Association Between Statewide Opioid Prescribing Interventions and Opioid Prescribing Patterns in North Carolina, 2006–2018

Courtney N Maierhofer; Shabbar I Ranapurwala; Bethany L DiPrete; Naoko Fulcher; Christopher L Ringwalt; Paul R Chelminski; Timothy J Ives; Nabarun Dasgupta; Vivian F Go; Brian W Pence

doi:10.1093/pm/pnab181

. 2021 Jun 27;22(12):2931–2940. doi: 10.1093/pm/pnab181

Association Between Statewide Opioid Prescribing Interventions and Opioid Prescribing Patterns in North Carolina, 2006–2018

Courtney N Maierhofer ^1,^✉, Shabbar I Ranapurwala ^1,², Bethany L DiPrete ¹, Naoko Fulcher ², Christopher L Ringwalt ^3,⁴, Paul R Chelminski ⁵, Timothy J Ives ^5,⁶, Nabarun Dasgupta ², Vivian F Go ⁴, Brian W Pence ¹

PMCID: PMC8665995 PMID: 34175958

Abstract

Objective

To examine the impact of three sequential statewide policy and legislative interventions on opioid prescribing practices among privately insured individuals in North Carolina.

Methods

An interrupted time series approach was used to examine level and trajectory changes of new and prevalent opioid prescription rates, days’ supply, and daily morphine milligram equivalents before and after implementation of a 1) prescription drug monitoring program, 2) state medical board initiative, and 3) legislative action. Analyses were conducted using individual-level claims data from a large private health insurance provider serving North Carolina residents, ages 18–64 years, from January 2006 to August 2018.

Results

Rates of new and prevalent prescription opioid patients were relatively unaffected by the prescription monitoring program but sharply declined in the months immediately following both medical board (−3.7 new and −19.3 prevalent patients per 10,000 person months) and legislative (−14.1 new and −26.7 prevalent patients) actions. Among all opioid prescriptions, days’ supply steadily increased on average over the study period but declined after legislative action (−1.5 days’ supply per year).

Conclusions

The voluntary prescription drug monitoring program launched in 2010 only marginally affected opioid prescribing patterns on its own, but its redeployment as an investigative and clinical tool in multifaceted public policy approaches by the state medical board and legislature later in the decade plausibly contributed to notable declines in prescription rates and days’ supply. This study lends new emphasis to the importance of enforcement mechanisms for state and national policies seeking to reverse this critical public health crisis.

Keywords: Drug Abuse, Opioids, Pain, Prescription Monitoring Program, PMP, Prescriptions

Introduction

While an increasing proportion of opioid-related deaths in the United States are attributable to illicit opioids (primarily fentanyl analogs/heroin), uncertainty remains about the appropriate societal response to opioid analgesics [1–3]. Pharmaceutical grade opioid agents are involved in about a third of all opioid overdose deaths nationally [4]. In response to two decades of increasing prescription overdose mortality, many states have enacted policy and legislative interventions to stem the epidemic [5].

A key target for such interventions has been reduction of the number of patients prescribed opioid analgesics. Initial opioid prescription dose and duration strongly predict future long-term opioid use [6]. Nearly all states have implemented prescription drug monitoring programs (PDMPs) to decrease patients’ ability to obtain prescriptions from multiple providers [7, 8]. Many states have also implemented legislation and policies aimed at identifying and potentially sanctioning high-volume prescribers or limiting the dose, duration, or number of initial opioid prescriptions for acute and chronic pain [9]. One approach is individual and disciplinary; the other is an attempt to change the prescribing environment. The impact of such efforts on opioid prescribing practices, especially over an extended period of time, remains understudied.

The objective of this paper was to examine the impact in North Carolina of three sequential statewide policy and legislative interventions on opioid prescribing practices: 1) establishment of a statewide PDMP, 2) launch of a state medical board policy focused on identifying and investigating high-volume prescribers, and 3) passage of state legislation limiting initial opioid prescriptions for injuries and surgeries.

Methods

An interrupted time series approach [10] was used to examine changes in the level and trajectory of opioid prescription rates, strength, and duration among privately insured individuals from 2006 to 2018, a period that spanned three key statewide policy and legislative changes.

Data Source

A total of 146 consecutive months of de-identified claims data were examined from a large single North Carolina provider of private health insurance. Data included the insured population’s demographics, place of residence, inpatient and outpatient professional and facility claims, and pharmacy claims.

Study Population

All North Carolina residents ages 18–64 years insured by the provider at any point from January 2006 through August 2018 (from first available to most recent available data at time of data pull) were included. Individuals entered the analysis after 6 months of continuous insurance coverage to allow time to identify prior prescriptions. Individuals with a gap in insurance coverage re-entered the analysis after six months of being re-insured. Individuals remained in analyses until either the end of insurance coverage or the end of the study period.

Outcomes

The monthly rate of new (incident) prescription opioid patients, all (prevalent) prescription opioid patients, and all opioid prescriptions over time were examined. New patients were defined as individuals filling an opioid prescription in any given month who had not filled any opioid prescription in the prior six months. Prevalent patients were defined as individuals filling an opioid prescription in any given month regardless of prescriptions in preceding months. For new prescription opioid patient rates, the numerator included the number of new patients each month; the denominator included all individuals who were currently insured that month without an observed opioid prescription in the previous 6 months (i.e., the population “at risk” of a new opioid prescription). The prevalent opioid patient rate was calculated as the number of all opioid patients in a given month divided by the number of all individuals who were currently insured that month regardless of prior prescriptions. The prevalent opioid prescription rate was calculated in the same way as the prevalent opioid patient rate, except that the numerator comprised all prescriptions in a given month rather than all individuals who received at least one prescription. Rates were reported per 10,000 person-months (PM).

Days’ supply dispensed and daily morphine milligram equivalent (MME) of both prevalent and new opioid prescriptions were also examined. For prevalent opioid prescriptions, the days’ supply and MME of all prescriptions were averaged within each calendar month. For new opioid prescriptions, the days’ supply and MME of the initial prescription for all patients who initiated opioids within each calendar month were averaged. Daily MME, based on the prescription’s National Drug Code, was calculated as the product of the dose per unit and the number of units dispensed, divided by the days’ supply of the prescription as recorded in the paid insurance claim. The resulting milligrams of medication per day was multiplied by an MME conversion factor from Centers for Disease Control and Prevention (CDC) tables [11]. These calculations were performed at the individual level in a person-day-level file prior to aggregation for analyses. Prescriptions that overlapped in time by more than 7 days were assumed to be simultaneous, whereas those that overlapped by seven or fewer days were assumed to represent early refills and were pushed forward accordingly.

Medications (outpatient only) excluded all opioid co-formulations to treat coughs, colds, or allergies (Supplementary Data).

Exposures

The impact of three distinct statewide interventions on prescription patterns was assessed: 1) the establishment of a PDMP; 2) a state medical board investigative initiative; and 3) state legislation to limit initial opioid prescriptions. PDMPs are centralized databases now established in all states but Missouri. In North Carolina, all pharmacies except those managed by the Veterans Administration and military installations are mandated to report all filled outpatient opioid prescriptions. Data from the PDMP, which took effect on January 1, 2010, can be accessed at the point of care. The state medical board launched the Safe Opioids Prescribing Initiative (SOPI) in May 2016 to address overprescribing of opioids; this initiative identifies and sends letters to clinicians who prescribe high doses of opioids to a large number of patients or who have had ≥2 patients who have died from opioid overdose-related causes [12]. The state legislature passed the Strengthen Opioid Misuse Prevention (STOP) Act which 1) mandates that opioid prescriptions prescribed by physician assistants and nurse practitioners be reviewed by a licensed physician, 2) limits the maximum days’ supply of an initial prescription for acute (five days) or post-surgical (seven days) pain, and 3) requires physicians to review the PDMP prior to prescribing a schedule II or III opioid or narcotic. The first and second provisions went into effect July 1, 2017, and January 1, 2018, respectively, while the third had yet to take effect by the end of the study period [13]. As a result of this law, the insurance provider mandated a maximum of seven days’ supply on new opioid prescriptions effective April 1, 2018.

The PDMP was defined as taking effect in January 2010, the medical board initiative as May 2016, and the state legislation as January 2018.

Covariates

We accounted for two national events that may affect the measurement of our outcomes. The rollout of the Affordable Care Act’s (ACA) Health Insurance Marketplace on January 1, 2014, brought many new individuals into the private insurance market who differed systematically from those previously insured, leading to observable changes in preliminary analyses. We accounted for this change in the underlying insured population by including an indicator term for the pre-ACA vs ACA period, defined as July 1, 2014, due to our six-month washout period. As we did not expect that the ACA would impact opioid prescribing practices, we did not include a change-in-slope term. Additionally, a tamper-resistant formulation of OxyContin (oxycodone extended-release, Purdue Pharma LP, Stamford, Connecticut) intended to deter abuse was released in August 2010 and achieved large-scale distribution in November 2010, identified in prior work as leading to a decrease in daily MME prescribed without affecting other indicators [14, 15]. We accounted for this formulation change by including an indicator variable for pre- vs post-introduction (December 2010) in our MME models only.

Statistical Analyses

An autoregressive integrated moving average (ARIMA) model was used to conduct interrupted time series analyses [10] with three inflection time points, allowing changes in opioid prescribing levels and trends before and after the three statewide interventions. Our model for these analyses is as follows:

\begin{array}{l} {outcome}_{t} = β_{0} + β_{1} ({time}_{t}) \\ + β_{2} (PDMP) + β_{3} ({PDMP trend}_{t}) \\ + β_{4} (medical board) + β_{5} ({medical board trend}_{t}) \\ + β_{6} (legislation) + β_{7} ({legislation trend}_{t}) \\ + β_{8} (ACA) + ε_{t}, \end{array}

where time is a continuous variable representing months since June 2006; PDMP, medical board, and legislation are dichotomous variables indicating time before and after each policy change; PDMP trend, medical board trend, and legislation trend are continuous variables set to 0 before the respective policy change and time in years after each policy change; and ACA is a dichotomous variable indicating pre-ACA vs ACA time to absorb the related change in the insured population. With this specification, β₁ estimates the pre-PDMP trend of the outcome; β₂, β₄, and β₆, estimate the absolute change in outcome immediately after the enactment of each policy (the immediate impact of the policy); and β₃, β₅, and β₇ estimate the difference in the pre- and post-policy trajectories of the outcome for each policy (the change in the trend line after each policy) [16]. In models where mean daily MME was the outcome, we added a dichotomous term to absorb the change in MME attributable to the release of the abuse-deterrent OxyContin formulation.

Results are presented from the full population as well as stratified by sex and age. While data were aggregated and analyzed by month, regression estimates are annualized and expressed as changes per year for ease of interpretation. Consistent with recommendations of the American Statistical Association and major medical journals to de-emphasize drawing conclusions based on statistical significance testing [17–19], this article does not focus on statistical significance tests, but rather results are reported with 95% confidence intervals (CIs) to provide a measure of precision. Analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA).

Compliance with Ethics Guidelines

This study was approved by the Institutional Review Board, Office of Human Research Ethics, University of North Carolina at Chapel Hill.

Results

The analysis sample comprised 3,034,688 unique privately insured North Carolina residents, of whom, on average in any given month, 877,000 were currently insured and 50.3% were female. The largest age group in the sample was those ages 45–54 years (26%), while the smallest was those ages 18–24 years (10%). There was a monthly average of 12,079 new opioid patients and a monthly average of 37,264 overall opioid patients and 49,280 opioid prescriptions (Table 1).

Table 1.

Number of monthly insured patients, prescription opioid patients, and opioid prescriptions from a single North Carolina provider of private health insurance, July 2006 through August 2018

Categories per Month	Number Insured	Number of New Prescription Opioid Patients^∗	Number of All Prescription Opioid Patients^†	Number of All Opioid Prescriptions^†
Categories per Month	Mean (Range)	Mean (Range)	Mean (Range)	Mean (Range)
Overall	876,663.1 (769,329–1,025,817)	12,079.3 (8258–14,620)	37,264.7 (29,396–48,246)	49,280.5 (36,564–63,896)
Sex
Female	440,525.5 (383,429–522,851)	6598.3 (4710–8133)	20,389.1 (16,709– 26,740)	26,735.0 (20,754–35,021)
Male	436,137.7 (385,900–502,966)	5480.9 (3509–6563)	16,875.5 (12,513–21,567)	22,545.5 (15,655–28,894)
Age (years)
18–24	91,650.7 (66,064–116,743)	1039.9 (497–1607)	1599.2 (684–2314)	1999.2 (756–2888)
25–34	165,012.1 (146,054–189,617)	2221.0 (1353–2877)	4806.7 (2609–6081)	6416.0 (3163–8444)
35–44	198,074.1 (161,811–219,418)	2795.4 (1604–3711)	8004.8 (5052–534)	10,784.5 (6325–13,402)
45–54	225,181.3 (195,134–257,219)	3181.6 (2165–3826)	11,646.3 (9000–14,852)	15,582.6 (11,269–19,824)
55–64	196,744.9 (150,484–253,581)	2,841.3 (2116–3929)	11,207.7 (7337–16,584)	14,498.3 (9769–21,786)

Open in a new tab

^∗

New prescription opioid patient population includes person-months where the individual has been insured continuously for ≥6 months and has no opioid prescription in the prior 6 months.

^†

All prescription opioid patient population includes person-months where the individual has been insured continuously for ≥6 months regardless of prior opioid prescriptions.

New Opioid Patients

Prescribing Rate

The incidence rate of new opioid patients decreased by over one-third from 2006 to 2018. As seen in Figure 1, the new patient rate followed a fairly slow decline prior to 2010 that accelerated after the introduction of the PDMP, accelerated further with the medical board initiative, and then declined even more sharply following legislative action (Figure 1). The graphical evidence displayed is supported by ARIMA model estimates that indicate a slow decline prior to 2010 (decline in incidence rate of −1.6 (95% CI: −3.7, 0.5) new patients per 10,000 per year), a faster decline during the PDMP period (additional decline: −4.7 [95% CI: −7.2, 2.2] new patients per 10,000 per year) and an even faster decline under the medical board investigative initiative (additional decline: −2.8 [95% CI: −9.2, 3.7] new patients per 10,000 per year) (Table 2). There is no strong evidence of an immediate decline, or disjunction, in the new patient rate with the advent of either the PDMP (0.1 [95% CI: −5.4, 5.6] new patients per 10,000 PM) or medical board initiative (−3.7 [95% CI: −10.8, 3.3] new patients per 10,000 PM); however, a notable drop occurred immediately after legislative action in 2018 (−14.1 [95% CI: −26.2, −2.1] new patients per 10,000 PM).

Table 2.

Association of three policies on opioid prescribing behaviors, controlling for a national policy, in North Carolina from July 2006 through August 2018

	Pre-PDMP	Post-PDMP		Post-Medical Board Initiative		Post-Legislation
Outcome Categories	Trend (95% CI)	Absolute Difference (95% CI)	Change in Trend (95% CI)	Absolute Difference (95% CI)	Change in Trend (95% CI)	Absolute Difference (95% CI)	Change in Trend (95% CI)
	β1	β2	β3	β4	β5	β6	β7
New prescription opioid patients^∗
Prescribing rate^†	−1.6 (−3.7, 0.5)	0.1 (−5.4, 5.6)	−4.7 (−7.2, −2.2)	−3.7 (−10.8, 3.3)	−2.8 (−9.2, 3.7)	−14.1 (−26.2, −2.1)	5.0 (−20.9, 30.9)
Mean days’ supply	0.1 (0.1, 0.1)	0 (−0.2, 0.1)	0.0 (−0.1, 0.0)	0.0 (−0.2, 0.1)	−0.4 (−0.6, −0.3)	−0.1 (−0.4, 0.1)	−2.2 (−2.7, −1.7)
Mean daily MME	−0.6 (−0.7, −0.4)	0.1 (−0.3, 0.6)	0.3 (0.1, 0.5)	−0.9 (−1.5, −0.4)	−0.3 (−0.8, 0.3)	−0.5 (−1.5, 0.5)	−0.8 (−2.9, 1.3)
All prescription opioid patients^‡
Prescribing rate^†	6.7 (2.8, 10.7)	−2.5 (−13.1, 8.1)	−13.3 (−18.1, −8.5)	−19.3 (−32.8, −5.7)	−3.4 (−15.9, 9.0)	−26.7 (−50.0, −3.5)	−38.4 (−88.2, 11.5)
All opioid prescriptions^‡
Prescribing rate^†	2.9 (−4, 9.8)	−5.1 (−23.5, 13.3)	−17.3 (−25.6, −9)	−23.4 (−46.9, 0.1)	−2.9 (−24.5, 18.7)	−53.9 (−94.3, −13.5)	−12.8 (−99.4, 73.7)
Mean days’ supply	0.5 (0.5, 0.5)	0.0 (−0.1, 0.1)	−0.1 (−0.1, −0.1)	0.0 (−0.1, 0.1)	−0.1 (−0.2, 0.0)	0.1 (−0.1, 0.3)	−1.5 (−2.0, −1.1)
Mean daily MME	−0.1 (−0.3, 0.0)	0.3 (−0.2, 0.7)	−0.4 (−0.6, −0.2)	−0.1 (−0.6, 0.4)	−1.8 (−2.3, −1.4)	0.2 (−0.6, 1.1)	−0.6 (−2.4, 1.2)

Open in a new tab

Note: PDMP = Prescription Drug Monitoring Program; Medical Board Initiative = Safe Opioid Prescribing Initiative; Legislation=Strengthen Opioid Misuse Prevention Act; CI = confidence interval; MME = morphine milligrams equivalent. All models controlled for the implementation of the Affordable Care Act in 2014; daily MME models additionally controlled for new formulation of OxyContin on December 1, 2010 (released in August 2010 and broadly distributed in November 2010).

^∗

New prescription opioid patient population includes person-months where the individual has been insured continuously for >=6 months and has no opioid prescription in the prior 6 months.

^†

Prescribing rates per 10,000 insured person-months; trends calculated per 10,000 per year.

^‡

All prescription opioid patient population includes person-months where the individual has been insured continuously for >=6 months regardless of prior opioid prescriptions.

Days’ Supply

The days’ supply in new opioid patients’ first opioid prescription increased gradually from a mean of just under 6 days’ supply at the beginning of the study period, with no evident change following the PDMP launch, to a mean of just over 6 days’ supply at the time of the medical board initiative (Figure 2). Days’ supply then declined and fell sharply to approximately 4 days’ supply with the legislative action. From the regression model, days’ supply increased at a rate of 0.1 (95% CI: 0.1, 0.1) additional days’ supply per year pre-PDMP, continued to increase at a similar rate post-PDMP (change in rate: 0.0 [95% CI: −0.1, 0.0] additional days per year), began to decline after the medical board initiative (change in rate: −0.4 [95% CI: −0.6, −0.3] days’ supply per year), and declined even more rapidly after legislative action (change in rate: −2.2 [95% CI: −2.7, −1.7] days’ supply per year) (Table 2).

Figure 2. — Mean days’ supply of opioid prescriptions among privately insured North Carolina residents from July 2006 through August 2018. PDMP = Prescription Drug Monitoring Program; ACA=Affordable Care Act; Medical Board Initiative = Safe Opioid Prescribing Initiative; Legislation = Strengthen Opioid Misuse Prevention Act. New prescription opioid patient population includes person-months where the individual has been insured continuously for ≥6 months and has no opioid prescription in the prior 6 months. All prescription opioid patient population includes person-months where the individual has been insured continuously for ≥6 months regardless of prior opioid prescriptions.

Daily MME

Among new prescription opioid patients, mean daily MME in the first prescription declined from approximately 48 in 2010 to 35 in 2018 (Figure 3). Over half of this decline coincided with the release of the abuse-deterrent OxyContin 80 mg formulation in 2010, which has been observed in other data to have led a large number of patients to switch away from this formulation resulting in a sharp population-level drop in mean daily MME [14, 15]. Otherwise, the mean daily MME declined gradually throughout the study period. The ARIMA model estimated that the pre-PDMP rate declined −0.6 (95% CI: −0.7, −0.4) daily MME per year; this decline slowed slightly post-PDMP (0.3 [95% CI: 0.1, 0.5]), but re-accelerated under the medical board initiative (−0.3 [95% CI: −0.8, 0.3]) and accelerated further after legislative action (−0.8 [95% CI: −2.9, 1.3]). The mean daily MME level fell immediately when the medical board initiative was launched (−0.9 [95% CI: −1.5, −0.4]) and when legislative action took effect (−0.5 [95% CI: −1.5, 0.5]) (Table 2).