Medical prescription opioid use remains common despite increasingly stringent prescribing policies (Lee et al., 2021; National Academies of Sciences, 2017). Current Centers for Disease Control and Prevention (CDC) guidelines (Dowell et al., 2016) suggest prescribing opioids for pain management only if pain and function benefits outweigh risks. Clinical indications for opioid initiation vary widely and commonly include acute, postsurgical, chronic cancer-related, and chronic non-cancer-related pain, with varying risk-benefit relationships across these indications (Liu et al., 2013; Mikosz et al., 2020; Mundkur et al., 2018; Pasricha et al., 2018; Shah et al., 2017). Because prescription opioid use in patients with acute and postsurgical pain has been associated with new-onset long-term opioid use, opioid use disorder (OUD), and overdose (Baker 2017; Hadlandsmyth et al., 2019; Hah et al., 2017), most policy efforts have focused on limiting initial opioid prescription dose and duration for these pain indications.

Although prescribing policies to reduce opioid misuse are necessary, policy efforts may have unintended consequences (Centers for Disease Control and Prevention, 2019; Dowell et al., 2019; Ranapurwala et al., 2021). One unintended consequence is the under-prescribing of opioids in patients with chronic non-cancer and cancer pain, either due to imprecise policy language or an increasingly hostile prescribing environment (Kroenke et al., 2019). Misapplication of guidelines, such as abrupt discontinuation or rapid taper of high-dose opioids (Ranapurwala et al., 2021), may compromise patient safety (Centers for Disease Control and Prevention, 2019; Dowell et al., 2019) by causing patients with sudden uncontrolled pain to seek diverted or illegal opioids, increasing risk of OUD, overdose, and suicide (Binswanger et al., 2020; Bohnert and Ilgen, 2019; Coffin et al., 2020; Dowell et al., 2019; Hartung et al., 2020; Muzzio et al., 2021; Perez et al., 2020). Policies’ unintended prescribing impacts must be understood so that future policies can safely and clearly distinguish between problematic and legitimate prescribing—in particular, addressing appropriate limits for the management of acute and postsurgical pain without negatively impacting clinically indicated management of chronic non-cancer and cancer pain (Kroenke et al., 2019).

In response to the CDC’s 2016 opioid prescribing guidelines for chronic non-cancer pain (Dowell et al., 2016), North Carolina enacted state policies to improve patient care and safety by focusing on prescribers. In May 2016, the state medical board, intending to restrict opioid prescribing at high doses and high volume, launched an investigative initiative to identify clinicians who prescribe large numbers of high-dose opioids or had ≥2 patients die from opioid overdose within a 12-month period (North Carolina Medical Board, 2016; North Carolina Medical Society, 2017). In January 2018, North Carolina enacted legislation intended to, in part, limit the initial prescription of opioids for outpatient acute and postsurgical pain to five and seven days, respectively (General Assembly of North Carolina Session, 2017; North Carolina Medical Board, 2017). We conducted a study to examine the impacts of the two statewide North Carolina policies on opioid prescribing patterns among privately insured postsurgical, acute, and chronic pain patients, overall and by history of cancer.

The importance of implementing evidence-based, safe and appropriate opioid prescribing policies is undisputed (Dowell et al., 2019; Lowenstein et al., 2018), however, the potential misapplication of current policies remains understudied and may have substantial adverse implications for patient welfare. To understand the consequences of two opioid prescribing policies, we examined changes in initial opioid prescription rates, days’ supply, and daily MME for privately insured persons with acute, postsurgical, and chronic pain, overall and by cancer history, between 2012 and 2018.

In a prior study examining the impact of North Carolina policies on opioid prescribing practices in the same population (Maierhofer et al., 2021), we demonstrated that prescribing rates and days’ supply declined overall after both the statewide medical board initiative and legislative action. The present analysis presents a more nuanced picture considering pain indication and cancer history: some of the previously reported overall declines may have been driven by declines in unintended or unexpected subpopulations.

The first policy, the NCMB’s investigative initiative or SOPI (North Carolina Medical Society, 2017), intended to curtail opioid prescribing at high doses and high volume, especially in persons with chronic pain. However, we observed no overall changes in daily MME among patients with chronic pain, but immediate and sustained declines among patients with acute and postsurgical pain following the initiative. Post-initiative prescribing rates in single-series models were in accordance with the medical board’s intended consequences with immediate and sustained declines observed among persons with chronic pain overall and by cancer history; postsurgical and acute pain prescribing rates did not appear to be impacted by the initiative. Conversely, in models with a benzodiazepine control series, no apparent changes in prescribing rates occurred for those with chronic pain. For days’ supply, we observed trend declines overall and within most subgroups in single-series models, but there were no notable prescribing changes among those with chronic pain in controlled analyses. These differences between models demonstrate that at the time of the medical board initiative implementation, there were likely other co-interventions that impacted the prescribing rates and days’ supply of both benzodiazepines and opioids. Using another controlled substance as a control series allowed for the control of potential confounding from co-interventions, achieving more unbiased results than those from the single-series models.

The second policy, the STOP Act (North Carolina Medical Board, 2017), is a state legislative action created to address “overprescribing” of opioids, most directly through limiting initial outpatient prescriptions’ days’ supply for acute and postsurgical pain. Notably in CITS analyses, there was no evidence of overall prescribing declines in days’ supply for acute and postsurgical pain groups, but strong evidence of unintended trend declines in all chronic pain groups relative to benzodiazepines. In the single-series models, mean days’ supply trends modestly decreased in postsurgical and acute pain groups as intended by the initiative; however, comparatively, chronic pain prescriptions were more strongly impacted following the policy. Similar to the medical board initiative, the results from controlled analyses were likely more unbiased estimates of the impact of STOP Act in North Carolina.

Pain medicine societies and experts have expressed concerns regarding unintended effects of opioid prescribing guidelines and mandates (Centers for Disease Control and Prevention, 2019; Dowell et al., 2019; Kroenke et al., 2019; Ranapurwala et al., 2021), in particular the rapid tapering of opioid prescriptions in those with chronic pain, which is associated with overdose from illicit opioid use (Binswanger et al., 2020; Coffin et al., 2020; Hartung et al., 2020). Several studies using national databases demonstrate decreasing trends in high-dose opioid prescriptions and prescribing duration following the release of the 2016 CDC Guidelines (Bohnert et al., 2018; Goldstick et al., 2021). For example, a recent study using national claims data (Goldstick et al., 2021) found that opioid-naïve patients were less likely to receive initial opioid prescriptions ≥50 daily MME and received fewer days’ supply after the CDC guidelines. Although important for informing opioid prescribing policies, the bulk of such published studies have focused on the policies’ intended population, and not populations in which unintended prescribing consequences may occur (Agarwal et al., 2019; Hincapie-Castillo et al., 2021; Oueini et al., 2021). SOPI was modeled after the CDC guidelines for chronic pain and was implemented a month after the guidelines’ publication (North Carolina Medical Board, 2016; North Carolina Medical Society 2017), but through a granular examination by pain indication and cancer history, we observed the unintended consequences in prescribing practices described above, as well as those that occurred following the legislative action.

Three months following the STOP Act, in response to that legislation, the private health insurer implemented internal reimbursement policies requiring a limit of seven days’ supply for initial immediate-release opioid prescriptions for acute pain and prior authorization for initial extended-release prescriptions. When shifting the interruption timepoint from the legislation implementation date to the reimbursement policy start date, we observed considerable immediate declines in days’ supply among persons with chronic pain, with no apparent declines in other pain groups. The insurer’s policy changes likely contributed to the post-legislation trend declines observed in persons with chronic pain in the main analyses, and may have unintentionally reduced the days’ supply of chronic pain prescriptions instead of the intended acute and postsurgical pain prescriptions. It may be that chronic pain patients with previous or current opioid prescriptions who change health insurance providers are considered to be opioid-naïve by the insurance provider, thereby limiting prescriptions. While opioid prescribing remains high compared to the pre-2000 years, it is critical to avoid limiting legitimate opioid prescriptions to patients who can safely benefit from pharmacotherapy. Policies and mandates must be well communicated to both prescribers and insurers to avoid potentially harmful unintended effects. Additionally, future studies must examine longer follow-up periods to better understand the impact of the state legislation and reimbursement policy on days’ supply.

The unexpected prescribing reactions to these policies may partly be associated with providers’ perceived risk of regulatory or legal sanction for opioid prescribing (Bolen 2006; Jung and Reidenberg, 2006; Passik and Kirsh, 2006). Given the high frequency of opioid prescribing for acute pain, misapplication of guidelines may occur for this indication if prescribing fear is consistently present in providers’ minds without considering the intended population of a new policy. Prescribing fears may also have contributed to decreased prescribing rates in those with chronic pain after the implementation of the STOP Act, especially since the severity of the opioid epidemic intensified during this period. Additionally, although benzodiazepines were not targeted in these statewide policies, benzodiazepine prescribing may have been unintentionally influenced through provider fear of potential patient overdose with the co-prescribing of opioids and benzodiazepines (U.S. Food and Drug Administration, 2017); however, we assume influence to be minimal without the direct mention of benzodiazepines in SOPI and the STOP Act.

This study has several limitations. We focused analyses on persons for whom we had a reasonable basis of characterizing their pain indication as acute, postsurgical, or chronic using a hierarchical classification based on ICD and CPT codes. A large percentage (45%) of insured persons with a newly prescribed opioid did not have a contemporaneous ICD/CPT code on which to base a pain indication and were therefore excluded from analyses. Because we included only major surgeries in our postsurgical criteria and lacked information on dental procedures (not systematically collected within our data), persons with new opioid prescriptions related to minor surgeries or dental procedures with no other pain indication may account for a proportion of excluded persons. These exclusions may limit generalizability of the results, given this group had a higher rate of new opioid prescriptions than any of the other pain indication groups (data not shown). Results for the overall population, which includes the three pain groups and all other persons without a characterized pain indication, can be found in our prior publication (Maierhofer et al., 2021). Furthermore, the results of this study may not be generalizable to uninsured or publicly insured persons, or to US states with prescription opioid dispensing rates dissimilar to those of North Carolina. During the study period, North Carolina’s overall opioid prescribing trends were higher than the US average (Centers for Disease Control and Prevention, 2020), and consequently these findings are likely more applicable to the privately insured populations in bordering southeastern states with similar opioid prescribing patterns. Because claims data capture information on limited demographics, it was not possible to examine prescribing trends by race and ethnicity. Claims data reflect only prescription claims and not prescription opioid consumption or non-prescription opioid use. Lastly, the cancer group included all persons with a cancer history which may not reflect current cancer diagnosis, and opioid prescriptions in this group may not always be related to cancer pain.