNCP: n = 359 759
Virginia	7/1/15	Prescribers must check the PDMP for every new opioid prescription	7/1/13-6/30/17	FL, GA, IA, KS, KY, MI, MO, MS, MT, ND, NE, NM, OR, SD, TN, WV, WY	All adults aged ≥18 y: n = 1 336 033 Adults with CNCP: n = 346 059
Mandatory PDMP enrollment law
Colorado	1/1/15	Prescribers must be enrolled in the PDMP to prescribe a controlled substance	1/1/13-12/31/16	AK, AZ, FL, IA, KS, KY, LA, MI, MO, MS, MT, NC, ND, NE, NM, OR, SC, SD, TN, UT, WA, WY	All adults aged ≥18 y: n = 2 502 353 Adults with CNCP: n = 572 550
Idaho	7/1/14	Prescribers must be enrolled in the PDMP to prescribe a controlled substance	7/1/12-6/30/16	AK, CA, AZ, DE, FL, IA, KS, KY, LA, MI, MN, MO, MT, NC, ND, NE, OR, SC, SD, UT, WA, WV, WY	All adults aged ≥18 y: n = 1 912 471 Adults with CNCP: n = 442 996

CNCP = chronic noncancer pain; MME = morphine milligram equivalents; PDMP = prescription drug monitoring program.

^*Treatment states were defined as states that implemented a single law of interest in a 4-year period 2 years before or after the law and implemented no other opioid prescribing laws during that period. Control states did not have the law of interest and did not change any other types of opioid prescribing laws during the study period. Additional details on law provisions are provided in Section A of the Supplement.

^†Defined as the date providers were required to adhere to the law. This date was identified through legal research and confirmed in qualitative interviews with leaders from the 13 treatment states whose professional roles involved overseeing implementation of the law of interest.

^‡Across all 13 treatment states and their control groups, the overall sample included 7 694 514 unique adults aged ≥18 y and 1 976 355 unique adults aged ≥18 y with chronic noncancer pain conditions. Some adults were represented in multiple state-specific analyses because the same states were included in multiple control groups or because 2 states (New York and Ohio) served as treatment states in 2 separate analyses.

Data

We used the IBM MarketScan commercial insurance claims data for 2008 to 2019, which includes inpatient, outpatient, and pharmacy claims, covered by more than 350 different insurers, for health care services received by about 25% of U.S. commercial insurance beneficiaries and their families (48).

The state opioid prescribing law data were compiled by 2 public health attorneys (A.D.M. and L.R.) using systematic searches of the Westlaw legal database. Once identified, laws were reviewed to determine their provisions and implementation dates, defined as the date prescribers were required to adhere to them. For the 13 treatment states, we verified the implementation date in qualitative interviews with state leaders whose professional roles involved overseeing implementation of a state law of interest (47). For all 13 treatment states, interviewees confirmed that full implementation began on the date identified. Additional details are provided in Section A of the Supplement (available at Annals.org).

The legal data included the 4 types of state laws of interest in this study: opioid prescribing cap laws, pill mill laws, mandatory PDMP query laws, and mandatory PDMP enrollment laws. The data also included an earlier generation (largely before 2010) of state laws, including voluntary PDMP laws establishing a state PDMP, pharmacy identification laws requiring identification to pick up an opioid prescription, laws requiring a physical examination before opioids are prescribed, and “doctor-shopping” laws prohibiting people from seeking overlapping opioid prescriptions from multiple providers. Although most states implemented these laws before our study period, a few implemented them in the early 2010s. We accounted for these potentially confounding laws in our study design: If a state implemented 1 of these laws within 2 years of 1 of the 4 laws of interest, it was ineligible for inclusion as a treatment state, and states were ineligible to be control states if they implemented 1 of these laws during the study period.

Sample

For each treatment state and its control group, we identified adults aged 18 years or older who were continuously enrolled in insurance for the entire 4-year period. We identified a subsample of adults meeting the same continuous enrollment criterion who were diagnosed with conditions that commonly lead to chronic noncancer pain (arthritis, low back pain, serious headache, fibromyalgia, and neuropathic pain). Adults were included in this chronic noncancer pain sample if they had 1 inpatient or 2 outpatient diagnosis codes for a given condition in the prelaw period. People meeting this criterion for multiple diagnoses were coded as having each relevant condition. We excluded those diagnosed with cancer, excluding skin cancer, at any point during the study period. Appendix Table 1 (available at Annals.org) lists diagnosis codes.

Measures

State laws were coded as binary indicators that changed from 0 to 1 in the first month that a state law was effective for 15 or more days in that month. Outcome measures were constructed at the state-month level. Opioid medications used primarily to treat pain were identified using the CDC Opioid NDC and Oral MME Conversion File. Opioid-agonist medications used primarily to treat opioid use disorder were excluded (Appendix, available at Annals.org). Opioid prescribing outcomes included measures of the proportion of patients receiving at least 1 opioid prescription per state per month; measures of mean days’ supply of opioid prescriptions and mean morphine milligram equivalents (MME) per day per patient prescribed opioids per state per month; and the proportion of patients prescribed opioids in a state-month who received any opioid prescription with at least 7 days’ supply or had at least 50 MME/d across all opioid prescriptions, accounting for overlapping prescriptions, at any point in the month. The latter 2 measures are indicators of high-risk prescriptions associated with increased risk for addiction and overdose (9).

In the chronic noncancer pain cohort, we also measured the proportion of patients receiving any clinical guideline-concordant nonopioid pain medication or procedure per state per month. The chronic pain clinician author (M.C.B.) identified guideline-concordant nonopioid treatments for arthritis, low back pain, serious headache, fibromyalgia, and neuropathic pain through review of clinical guidelines (49–52) and developed algorithms to code these treatments in claims data (Appendix).

Covariates included mean patient age; mean patient Elixhauser Comorbidity Index; the proportion of patients who were male; the proportion with a mental illness diagnosis; the proportion with a substance use disorder diagnosis; and the proportions of the state population that were Black, were Hispanic, were living below the federal poverty line, were employed, and had no postsecondary degree, per state, per year.

Statistical Analysis

We used an augmented synthetic control approach (53). The synthetic control approach, which has been used in a wide range of policy evaluations (36, 54–58), uses longitudinal data from a weighted combination of control states to create a control group that estimates what would have happened to the outcome of interest in the treatment state in the absence of the policy change. This approach weights the control states so that the magnitude of and time trends in outcomes and covariates during the prelaw period align as closely as possible in the treatment and control states. Covariates are included to capture factors that could drive differential trends in outcomes across treatment and control states over time (59) and to avoid overfitting to the outcome trends. The original synthetic control approach (55), without augmentation, can produce biased results in the absence of excellent pretreatment balance in outcomes in the treatment state and its synthetic control (53). The augmented synthetic control approach overcomes this by combining the synthetic control method with a parametric outcome model (53). The key assumptions underlying this approach are that there are enough prelaw periods to estimate what would have happened to the outcome of interest in the treatment state in the absence of the law and that absent the law, trends in outcomes in the treatment state would have followed the same trajectory as trends in outcomes in the weighted group of control states (the “synthetic control”).

Using the augmented synthetic control approach, our study sought to answer the question, “What is the effect of implementation of a state opioid prescribing law on receipt of opioid prescriptions and nonopioid pain treatments among commercially insured adults in that state, relative to the expected levels of opioid prescribing and nonopioid pain treatment in the same population absent the law?” In our study, the synthetic control approach used the prelaw outcomes and covariates described earlier to construct the weighted control group that best approximated the prelaw outcome and covariate trends in the treatment state. We then augmented the synthetic control approach using a linear regression model of the outcome, controlling for each of the covariates described earlier and including state fixed effects, which account for unobserved state-specific factors, to estimate the change in outcome attributable to the law in the treatment state. We estimated the change in outcomes attributable to the law averaged over the entire 2-year postlaw period as well as monthly effects, which allowed us to examine whether the magnitude of change in outcome attributable to the law varied month by month over the 2-year postlaw period. The change in outcome attributable to the law estimated using the augmented synthetic control method is the mean difference between the change in outcomes before and after the law in the treatment state versus its weighted control group, also known as a difference-in-differences.

We conducted 4 sensitivity analyses, all of which used the same analytic approach as our main analyses. First, we conducted analyses using control state samples with the same underlying opioid prescribing legal environment as the treatment state, except for the law of interest. For example, if a treatment state had a doctor-shopping law in place for the entire study period, control states were limited to states with a doctor-shopping law for the same period. Second, we conducted analyses limiting the sample of patients with chronic noncancer pain to the subgroup who had opioid prescriptions in the prelaw period. Third, we examined state laws’ effects on outcomes among patients diagnosed with each specific chronic noncancer pain condition. Fourth, given research suggesting that state cannabis laws may influence use of prescription opioids (60), we conducted analyses excluding states that enacted medical or recreational cannabis laws during the study period.

Role of the Funding Source

The National Institute on Drug Abuse had no role in the study design, conduct, or analysis.

Results

The state-specific analytic sample sizes and balance in prelaw characteristics in the treatment group, unweighted control states, and weighted synthetic control are shown in Section B of the Supplement. There was no evidence of nonparallel preperiod trends in outcomes in the augmented synthetic control analyses, as shown graphically in Section C of the Supplement. As it is designed to do, the augmented synthetic control approach achieved excellent balance in prelaw outcomes in each treatment state and its synthetic control: Mean outcomes over the 2-year prelaw period were identical across all treatment states and synthetic controls (Appendix Tables 2 to 7, available at Annals.org).

Results of augmented synthetic control analyses showed small-in-magnitude, non-statistically significant changes in outcomes attributable to state opioid prescribing laws in both the overall adult and chronic noncancer pain samples. In the first 2 years of implementation, the 13 state laws were each associated with a change of less than 1 percentage point, with 95% CIs that did not exceed a 2-percentage point increase or decrease, in the monthly proportion of patients receiving any opioid prescription in both the overall adult and chronic noncancer pain samples (Figure 1). Changes ranged from a −0.75 (95% CI, −1.42 to 0.07)–percentage point decrease in the proportion of patients with chronic noncancer pain with an opioid prescription attributable to Idaho’s mandatory PDMP enrollment law to a 0.47 (CI, −0.06 to 0.89)–percentage point increase attributable to Kentucky’s opioid prescribing cap law. The difference-in-differences calculations that produced these estimates of the change in outcomes attributable to the law are shown in Appendix Tables 2 to 7. For example, an estimated 9.18% of patients with chronic noncancer pain in Idaho and its synthetic control received an opioid prescription in each month of the 2 years before implementation of the mandatory PDMP enrollment law. An estimated 9.03% of patients in Idaho and 9.78% of patients in the synthetic control received an opioid prescription, per month, in the first 2 years of law implementation ([9.03–9.18] – [9.78–9.18] = −0.75).

Figure 1.

Changes in the proportion of patients receiving any opioid prescription, per month, attributable to the law in its first 2 years of implementation.

Augmented synthetic control analyses were used to estimate changes in outcomes between the prelaw and postlaw periods in the treatment state and its control group. The results in this figure show the percentage point difference in the proportion of patients with any opioid prescription, per month, in the 2 years before versus after the law implementation date in the treatment state versus its control group. Analyses were adjusted for mean patient age; mean patient Elixhauser Comorbidity Index; the proportion of patients who were male; the proportion with a mental illness diagnosis; the proportion with a substance use disorder diagnosis; and the proportions of the state population that were Black, were Hispanic, were living below the federal poverty line, were employed, and had no postsecondary degree, per state, per year. The chronic noncancer pain sample included patients diagnosed with arthritis, low back pain, headache, fibromyalgia, or neuropathic pain in the prelaw period. Across all 13 treatment states and their control groups, the overall sample included 7 694 514 unique adults aged ≥18 years and 1 976 355 unique adults aged ≥18 years with chronic noncancer pain conditions. State-specific analytic sample sizes are shown in Table 1. PDMP = prescription drug monitoring program.

In both the overall adult and chronic noncancer pain samples, the state opioid prescribing laws were each associated with a change of less than 1 day, with upper and lower 95% CI bounds not exceeding a 2-day increase or decrease, in days’ supply of opioid prescriptions per month among patients prescribed opioids (Table 2). Change in days’ supply attributable to state laws ranged from a −0.69 (CI, −1.75 to 0.37)-day reduction attributable to Mississippi’s pill mill law in the chronic noncancer pain sample to a 0.68 (CI, −0.31 to 1.66)-day increase attributable to Idaho’s PDMP enrollment law in the overall adult sample. Each of the 13 state opioid prescribing laws was associated with a change of less than 4 in the average monthly MME per day per patient prescribed opioids in the overall adult and chronic noncancer pain samples, ranging from a −3.73 (CI, −8.89 to 1.43)-MME/d decrease attributable to Mississippi’s pill mill law among patients in the overall adult sample to a 3.05 (CI, −2.50 to 8.60)-MME/d increase attributable to Delaware’s opioid prescribing cap law in the chronic noncancer pain sample; across all estimates, 95% CI bounds did not exceed a 9-MME/d or a 17% relative increase or decrease. State laws were associated with non-statistically significant changes of less than 2 percentage points in the monthly proportion of patients with chronic noncancer pain receiving any guideline-concordant nonopioid pain medication or procedure, with confidence bounds not exceeding a 5-percentage point change (Figure 2).

Table 2.

Change in Monthly Volume and Dose of Opioid Prescriptions, Among Patients Receiving ≥1 Prescription, Attributable to the Law During Its First 2 Years of Implementation^*

State Laws	Opioid Prescription Volume		Opioid Prescription Dose
	Change in Days’ Supply per Patient Prescribed Opioids, per Month, Attributable to Law (95% CI)	Change in Proportion of Patients Prescribed Opioids Who Received Any Opioid Prescription With ≥7 Days’ Supply, per Month, Attributable to Law (95% CI), percentage points	Change in Dose per Patient Prescribed Opioids, per Month, Attributable to Law (95% CI), MME/d	Change in Proportion of Patients Prescribed Opioids With ≥50 MME/d, per Month, Attributable >to Law (95% CI), percentage points
Overall sample
Prescribing cap law
Delaware	0.32 (−1.27 to 1.91)	0.46 (−5.53 to 6.45)	0.81 (−2.83 to 4.45)	−1.21 (−4.55 to 2.14)
Kentucky	0.14 (−0.62 to 0.91)	0.41 (−1.89 to 2.71)	1.11 (−0.45 to 2.67)	−0.05 (−1.96 to 1.85)
New York	−0.21 (−0.75 to 0.33)	−2.57 (−4.98 to 0.17)	−0.40 (−2.64 to 1.83)	−0.53 (−2.32 to 1.26)
Ohio	0.32 (−1.27 to 1.91)	0.46 (−5.53 to 6.45)	0.81 (−2.83 to 4.45)	−1.21 (−4.55 to 2.14)
Pill mill law
Mississippi	−0.39 (−1.16 to 0.39)	−0.52 (−3.62 to 2.57)	−3.73 (−8.89 to 1.43)	−2.80 (−6.52 to 0.92)
Ohio	−0.09 (−0.45 to 0.26)	−0.35 (−1.50 to 0.79)	−0.55 (−1.77 to 0.66)	−0.98 (−2.32 to 0.36)
Texas	−0.12 (−0.60 to 0.36)	0.28 (−1.20 to 1.75)	−1.97 (−5.70 to 1.75)	−3.55 (−6.12 to 0.97)
PDMP query law
New York	−0.002 (−0.52 to 0.51)	−1.25 (−2.95 to 0.45)	0.80 (−1.04 to 2.64)	1.34 (−1.17 to 2.78)
Oklahoma	0.41 (−0.41 to 1.23)	1.40 (−1.57 to 4.37)	−2.17 (−6.11 to 1.77)	−1.46 (−4.29 to 1.37)
Pennsylvania	0.38 (−0.47 to 1.23)	1.03 (−1.79 to 3.84)	1.01 (−1.51 to 3.53)	0.88 (−0.99 to 2.75)
Virginia	−0.39 (−1.34 to 0.57)	−1.41 (−5.31 to 2.49)	−0.56 (−2.70 to 1.58)	−0.93 (−3.00 to 1.14)
PDMP enrollment law
Colorado	0.01 (−0.57 to 0.59)	−0.79 (−3.25 to 1.67)	−1.81 (−5.80 to 2.17)	−1.42 (−4.08 to 1.25)
Idaho	0.68 (−0.31 to 1.66)	1.56 (−1.49 to 4.61)	0.93 (−1.03 to 2.89)	1.04 (−1.06 to 3.68)
Chronic noncancer pain sample
Prescribing cap law
Delaware	0.08 (−1.53 to 1.69)	0.35 (−5.38 to 6.08)	3.05 (−2.50 to 8.60)	2.50 (−1.58 to 6.59)
Kentucky	0.07 (−0.93 to 1.06)	0.67 (−2.29 to 3.62)	0.96 (−1.34 to 3.25)	0.67 (−2.12 to 3.46)
New York	−0.003 (−1.06 to 1.05)	−2.63 (−6.35 to 1.10)	−1.43 (−6.04 to 3.18)	−1.28 (−3.76 to 1.20)
Ohio	−0.06 (−0.92 to 0.81)	−3.22 (−7.56 to 1.12)	−0.47 (−2.45 to 1.51)	−2.25 (−4.98 to 0.48)
Pill mill law
Mississippi	−0.69 (−1.75 to 0.37)	−0.18 (−3.31 to 2.95)	−3.47 (−8.24 to 1.29)	−4.18 (−8.26 to 0.11)
Ohio	−0.06 (−0.55 to 0.43)	−0.57 (−2.06 to 0.91)	−1.11 (−2.58 to 0.35)	−1.33 (−2.97 to 0.30)
Texas	0.15 (−0.52 to 0.83)	1.12 (−1.57 to 3.82)	−0.06 (−3.14 to 3.02)	−2.78 (−6.25 to 0.68)
PDMP query law
New York	−0.42 (−1.36 to 0.51)	−1.07 (−3.70 to 1.55)	2.04 (−0.76 to 4.84)	2.19 (−1.21 to 5.59)
Oklahoma	0.10 (−0.80 to 1.00)	0.07 (−2.54 to 2.67)	−1.49 (−5.72 to 2.74)	−1.13 (−3.62 to 1.36)
Pennsylvania	−0.07 (−0.57 to 0.43)	−0.62 (−3.11 to 1.88)	1.24 (−2.61 to 5.09)	1.39 (−1.78 to 4.57)
Virginia	−0.28 (−1.39 to 0.82)	−1.58 (−6.27 to 3.10)	−1.10 (−4.00 to 1.80)	−1.03 (−3.92 to 1.87)
PDMP enrollment law
Colorado	0.09 (−0.71 to 0.88)	−0.82 (−3.66 to 2.01)	−1.93 (−6.48 to 2.63)	−1.44 (−4.24 to 1.36)
Idaho	0.59 (−0.82 to 1.99)	1.94 (−2.79 to 6.67)	1.37 (−1.65 to 4.39)	1.28 (−2.07 to 4.64)

MME = morphine milligram equivalents; PDMP = prescription drug monitoring program.

^*Augmented synthetic control analyses were used to estimate changes in outcomes between the prelaw and postlaw periods in the treatment state and its control group. The results in this table show the difference in the days’ supply and MME of opioid prescriptions per day, per patient prescribed opioids, per month; the percentage point difference in the proportion of patients prescribed opioids with any prescription with ≥7 days’ supply; and the percentage point difference in the proportion of patients prescribed opioids with ≥50 MME/d, per month, in the 2 years before versus after the law implementation date in the treatment versus control states. The analyses were adjusted for mean patient age; mean patient Elixhauser Comorbidity Index; the proportion of patients who were male; the proportion with a mental illness diagnosis; the proportion with a substance use disorder diagnosis; and the proportions of the state population that were Black, were Hispanic, were living below the federal poverty line, were employed, and had no postsecondary degree, per state, per year. Across all 13 treatment states and their control groups, the overall sample included 7 694 514 unique adults aged ≥18 y and 1 976 355 unique adults aged ≥18 y with chronic noncancer pain conditions. State-specific analytic sample sizes are shown in Table 1.

Figure 2.

Changes in the proportion of patients with chronic noncancer pain receiving any clinical guideline-concordant nonopioid pain medication or procedure, per month, attributable to the law in its first 2 years of implementation.

Augmented synthetic control analyses were used to estimate changes in outcomes between the prelaw and postlaw periods in the treatment state and its control group. The results in this figure show the percentage point difference in the proportion of patients with any clinical guideline–concordant nonopioid pain medication or procedure, per month, in the 2 years before versus after the law implementation date in the treatment state versus its control group. Analyses were adjusted for mean patient age; mean patient Elixhauser Comorbidity Index; the proportion of patients who were male; the proportion with a mental illness diagnosis; the proportion with a substance use disorder diagnosis; and the proportions of the state population that were Black, were Hispanic, were living below the federal poverty line, were employed, and had no postsecondary degree, per state, per year. The chronic noncancer pain sample included patients diagnosed with arthritis, low back pain, headache, fibromyalgia, or neuropathic pain in the prelaw period. Across all 13 treatment states and their control groups, the overall sample included 7 694 514 unique adults aged ≥18 years and 1 976 355 unique adults aged ≥18 years with chronic noncancer pain conditions. State-specific analytic sample sizes are shown in Table 1. PDMP = prescription drug monitoring program.

The magnitude of the change in outcomes attributable to the laws remained consistent in each of the 24 months in the postlaw period (Section D of the Supplement), with the possible exceptions of pill mill laws in Mississippi and Texas. In Mississippi, analyses suggested possible decreases in MME per day per patient prescribed opioids per month attributable to the law beginning in the second year of implementation. In Texas, analyses suggested a decrease in the proportion of patients prescribed at least 50 MME of opioids per day beginning in the third month of the 2-year postlaw period and a gradual upward trend over the course of the postlaw period in the monthly proportion of patients receiving any guideline-concordant procedure attributable to the law. Results of all sensitivity analyses were consistent with the main findings (Section E of the Supplement).

Discussion

We did not find an association between state opioid prescribing laws and receipt of opioid prescriptions or guideline-concordant nonopioid pain treatments among commercially insured adults. Across the 13 states that implemented laws, the change in treatment attributable to the law was consistently small in magnitude and not statistically significant. The magnitude of effects and the breadth of CIs varied across outcomes. In both the overall adult and chronic noncancer pain samples, each of the 13 state laws studied was associated with a change of less than 1 percentage point in the proportion of patients receiving any opioid prescription, with 95% CIs that did not exceed a 2-percentage point increase or decrease. Confidence intervals were widest for estimates of changes in MME per day attributable to the laws, encompassing values that could include clinically meaningful shifts. Each state opioid prescribing law was associated with a change of less than 4 in average monthly MME per day per patient prescribed opioids, with CI bounds across state-specific estimates ranging from a decrease of −8.89 MME/d to an increase of 8.60 MME/d attributable to an opioid prescribing law. Although our study was designed to help explain potential heterogeneity in laws’ effects across states by triangulating state-specific augmented synthetic control analyses with information about laws’ provisions and implementation (14), results indicated minimal or no heterogeneity in the degree to which opioid prescribing laws influenced outcomes across states. The findings suggest that the decreasing volume of opioid prescribing in the United States may be driven more by shifting clinical guidelines, professional norms, or other factors than by these laws.

One possible explanation for these findings is sub-optimal implementation. Although our qualitative study identified no implementation delays, state leaders noted implementation challenges that may prevent these laws from having their intended effects, including lack of health information technology capacity (47). Our study examined laws’ effects at the state level, which may mask health system-level effects: A health system electronic medical record flagging prescriptions that did not align with New Jersey’s opioid prescribing cap law was associated with reduced prescription opioid dose and tablet quantity (42). Laws’ provisions could also explain the lack of observed effects. Like most state opioid prescribing cap laws (15), the 4 state laws limiting the dose or duration of opioid prescriptions included in our study had professional judgment exemptions allowing clinicians to override the law’s prescribing limit. None of the pill mill laws included in our study had provisions imposing criminal penalties for failure to adhere to the law, and none of the PDMP laws had provisions allowing law enforcement access to PDMP data without a warrant, subpoena, or active investigation; prior research suggests that these provisions may be associated with reduced opioid prescribing (35, 61,62).

The prior literature evaluating state opioid prescribing laws used a range of study designs and data sources and showed mixed results (32–44, 63, 64). This prior research predominantly used statistical methods to control for multiple state opioid prescribing laws implemented at or around the same time within states. In contrast, our study used a design-based approach to isolate the effects of state opioid prescribing laws on outcomes. Our study is unique in examining these laws’ effects on opioid and nonopioid pain treatment in longitudinal cohorts of patients with chronic noncancer pain.

This study’s commercially insured sample may not be generalizable to other groups. Although our design minimized the potential for confounding from other state opioid prescribing laws, there could be unmeasured confounding from other state or health system initiatives to curb opioid prescribing during the study period, though to bias results, these initiatives would have to systematically differ over time across treatment and control states (59). As in all state policy evaluations, statistical power is limited by the finite number of states; however, in our case, concern about lack of power is lessened by the consistently small-in-magnitude effect estimates and narrow confidence bounds seen in most analyses. Our analysis did not assess clinical appropriateness of pain treatment or patient-reported outcomes relevant to pain management. We were unable to measure the effects of the laws on opioid overdose because all but 1 of the treatment states (Ohio) changed laws related to naloxone access during their study period (65), precluding the ability to attribute changes in overdose outcomes to the opioid prescribing laws of interest. In a separate study, we found no effects of Ohio’s pill mill law on overdose death from any type of opioid (36). Our study was designed to isolate the effects of a single type of state opioid prescribing law on treatment patterns. The combined effect of multiple laws should be evaluated in future research.

Our study did not identify substantial changes in patterns of opioid prescribing or nonopioid pain treatment attributable to state laws among commercially insured patients. Declining trends in opioid prescribing may be driven more by changing clinical guidelines, professional norms, or other factors than by the direct effects of state laws.

Appendix: Treatment and Diagnosis Lists and Codes

List of Clinical Guideline–Concordant Pain Medications and Procedures

Lists of the names and National Drug Code (NDC) or procedure codes for the clinical guideline-concordant nonopioid pain medications and procedures for treatment of pain related to arthritis, low back pain, headache, fibromyalgia, or neuropathic pain are available at https://github.com/ sbandar2/CNCP_Trt_Codes. This website includes spreadsheets that list the treatment, indicates the chronic noncancer pain conditions for which guidelines recommend that treatment, and includes the NDC or Current Procedural Terminology (CPT) codes used to identify the treatment in insurance claims data. The website includes the following 3 files:

“Medications”:

A list of clinical guideline–concordant nonopioid pain medications and associated NDC codes

“Procedures”:

A list of clinical guideline–concordant procedures and associated CPT codes.

“Chronic Pain ICD codes”:

This file includes the same information as in Appendix Table 1: the International Classification of Diseases (ICD) codes used to identify arthritis, low back pain, headache, fibromyalgia, and neuropathic pain in this study.

Mental Illness and Substance Use Diagnosis

Mental illness:

Two outpatient diagnoses or 1 inpatient diagnosis of ICD-9-CM codes 295 to 302 and 306 to 314 and ICD-10-CM codes F20 to F69, F84, and F90 to F99.

Substance use:

Two outpatient diagnoses or 1 inpatient diagnosis of ICD-9-CM codes 291, 292, 303, 304, and 305 (excluding 305.1 [tobacco use disorder] and 305.8 [antidepressant abuse]) and ICD-10-CM codes F10 to F19 (excluding F17.2x).

Opioid Agonist Medications Used Primarily to Treat Opioid Use Disorder (Excluded in Our Study)

Bunavail, buprenorphine-naloxone, Probuphine, sub-oxone, Subutex, Zubsolv.

Methadone was excluded if prescribed to be taken once per day. If it was prescribed to be taken twice a day or more often, it was included because this indicates likely use for pain rather than opioid use disorder.

Cancer Diagnoses (Excluded in Our Study)

We excluded people diagnosed with cancer other than skin cancer. ICD-9 and ICD-10 cancer diagnosis codes were identified via the National Cancer Institute Surveillance, Epidemiology, and End Results Program Casefinding List (https://seer.cancer.gov/tools/casefinding).

Appendix Table 1.

Chronic Noncancer Pain Diagnosis Codes^†

Diagnosis	ICD-9 Codes	ICD-10 Codes
Low back pain	721.3, 721.42, 722.52, 722.73, 724.02, 724.03, 724.2, 724.3, 724.4, 724.5	M43.06, M43.07, M43.16, M43.17, M47.16, M47.26, M47.27, M47.816, M47.817, M47.896, M47.897, M48.06*, M48.07, M51.06, M51.16, M51.17, M51.26, M51.27, M51.36, M51.37, M54.16, M54.17, M54.3*, M54.4*, M54.5, M54.89, M54.9, S39.012$, S39.023$, S39.092$
Headache	307.81, 339.*, 346.*, 349, 723.8, 784	G43.*, G44.*, G97.1, M54.81, R51
Arthritis	714.*, 715.*	M05-M06*, M15-M19*
Neuropathic pain	350.*, 353.*, 354.*, 355.*, 356.*, 53.12, 53.13, 337.2*, 250.6*	G50.*, G51.*, G52.*, G53.*, G54.0-G54.6, G54.8-G59.*, G60.*, B02.22, B02.23, B02.29, G90.5x, E13.4*
Fibromyalgia	729.1	M79.7

ICD = International Classification of Diseases.

^†Asterisks are wild cards for numbers and blanks; dollar signs are wild cards for letters and blanks.

Appendix Table 2.

Change in Proportion of Patients Receiving Any Opioid Prescription, per Month, Attributable to the Law During Its First 2 Years of Implementation^*

State	Adults Aged ≥18 y					Adults Aged ≥18 y With Chronic Noncancer Pain
	Mean Proportion Receiving Any Opioid Prescription in Each Month, %		Difference, percentage points	Difference-in-Differences Estimate of Change Attributable to Law, percentage points	Mean Proportion Receiving Any Opioid Prescription in Each Month, %		Difference, percentage points	Difference-in-Differences Estimate of Change Attributable to Law, percentage points
	In the 2 Years Before Law	In the 2 Years After Law			In the 2 Years Before Law	In the 2 Years After Law
Delaware	3.08	2.53	−0.55	−0.19 (−0.50 to 0.12)	6.54	5.36	−1.18	−0.02 (−0.58 to 0.54)
Synthetic control	3.08	2.72	−0.36		6.54	5.38	−1.16
Kentucky	4.23	3.98	−0.25	0.19 (−0.01 to 0.38)	9.91	9.10	−0.81	0.47 (−0.06 to 0.89)
Synthetic control	4.23	3.79	−0.44		9.91	8.63	−1.28
New York	1.27	1.12	−0.15	0.05 (−0.19 to 0.29)	4.00	3.40	−0.60	0.17 (−0.30 to 0.64)
Synthetic control	1.27	1.07	−0.20		4.00	3.23	−0.77
Ohio	3.18	2.58	−0.60	−0.21 (−0.56 to 0.14)	7.48	5.84	−1.64	−0.51 (−1.25 to 0.22)
Synthetic control	3.18	2.79	−0.39		7.48	6.35	−1.13
Mississippi	3.77	4.05	0.28	−0.19 (−0.65 to 0.27)	9.37	9.33	−0.04	−0.17 (−0.95 to 0.62)
Synthetic control	3.77	4.24	0.47		9.37	9.50	0.13
Ohio	3.61	3.73	0.12	0.07 (−0.13 to 0.27)	9.09	8.71	−0.38	−0.35 (−0.73 to 0.03)
Synthetic control	3.61	3.66	0.05		9.09	9.06	−0.03
Texas	3.15	3.43	0.28	−0.05 (−0.20 to 0.11)	8.99	8.67	−0.32	−0.28 (−0.66 to 0.10)
Synthetic control	3.15	3.46	0.31		8.99	8.95	−0.04
New York	2.31	2.14	−0.17	−0.28 (−0.74 to 0.18)	5.14	4.48	−0.66	−0.57 (−1.55 to 0.42)
Synthetic control	2.31	2.42	0.11		5.14	5.05	−0.09
Oklahoma	4.99	4.93	−0.06	−0.02 (−0.27 to 0.23)	12.19	11.41	−0.78	−0.23 (−0.87 to 0.42)
Synthetic control	4.99	4.95	−0.04		12.19	11.64	−0.55
Pennsylvania	2.40	2.30	−0.10	−0.05 (−0.24 to 0.13)	6.14	5.59	−0.55	0.02 (−0.29 to 0.33)
Synthetic control	2.40	2.35	−0.05		6.14	5.57	−0.57
Virginia	3.17	3.06	−0.11	−0.09 (−0.31 to 0.13)	7.86	7.11	−0.75	−0.36 (−1.00 to 0.27)
Synthetic control	3.17	3.15	−0.02		7.86	7.47	−0.39
Colorado	2.89	2.78	−0.11	0.01 (−0.12 to 0.13)	7.16	6.46	−0.70	−0.24 (−0.60 to 0.13)
Synthetic control	2.89	2.77	−0.12		7.16	6.70	−0.46
Idaho	2.97	3.01	0.04	−0.05 (−0.20 to 0.10)	9.18	9.03	−0.15	−0.75 (−1.42 to 0.07)
Synthetic control	2.97	3.06	0.09		9.18	9.78	0.60

^*Augmented synthetic control analyses were used to estimate changes in outcomes between the prelaw and postlaw periods in the treatment state and its control group. This approach is designed to minimize differences in outcomes averaged over the prelaw period in the treatment state versus its synthetic control. The results in this table show the percentage point difference in the proportion of patients with any opioid prescription, per month, in the 2 years before versus after the law implementation date in the treatment state versus its control group. Analyses were adjusted for mean patient age; mean patient Elixhauser Comorbidity Index; the proportion of patients who were male; the proportion with a mental illness diagnosis; the proportion with a substance use disorder diagnosis; and the proportions of the state population that were Black, were Hispanic, were living below the federal poverty line, were employed, and had no postsecondary degree, per state, per year. The chronic noncancer pain sample included patients diagnosed with arthritis, low back pain, headache, fibromyalgia, or neuropathic pain in the prelaw period. Across all 13 treatment states and their control groups, the overall sample included 7 694 514 unique adults aged ≥18 y and 1 976 355 unique adults aged ≥18 y with chronic noncancer pain conditions. State-specific analytic sample sizes are shown in Table 1 in the main text.

Appendix Table 3.

Change in Days’ Supply of Prescribed Opioids per Patient Prescribed Opioids, per Month, Attributable to the Law During Its First 2 Years of Implementation^*

State	Adults Aged ≥18 y					Adults Aged ≥18 y With Chronic Noncancer Pain
	Mean Monthly Days’ Supply of Prescribed Opioids Among Patients With ≥1 Prescription		Difference	Difference-in-Differences Estimate of Change Attributable to Law	Mean Monthly Days’ Supply of Prescribed Opioids Among Patients With ≥1 Prescription		Difference	Difference-in-Differences Estimate of Change Attributable to Law
	In the 2 Years Before Law	In the 2 Years After Law			In the 2 Years Before Law	In the 2 Years After Law
Delaware	17.78	18.39	0.61	0.32 (−1.27 to 1.91)	21.19	22.12	0.93	0.08 (−1.53 to 1.69)
Synthetic control	17.78	18.07	0.29		21.19	22.04	0.85
Kentucky	20.24	20.41	0.17	0.14 (−0.62 to 0.91)	23.72	24.29	0.57	0.07 (−0.93 to 1.06)
Synthetic control	20.24	20.27	0.03		23.72	24.22	0.50
New York	17.04	17.57	0.53	−0.21 (−0.75 to 0.33)	19.85	21.12	1.27	−0.003 (−1.06 to 1.05)
Synthetic control	17.04	17.78	0.74		19.85	21.12	1.27
Ohio	17.78	17.78	0.00	0.32 (−1.27 to 1.91)	21.36	21.82	0.46	−0.06 (−0.92 to 0.81)
Synthetic control	18.39	18.07	−0.32		21.36	21.88	0.52
Mississippi	13.26	14.66	1.40	−0.39 (−1.16 to 0.39)	17.33	19.38	2.05	−0.69 (−1.75 to 0.37)
Synthetic control	13.26	15.05	1.79		17.33	20.07	2.74
Ohio	17.43	18.75	1.32	−0.09 (−0.45 to 0.26)	21.01	23.12	2.11	−0.06 (−0.55 to 0.43)
Synthetic control	17.43	18.84	1.41		21.01	23.18	2.17
Texas	14.00	15.33	1.33	−0.12 (−0.60 to 0.36)	19.58	21.59	2.01	0.15 (−0.52 to 0.83)
Synthetic control	14.00	15.45	1.45		19.58	21.44	1.86
New York	17.90	18.82	0.92	−0.002 (−0.52 to 0.51)	20.69	22.04	1.35	−0.42 (−1.36 to 0.51)
Synthetic control	17.90	18.82	0.92		20.69	22.46	1.77
Oklahoma	19.29	20.50	1.21	0.41 (−0.41 to 1.23)	22.65	24.32	1.67	0.10 (−0.80 to 1.00)
Synthetic control	19.29	20.09	0.80		22.65	24.22	1.57
Pennsylvania	17.23	18.17	0.94	0.38 (−0.47 to 1.23)	21.26	22.69	1.43	−0.07 (−0.57 to 0.43)
Synthetic control	17.23	17.79	0.56		21.26	22.76	1.50
Virginia	15.06	15.58	0.52	−0.39 (−1.34 to 0.57)	18.73	20.09	1.36	−0.28 (−1.39 to 0.82)
Synthetic control	15.06	15.97	0.91		18.73	20.37	1.64
Colorado	14.54	15.42	0.88	0.01 (−0.57 to 0.59)	18.43	20.18	1.75	0.09 (−0.71 to 0.88)
Synthetic control	14.54	15.41	0.87		18.43	20.09	1.66
Idaho	16.22	17.46	1.24	0.68 (−0.31 to 1.66)	21.30	23.13	1.83	0.59 (−0.82 to 1.99)
Synthetic control	16.22	16.78	0.56		21.30	22.54	1.24

^*Augmented synthetic control analyses were used to estimate changes in outcomes between the prelaw and postlaw periods in the treatment state and its control group. This approach is designed to minimize differences in outcomes averaged over the prelaw period in the treatment state versus its synthetic control. The results in this table show the difference in the days’ supply per patient prescribed opioids per month in the 2 years before versus after the law implementation date in the treatment versus control states. Analyses were adjusted for mean patient age; mean patient Elixhauser Comorbidity Index; the proportion of patients who were male; the proportion with a mental illness diagnosis; the proportion with a substance use disorder diagnosis; and the proportions of the state population that were Black, were Hispanic, were living below the federal poverty line, were employed, and had no postsecondary degree, per state, per year. Across all 13 treatment states and their control groups, the overall sample included 7 694 514 unique adults aged ≥18 y and 1 976 355 unique adults aged ≥18 y with chronic noncancer pain conditions. State-specific analytic sample sizes are shown in Table 1 in the main text.

Appendix Table 4.

Change in Proportion of Patients Prescribed Opioids Who Received Any Opioid Prescription With ≥7 Days’ Supply, per Month, Attributable to the Law During Its First 2 Years of Implementation^*

State	Adults Aged ≥18 y					Adults Aged ≥18 y With Chronic Noncancer Pain
	Mean Proportion of Patients Prescribed Opioids Receiving Any Prescription With ≥7 Days’ Supply in Each Month, %		Difference, percentage points	Difference-in-Differences Estimate of Change Attributable to Law, percentage points	Mean Proportion of Patients Prescribed Opioids Receiving Any Prescription With ≥7 Days’ Supply in Each Month, %		Difference, percentage points	Difference-in-Differences Estimate of Change Attributable to Law, percentage points
	In the 2 Years Before Law	In the 2 Years After Law			In the 2 Years Before Law	In the 2 Years After Law
Delaware	57.78	57.24	−0.54	0.46 (−5.53 to 6.45)	70.00	70.91	0.91	0.35 (−5.38 to 6.08)
Synthetic control	57.78	56.78	−1.00		70.00	70.56	0.56
Kentucky	65.32	64.23	−1.09	0.41 (−1.89 to 2.71)	77.32	77.73	0.41	0.67 (−2.29 to 3.62)
Synthetic control	65.32	63.82	−1.50		77.32	77.06	−0.26
New York	57.75	57.15	−0.60	−2.57 (−4.98 to 0.17)	68.11	69.52	1.41	−2.63 (−6.35 to 1.10)
Synthetic control	57.75	59.72	1.97		68.11	72.15	4.04
Ohio	57.78	57.24	−0.54	0.46 (−5.53 to 6.45)	70.35	66.60	−3.75	−3.22 (−7.56 to 1.12)
Synthetic control	57.78	56.78	−1.00		70.35	69.82	−0.53
Mississippi	41.85	47.94	6.09	−0.52 (−3.62 to 2.57)	56.88	64.24	7.36	−0.18 (−3.31 to 2.95)
Synthetic control	41.85	48.46	6.61		56.88	64.42	7.54
Ohio	54.53	59.28	4.75	−0.35 (−1.50 to 0.79)	67.31	73.45	6.14	−0.57 (−2.06 to 0.91)
Synthetic control	54.53	59.63	5.10		67.31	74.02	6.71
Texas	43.19	48.21	5.02	0.28 (−1.20 to 1.75)	62.12	68.04	5.92	1.12 (−1.57 to 3.82)
Synthetic control	43.19	47.93	4.74		62.12	66.92	4.80
New York	60.06	62.83	2.77	−1.25 (−2.95 to 0.45)	69.31	73.65	4.34	−1.07 (−3.70 to 1.55)
Synthetic control	60.06	64.08	4.02		69.31	74.72	5.41
Oklahoma	63.70	68.56	4.86	1.40 (−1.57 to 4.37)	74.93	80.57	5.64	0.07 (−2.54 to 2.67)
Synthetic control	63.70	67.16	3.46		74.93	80.50	5.57
Pennsylvania	55.48	59.31	3.83	1.03 (−1.79 to 3.84)	69.69	75.01	5.32	−0.62 (−3.11 to 1.88)
Synthetic control	55.48	58.28	2.80		69.69	75.63	5.94
Virginia	48.17	49.92	1.75	−1.41 (−5.31 to 2.49)	61.62	66.16	4.54	−1.58 (−6.27 to 3.10)
Synthetic control	48.17	51.33	3.16		61.62	67.74	6.12
Colorado	44.61	48.25	3.64	−0.79 (−3.25 to 1.67)	59.22	65.23	6.01	−0.82 (−3.66 to 2.01)
Synthetic control	44.61	49.04	4.43		59.22	66.05	6.83
Idaho	52.54	58.26	5.72	1.56 (−1.49 to 4.61)	69.25	76.42	7.17	1.94 (−2.79 to 6.67)
Synthetic control	52.54	56.70	4.16		69.25	74.48	5.23

^*Augmented synthetic control analyses were used to estimate changes in outcomes between the prelaw and postlaw periods in the treatment state and its control group. This approach is designed to minimize differences in outcomes averaged over the prelaw period in the treatment state versus its synthetic control. The results in this table show the percentage point difference in the proportion of patients prescribed opioids with any opioid prescription with ≥7 days’ supply, per month, in the 2 years before versus after the law implementation date in the treatment versus control states. Analyses were adjusted for mean patient age; mean patient Elixhauser Comorbidity Index; the proportion of patients who were male; the proportion with a mental illness diagnosis; the proportion with a substance use disorder diagnosis; and the proportions of the state population that were Black, were Hispanic, were living below the federal poverty line, were employed, and had no postsecondary degree, per state, per year. Across all 13 treatment states and their control groups, the overall sample included 7 694 514 unique adults aged ≥18 y and 1 976 355 unique adults aged ≥18 y with chronic noncancer pain conditions. State-specific analytic sample sizes are shown in Table 1 in the main text.

Appendix Table 5.

Change in MME per Day per Patient Prescribed Opioids, per Month, Attributable to the Law During Its First 2 Years of Implementation^*

State	Adults Aged ≥18 y					Adults Aged ≥18 y With Chronic Noncancer Pain
	Mean Monthly Dose Among Patients With ≥1 Opioid Prescription, MME/d		Difference, MME/d	Difference-in-Differences Estimate of Change Attributable to Law, MME/d	Mean Monthly Dose Among Patients With ≥1 Opioid Prescription, MME/d		Difference, MME/d	Difference-in-Differences Estimate of Change Attributable to Law, MME/d
	In the 2 Years Before Law	In the 2 Years After Law			In the 2 Years Before Law	In the 2 Years After Law
Delaware	63.85	61.19	−2.66	0.81 (−2.83 to 4.45)	67.46	66.02	−1.44	3.05 (−2.50 to 8.60)
Synthetic control	63.85	60.38	−3.47		67.46	62.97	−4.49
Kentucky	42.76	40.04	−2.72	1.11 (−0.45 to 2.67)	43.46	40.42	−3.04	0.96 (−1.34 to 3.25)
Synthetic control	42.76	38.93	−3.83		43.46	39.46	−4.00
New York	55.53	53.09	−2.44	−0.40 (−2.64 to 1.83)	60.28	57.43	−2.85	−1.43 (−6.04 to 3.18)
Synthetic control	55.53	53.49	−2.04		60.28	58.86	−1.42
Ohio	63.85	61.19	−2.66	0.81 (−2.83 to 4.45)	43.36	37.99	−5.37	−0.47 (−2.45 to 1.51)
Synthetic control	63.85	60.38	−3.47		43.36	38.46	−4.90
Mississippi	52.06	40.79	−11.27	−3.73 (−8.89 to 1.43)	52.68	43.70	−8.98	−3.47 (−8.24 to 1.29)
Synthetic control	52.06	44.52	−7.54		52.68	47.17	−5.51
Ohio	42.39	38.04	−4.35	−0.55 (−1.77 to 0.66)	43.00	39.25	−3.75	−1.11 (−2.58 to 0.35)
Synthetic control	42.39	38.59	−3.80		43.00	40.36	−2.64
Texas	50.28	43.08	−7.20	−1.97 (−5.70 to 1.75)	53.83	51.33	−2.50	−0.06 (−3.14 to 3.02)
Synthetic control	50.28	45.05	−5.23		53.83	51.39	−2.44
New York	60.05	60.42	0.37	0.80 (−1.04 to 2.64)	47.39	49.15	1.76	2.04 (−0.76 to 4.84)
Synthetic control	60.05	59.62	−0.43		47.39	47.11	−0.28
Oklahoma	54.25	50.33	−3.92	−2.17 (−6.11 to 1.77)	58.85	55.83	−3.02	−1.49 (−5.72 to 2.74)
Synthetic control	54.25	52.50	−1.75		58.85	57.32	−1.53
Pennsylvania	51.25	49.43	−1.82	1.01 (−1.51 to 3.53)	56.16	55.84	−0.32	1.24 (−2.61 to 5.09)
Synthetic control	51.25	48.42	−2.83		56.16	54.60	−1.56
Virginia	49.27	45.98	−3.29	−0.56 (−2.70 to 1.58)	53.04	50.43	−2.61	−1.10 (−4.00 to 1.80)
Synthetic control	49.27	46.54	−2.73		53.04	51.53	−1.51
Colorado	56.11	52.82	−3.29	−1.81 (−5.80 to 2.17)	60.69	58.73	−1.96	−1.93 (−6.48 to 2.63)
Synthetic control	56.11	54.63	−1.48		60.69	60.66	−0.03
Idaho	58.38	59.41	1.03	0.93 (−1.03 to 2.89)	58.79	60.64	1.85	1.37 (−1.65 to 4.39)
Synthetic control	58.38	58.48	0.10		58.79	59.27	0.48

MME = morphine milligram equivalents.

^*Augmented synthetic control analyses were used to estimate changes in outcomes between the prelaw and postlaw periods in the treatment state and its control group. This approach is designed to minimize differences in outcomes averaged over the prelaw period in the treatment state versus its synthetic control. The results in this table show the difference in the daily MME of opioid prescriptions per patient prescribed opioids per month in the 2 years before versus after the law implementation date in the treatment versus control states. Analyses were adjusted for mean patient age; mean patient Elixhauser Comorbidity Index; the proportion of patients who were male; the proportion with a mental illness diagnosis; the proportion with a substance use disorder diagnosis; and the proportions of the state population that were Black, were Hispanic, were living below the federal poverty line, were employed, and had no postsecondary degree, per state, per year. Across all 13 treatment states and their control groups, the overall sample included 7 694 514 unique adults aged ≥18 y and 1 976 355 unique adults aged ≥18 y with chronic noncancer pain conditions. State-specific analytic sample sizes are shown in Table 1 in the main text.

Appendix Table 6.

Change in Proportion of Patients Prescribed Opioids With ≥50 MME/d, per Month, Attributable to the Law During Its First 2 Years of Implementation^*

State	Adults Aged ≥18 y					Adults Aged ≥18 y With Chronic Noncancer Pain
	Mean Proportion Prescribed Opioids With ≥50 MME/d in Each Month, %		Difference, percentage points	Difference-in-Differences Estimate of Change Attributable to Law, percentage points	Mean Proportion Prescribed Opioids With ≥50 MME/d in Each Month, %		Difference, percentage points	Difference-in-Differences Estimate of Change Attributable to Law, percentage points
	In the 2 Years Before Law	In the 2 Years After Law			In the 2 Years Before Law	In the 2 Years After Law
Delaware	32.66	29.23	−3.43	−1.21 (−4.55 to 2.14)	33.41	32.66	−0.75	2.50 (−1.58 to 6.59)
Synthetic control	32.66	30.44	−2.22		33.41	30.16	−3.25
Kentucky	21.52	18.65	−2.87	−0.05 (−1.96 to 1.85)	21.74	19.29	−2.45	0.67 (−2.12 to 3.46)
Synthetic control	21.52	18.70	−2.82		21.74	18.62	−3.12
New York	25.14	25.14	0.00	−0.53 (−2.32 to 1.26)	28.23	26.44	−1.79	−1.28 (−3.76 to 1.20)
Synthetic control	23.70	24.23	0.53		28.23	27.72	−0.51
Ohio	32.66	29.23	−3.43	−1.21 (−4.55 to 2.14)	21.60	16.33	−5.27	−2.25 (−4.98 to 0.48)
Synthetic control	32.66	30.44	−2.22		21.60	18.58	−3.02
Mississippi	30.81	19.09	−11.72	−2.80 (−6.52 to 0.92)	30.19	19.40	−10.79	−4.18 (−8.26 to 0.11)
Synthetic control	30.81	21.89	−8.92		30.19	23.58	−6.61
Ohio	22.04	17.89	−4.15	−0.98 (−2.32 to 0.36)	23.24	19.41	−3.83	−1.33 (−2.97 to 0.30)
Synthetic control	22.04	18.87	−3.17		23.24	20.74	−2.50
Texas	27.70	20.14	−7.56	−3.55 (−6.12 to 0.97)	29.22	23.79	−5.43	−2.78 (−6.25 to 0.68)
Synthetic control	27.70	23.69	−4.01		29.22	26.57	−2.65
New York	31.49	32.73	1.24	1.34 (−1.17 to 2.78)	24.81	26.72	1.91	2.19 (−1.21 to 5.59)
Synthetic control	31.49	31.39	−0.10		24.81	24.53	−0.28
Oklahoma	27.56	25.79	−1.77	−1.46 (−4.29 to 1.37)	29.19	27.81	−1.38	−1.13 (−3.62 to 1.36)
Synthetic control	27.56	27.25	−0.31		29.19	28.94	−0.25
Pennsylvania	22.76	23.10	0.34	0.88 (−0.99 to 2.75)	25.96	27.17	1.21	1.39 (−1.78 to 4.57)
Synthetic control	22.76	22.22	−0.54		25.96	25.78	−0.18
Virginia	23.32	22.30	−1.02	−0.93 (−3.00 to 1.14)	25.28	25.07	−0.21	−1.03 (−3.92 to 1.87)
Synthetic control	23.32	23.23	−0.09		25.28	26.10	0.82
Colorado	30.05	28.40	−1.65	−1.42 (−4.08 to 1.25)	31.55	30.63	−0.92	−1.44 (−4.24 to 1.36)
Synthetic control	30.05	29.82	−0.23		31.55	32.07	0.52
Idaho	48.87	49.72	0.85	1.04 (−1.60 to 3.68)	48.08	49.80	1.72	1.28 (−2.07 to 4.64)
Synthetic control	48.87	48.68	−0.19		48.08	48.52	0.44

MME = morphine milligram equivalents.

^*Augmented synthetic control analyses were used to estimate changes in outcomes between the prelaw and postlaw periods in the treatment state and its control group. This approach is designed to minimize differences in outcomes averaged over the prelaw period in the treatment state versus its synthetic control. The results in this table show the percentage point difference in the proportion of patients prescribed opioids with ≥50 MME/d, per month, in the 2 years before versus after the law implementation date in the treatment versus control states. Analyses were adjusted for mean patient age; mean patient Elixhauser Comorbidity Index; the proportion of patients who were male; the proportion with a mental illness diagnosis; the proportion with a substance use disorder diagnosis; and the proportions of the state population that were Black, were Hispanic, were living below the federal poverty line, were employed, and had no postsecondary degree, per state, per year. Across all 13 treatment states and their control groups, the overall sample included 7 694 514 unique adults aged ≥18 y and 1 976 355 unique adults aged ≥18 y with chronic noncancer pain conditions. State-specific analytic sample sizes are shown in Table 1 in the main text.

Appendix Table 7.

Change in Proportion of Patients With Chronic Noncancer Pain Conditions Receiving Any Clinical Guideline-Concordant Nonopioid Pain Medication or Procedure, per Month, Attributable to the Law During Its First 2 Years of Implementation^*

State	Clinical Guideline-Concordant Pain Medications					Clinical Guideline-Concordant Pain Procedures
	Mean Proportion Receiving Any Guideline-Concordant Nonopioid Pain Medication in Each Month, %		Difference, percentage points	Difference-in-Differences Estimate of Change Attributable to Law, percentage points	Mean Proportion Receiving Any Guideline-Concordant Procedure in Each Month, %		Difference, percentage points	Difference-in-Differences Estimate of Change Attributable to Law, percentage points
	In the 2 Years Before Law	In the 2 Years After Law			In the 2 Years Before Law	In the 2 Years After Law
Delaware	19.88	20.04	0.16	0.63 (−1.03 to 2.29)	21.51	19.90	−1.61	−0.09 (−1.32 to 1.15)
Synthetic control	19.88	19.41	−0.47		21.51	19.99	−1.52
Kentucky	28.88	28.60	−0.28	0.04 (−0.75 to 0.83)	15.00	14.24	−0.76	0.66 (−0.33 to 1.66)
Synthetic control	28.88	28.56	−0.32		15.00	13.58	−1.42
New York	16.94	17.30	0.36	0.07 (−0.55 to 0.68)	14.97	14.20	−0.77	0.90 (−0.17 to 1.97)
Synthetic control	16.94	17.23	0.29		14.97	13.30	−1.67
Ohio	26.08	25.58	−0.50	0.10 (−0.41 to 0.60)	16.75	15.58	−1.17	0.80 (−0.48 to 2.09)
Synthetic control	26.08	25.48	−0.60		16.75	14.78	−1.97
Mississippi	24.90	26.54	1.64	0.26 (−1.25 to 1.78)	8.68	7.26	−1.42	0.14 (−0.49 to 0.78)
Synthetic control	24.90	26.28	1.38		8.68	7.12	−1.56
Ohio	21.97	22.63	0.66	−0.01 (−0.41 to 0.41)	12.39	11.24	−1.15	0.32 (−0.15 to 0.79)
Synthetic control	21.97	22.64	0.67		12.39	10.92	−1.47
Texas	23.21	23.21	0.00	−0.48 (−1.32 to 0.37)	11.94	9.46	−2.48	1.04 (−0.10 to 1.99)
Synthetic control	22.66	23.14	0.48		11.94	8.42	−3.52
New York	16.28	16.45	0.17	−1.48 (−4.38 to 1.42)	15.98	15.74	−0.24	0.71 (−0.77 to 2.19)
Synthetic control	16.28	17.93	1.65		15.98	15.03	−0.95
Oklahoma	25.91	26.23	0.32	0.52 (−0.89 to 1.94)	13.46	12.46	−1.00	0.24 (−0.65 to 1.12)
Synthetic control	25.91	25.71	−0.20		13.46	12.22	−1.24
Pennsylvania	19.31	19.36	0.05	−0.35 (−1.01 to 0.31)	16.08	14.53	−1.55	−0.15 (−0.71 to 0.42)
Synthetic control	19.31	19.71	0.40		16.08	14.68	−1.40
Virginia	22.08	21.65	−0.43	−0.52 (−1.54 to 0.51)	13.59	12.39	−1.20	0.29 (−0.39 to 0.96)
Synthetic control	22.08	22.17	0.09		13.59	12.10	−1.49
Colorado	18.17	18.62	0.45	−0.34 (−1.21 to 0.53)	15.72	14.99	−0.73	0.75 (−0.27 to 1.77)
Synthetic control	18.17	18.96	0.79		15.72	14.24	−1.48
Idaho	22.46	23.88	1.42	−0.86 (−1.82 to 0.11)	20.46	19.09	−1.37	−0.82 (−3.55 to 1.90)
Synthetic control	22.46	24.74	2.28		20.46	19.91	−0.55

^*Augmented synthetic control analyses were used to estimate changes in outcomes between the prelaw and postlaw periods in the treatment state and its control group. This approach is designed to minimize differences in outcomes averaged over the prelaw period in the treatment state versus its synthetic control. The results in this table show the percentage point difference in the proportion of patients with any clinical guideline-concordant nonopioid pain medication or procedure, per month, in the 2 years before versus after the law implementation date in the treatment state versus its control group. Analyses were adjusted for mean patient age; mean patient Elixhauser Comorbidity Index; the proportion of patients who were male; the proportion with a mental illness diagnosis; the proportion with a substance use disorder diagnosis; and the proportions of the state population that were Black, were Hispanic, were living below the federal poverty line, were employed, and had no postsecondary degree, per state, per year. The chronic noncancer pain sample included patients diagnosed with arthritis, low back pain, headache, fibromyalgia, or neuropathic pain in the prelaw period. Across all 13 treatment states and their control groups, the overall sample included 7 694 514 unique adults aged ≥18 y and 1 976 355 unique adults aged ≥18 y with chronic noncancer pain conditions. State-specific analytic sample sizes are shown in Table 1 in the main text.