Since 1999, almost 500,000 deaths in the United States have resulted from opioid overdoses [1]. Although prescription opioids can be beneficial, their use can also present serious risks, including overdose and opioid use disorder. In 2019, more than 25% of opioid overdose deaths involved a prescription opioid [1].

Opioid prescribing practices have evolved in recent history. Dispensed opioid prescriptions in the United States increased from 72.4 per 100 persons in 2006 to 81.2 in 2010, when mean morphine milligram equivalents (MME) per capita were four times higher than in 1999 [2]. This period coincided with increases in rates of opioid overdose deaths [1], emergency department visits related to opioid use [3], and opioid use disorder [4].

To provide recommendations for pain management for adults in primary care settings, the U.S. Centers for Disease Control and Prevention (CDC) released in 2016 its Guideline for Prescribing Opioids for Chronic Pain (the CDC Guideline) [5]. The CDC Guideline provides evidence-based recommendations for the use of opioids in treating chronic pain in primary care settings (excluding cancer and palliative care) for patients 18 years of age and older and aims to improve patient–clinician communication about opioid therapy, improve the safety and effectiveness of pain management, and reduce the risks associated with long-term opioid use.

Although opioid prescribing began to decline in 2012 [2], the release of the CDC Guideline has been associated with accelerated temporal declines in opioid prescribing at a national level [6]. During 2012–2016, dispensed opioid prescriptions per 100,000 persons declined at a rate of 23 fewer prescriptions per month; after the release of the CDC Guideline, prescribing declined by 57 fewer prescriptions per month, a statistically significant acceleration [6]. Although this observation might have been related to secular trends in prescribing, accelerated declines in opioid prescribing exceeded changes in prescribing of other medications (e.g., benzodiazepines) during the same time period [6]. However, substantial variation exists in opioid prescribing by medical specialty [7]. During 2016–2017, the mean number of prescriptions dispensed was 215 per prescriber but ranged as high as 1,314 prescriptions per pain medicine specialist and as low as 14 prescriptions per pediatrician [7].

To date, the extent of prescribing changes among primary care practitioners (the intended audience of the CDC Guideline) and medical specialists after the release of the CDC Guideline has not been examined. In the assessment of policy change interventions, interrupted time-series analysis offers one of the most robust and commonly used statistical approaches for evaluating temporal changes in aggregate data [8]. Using interrupted time-series analysis, the present study aims to assess changes in opioid prescribing after the release of the CDC Guideline among medical disciplines likely to practice primary care, as well as among a variety of non–primary care specialties.

This analysis illustrates temporal trends in opioid dispensing among medical specialties in the periods before and after the release of the 2016 CDC Guideline [5]. Monthly dispensing rates were declining at the time of the CDC Guideline release [6], with declines observed across most medical specialties. These results demonstrate that accelerated national declines after the release of the CDC Guideline [6] were driven largely by declines in primary care specialties (i.e., family practice and internal medicine), those specifically targeted by the CDC Guideline; declines in annual opioid dispensing in these two specialties accounted for 50% of reductions across all specialties after the release of the CDC Guideline.

The present report represents the first findings of associations between the release of the CDC Guideline and changes in multiple opioid dispensing characteristics by distinct medical specialties. However, this work builds on other recent studies that have examined the impacts of the CDC Guideline. Declining national trends in opioid prescribing were shown to have accelerated after the release of the CDC Guideline, exceeding changes seen in the prescribing of other medications [6]. Among opioid-naïve patients with chronic pain, the duration of initial opioid prescriptions (which had been increasing) began decreasing after the release of the CDC Guideline [12], as did prescribing of doses of at least 50 MME per day [13]. As declines observed after the release of the CDC Guideline were similar across patients with different chronic pain indications (e.g., chronic headache and osteoarthritis), questions remain about the tailored patient-specific implementation of the CDC Guideline [14].

Additionally, the present findings indicate that specialties beyond primary care (e.g., surgery) also exhibited accelerated declines after the release of the CDC Guideline. These findings corroborate recent reports about declines in specific disciplines (such as surgery [15] and emergency medicine [16]) and expand on these findings by evaluating a wide range of specialties with unique and variable perspectives on the benefit–risk balance of opioid prescribing (such as addiction medicine, pain management, palliative care, and hematology/oncology). Collectively, these findings suggest that although the CDC Guideline was targeted specifically to primary care practitioners, clinicians in other specialties might have implemented aspects of the CDC Guideline intended for primary care settings (as the result of changes in individual prescribing or health system changes reinforcing the CDC Guideline [13]), resulting in concurrent accelerated declines. Alternatively, this could indicate that the CDC Guideline and other efforts at that time aimed at curbing excessive opioid prescribing and reducing opioid overdoses might have prompted specialists to independently reexamine their prescribing practices. This highlights the importance of specialty-specific guidance for opioid prescribing.

The analyses presented here have limitations. These data capture only prescriptions dispensed from retail pharmacies but do not capture mail-order, long-term care prescriptions, prescriptions dispensed from federally funded facilities (e.g., facilities affiliated with the Indian Health Service or Department of Veterans Affairs), or prescriptions that were written but not filled. Also excluded from this analysis were prescriptions from clinicians for whom a medical specialty could not be categorized (e.g., nurse practitioners and physician assistants), accounting for approximately 19% of dispensed opioid prescriptions. Although a majority of nurse practitioners and physician assistants practice specialties providing primary care [17, 18], they constitute a minority of all primary care practitioners [19] and might be unlikely to impact the direction of primary results observed. These data do not capture patient-level indicators, such as the clinical indications for a prescription or whether a patient took a full prescription when dispensed. In the absence of clinical indications, it is not possible to ascertain whether the reductions observed were due to decreases in non–evidence-based prescribing. Additionally, these analyses use ecological data to evaluate temporal trends; observed changes in opioid dispensing after March 2016 could be the result of a variety of factors in addition to (or separate from) the CDC Guideline, such as increased attention to opioid prescribing at the federal, state, and local levels during this period. Without a concurrent control group for comparison, this analysis cannot exclude the potential association between secular factors in prescribing trends and the results observed here. Such secular factors could include concurrent changes in opioid policies, such as increased resources funded through the 21^st Century Cures Act (December 2016) and increased utilization of prescription drug monitoring programs (available in all states by April 2017). Therefore, these results should not be viewed as causal, as historical biases and other trends might have contributed to the differences observed. Finally, this study did not examine trends in non-opioid or nonpharmacological treatments for pain, and therefore it could not assess whether decreases in opioid dispensing were offset by increases in other modalities for adequate pain management.

These findings serve to further contextualize changes in national opioid prescribing after the release of the CDC Guideline. This analysis offers evidence that declining trends in opioid dispensing accelerated among specialties likely to provide primary care, the intended audience of the CDC Guideline. These results also suggest that specialties beyond primary care also experienced accelerated declines in opioid dispensing after the release of the CDC Guideline, underscoring the importance of specialty-specific translation materials and trainings. Clinical management of pain requires integrated care, and use of prescription opioids should be based on an individualized assessment that balances the potential risks and benefits of opioid use along with other pain treatment modalities. The CDC Guideline provides a framework to support primary care practitioners in providing optimal care.