Nurse Practitioner Scope‐of‐Practice Laws and Opioid Prescribing

Abstract

Policy Points.

• The increased use of nurse practitioners represents a viable policy option to address continuing access‐to‐care deficiencies across the United States, but state scope‐of‐practice laws limit the ability of nurse practitioners to deliver health care.

• Groups in favor of restrictive scope‐of‐practice laws have argued that relaxing these laws will lead to increases in opioid prescriptions during an already severe opioid crisis, implicating patient safety concerns.

• An examination of a data set of 1.5 billion opioid prescriptions demonstrates that relaxing nurse practitioner scope‐of‐practice laws generally reduces opioid prescriptions. This evidence supports eliminating restrictive scope‐of‐practice laws that currently govern nurse practitioners in many states.

Context

As many parts of the United States continue to face physician shortages, the increased use of nurse practitioners (NPs) can improve access to care. However, state scope‐of‐practice (SOP) laws limit the ability of NPs to provide care by restricting the services they can provide and often requiring physician supervision of their practices. One important justification for the continuation of these restrictive SOP laws is preventing the overprescription of certain medications, particularly opioids.

Methods

This study examined a data set of approximately 1.5 billion individual opioid prescriptions between 2011 and 2018, which were aggregated to the individual provider‐year level. A series of difference‐in‐differences regression models was estimated to examine the association between laws allowing NPs to practice independently and opioid prescribing patterns among physicians and NPs. Opioid prescriptions were measured in total annual morphine milligram equivalents (MMEs) prescribed by individual providers.

Findings

Across all NPs and physicians, independent NP practice was associated with a statistically significant decline of 6%, 2%, 3%, 7%, and 5% in total annual MMEs prescribed to commercially insured, cash‐paying, Medicare, government‐assistance, and all patients, respectively. Medicaid patients saw no statistically significant change in annual MMEs. Across all payers, NPs generally increase and physicians generally decrease the number of opioids they prescribe following a grant of NP independence. These counterbalancing changes result in an overall net decline in MMEs.

Conclusions

No evidence supports the contention that allowing NPs to practice independently increases opioid prescriptions. The results support policy changes that allow NPs to practice independently.

More than a decade after the passage of the Affordable Care Act, access to health care continues to dominate health policy debates. The access debate has largely centered on ways to improve access to health insurance, with less emphasis on ways to improve access to health care providers. However, recent work has suggested that many areas of the country lack an adequate supply of providers, and this dearth of supply will only become more acute over the next decade.¹, ², ³, ⁴ One policy option that has the potential to mitigate provider shortages is the increased use of nurse practitioners (NPs) to provide care alongside physicians and other providers.⁵

Although NPs provide health care services across the country, their ability to do so is not equal in all areas. State scope‐of‐practice (SOP) laws limit the services NPs may provide and often require physician supervision of NP practices, both of which limit the ability of NPs to provide care.⁶, ⁷, ⁸ Recent research has found that relaxing SOP laws can increase access to care,⁶, ⁷, ⁸, ⁹ improve quality of care,⁸ reduce the use of intensive medical procedures,¹⁰ and lower the price of some health care services.¹¹ Based on this research and similar evidence, several national organizations have recommended that states grant NPs more autonomy.¹², ¹³, ¹⁴

Roughly half of all states have followed these recommendations. One important reason states may maintain restrictive SOP laws for NPs is the potential of NP independence to increase opioid prescriptions. The alleged reason for such an increase is straightforward: if NPs can prescribe opioids without physician supervision, then they may overprescribe opioids and worsen the ongoing opioid crisis.¹⁵, ¹⁶, ¹⁷, ¹⁸ Those opposed to reforming state laws to allow NPs to practice independently have explicitly made this overprescription argument in connection with their opposition.¹⁹ If, as alleged, an increase in opioid prescriptions does occur following a grant of independence to NPs, states may be justified in restricting the practices of NPs, despite evidence demonstrating that NPs provide care of similar quality to physicians²⁰, ²¹, ²², ²³ and that relaxing NP SOP laws can improve the health care system along multiple dimensions.⁷, ⁸, ⁹, ¹⁰, ¹¹ Importantly, the evidence base for the alleged increase in opioid prescriptions is not solid.

Recent studies have found conflicting evidence on whether relaxing NP SOP laws is associated with an increase in opioid prescriptions. A recent study by Alexander and Schnell reported that relaxing NP SOP laws increased opioid prescriptions.²⁴ While that study analyzed a data set that provided a nearly comprehensive picture of opioid prescribing, the authors’ primary measure of prescribing was at the county level. Additionally, the primary focus of the study was not opioid prescribing but mental health morbidity and mortality, which both declined following grants of independence. Conversely, a study by Hamilton found evidence that relaxing NP SOP laws decreased the use of prescription opioids by 9.8%.²⁵ The data analyzed, however, did not provide a comprehensive picture of opioid prescribing.²⁵

Two studies have examined a single year of prescriptions to Medicare beneficiaries. First, Ladd and colleagues analyzed Medicare data from 2013 and found evidence that NPs and physicians respond similarly to laws allowing NP independence.²⁶ They concluded that SOP laws were not an exclusive predictor of opioid prescribing.²⁶ Second, Lozada and colleauges analyzed Medicare data from 2015 and found that NPs practicing in states allowing independent prescribing were more likely to overprescribe opioids.²⁷ Because both of these studies involved cross‐sectional analyses of a single year of Medicare data, the authors' ability to isolate the association between NP independence and opioid prescribing from other confounders was limited.

Though not specific to SOP laws, recent research by Muench and colleagues concluded that NPs were less likely to prescribe an opioid but were more likely to prescribe stronger doses than physicians.²⁸ Relatedly, an analysis of treatment admissions for opioid abuse and opioid‐related deaths concluded that relaxing NP SOP laws increases the former and decreases the latter, but only in states requiring providers to access prescription drug monitoring programs (PDMPs).²⁹ While important, these studies do not directly address the association between NP SOP laws and opioid prescribing patterns.

This study provides new evidence on the association between NP SOP laws and opioid prescribing. Whereas prior studies have analyzed samples of prescription data or counts of opioid prescriptions at the county level,²⁴, ²⁵ this study analyzed a data set of approximately 1.5 billion opioid prescriptions associated with individual health care providers. Organizing these data by total annual morphine milligram equivalents (MMEs) prescribed by individual health care providers, this study examined changes in opioid prescribing by both physicians and NPs. This study included prescriptions to patients covered by commercial insurance, Medicare, Medicaid, and other government assistance, as well as those who paid cash for their prescriptions. The study findings can inform the ongoing state and national debates over NP SOP laws and may be relevant to the ongoing opioid epidemic as well.

This study does not suggest that blanket reductions in opioid prescriptions are unambiguously desirable. Opioids remain legal medications and are the most appropriate treatment in certain situations. Instead, this study aimed to address the question of whether allowing NPs to practice independently exacerbates the opioid crisis by increasing opioid prescription rates in general. The evidence developed here is directly relevant to this question but should not be interpreted as suggesting that a reduction in opioid prescriptions is always a positive development.

Methods

Classification of Scope‐of‐Practice Laws

NP SOP laws were classified as either “independent practice” or “restricted practice” based on prior work.⁷ The definition of independent practice used here is sometimes referred to as “full practice authority.” States that allowed NPs to both practice without any physician involvement (supervision or collaboration) and prescribe the same range of medications as physicians were classified as allowing independent practice. States either requiring physician supervision or collaboration or restricting the medications NPs can prescribe were classified as restricting the practices of NPs. Between 2011 and 2018—the period considered here—14 states changed their laws from restricted to independent practice: Maryland (2012), North Dakota (2012), Vermont (2012), Nevada (2014), Rhode Island (2014), Connecticut (2015), Minnesota (2015), Nebraska (2015), Delaware (2016), Utah (2016), West Virginia (2016), Illinois (2018), South Dakota (2018), and Virginia (2018). Of the remaining states, 13 (and the District of Columbia) allowed independent practice and 23 restricted the practices of NPs at all times between 2011 and 2018.

Data

Information on individual opioid prescriptions filled by patients at outpatient pharmacies between 2011 and 2018 came from Symphony Health's IDV® (Integrated Dataverse) data set. The data were collected from health insurance claims, nonretail invoices, and point‐of‐sale information obtained from individual pharmacies. The data set includes approximately 1.5 billion individual opioid prescriptions, which represent approximately 90% of all opioid prescriptions filled at outpatient pharmacies in the United States over the relevant time frame.

Each observation in the data set includes the following information about an individual prescription: the year the prescription was filled, the 11‐digit national drug code (NDC) for the prescription, the total days supply for the prescription, the quantity of drugs, an encrypted health care provider identifier, and the payer that covered the prescription. The provider identifier is encrypted but includes the provider's state of practice and the provider's taxonomy from the National Plan and Provider Enumeration System (NPPES). Providers were assigned to different SOP laws based on the listed state of practice in the NPPES and the year of the prescription. Payer categories include commercial insurance, cash, Medicare, Medicaid, other government assistance, and all payers combined. The data do not include information on specific commercial insurance plans, so all are included in a single “commercial” category.

From these raw data, a variable was constructed for the total annual MMEs that each provider prescribed to patients covered by each of the 6 payer categories. Total annual MMEs is the sum of the MMEs of all opioids prescribed by each provider in each year in each payer category. The MME of an individual opioid prescription is defined as:

$$\begin{array}{cc}& \mathrm{MME}\hfill \\ & =\frac{\left(\mathrm{Drug}\mathrm{Strength}\right)·\left(\mathrm{Drug}\mathrm{Quantity}\right)·\left(\mathrm{MME}\mathrm{Conversion}\mathrm{Factor}\right)}{\mathrm{Days}\mathrm{Supply}}\hfill \end{array}$$

Drug quantity and days supply came from the IDV® data set. The MME conversion factor and drug strength came from a data set compiled by the Prescription Drug Monitoring Program Training and Technical Assistance Center (PDMP TTAC). The PDMP TTAC data set is organized by 11‐digit NDCs. The IDV® and PDMP TTAC data sets were matched by NDCs prior to calculating MMEs. A logarithmic transformation was applied to the total annual MMEs because the MME data exhibit a substantial right skew. More details on this right skew and the construction of the primary independent variable of interest are available in the Online Appendix. The Online Appendix also reports additional models that do not rely on a logarithmic transformation, and these models yield similar results as the primary results reported here.

The primary analysis focuses on total annual MMEs as the outcome of interest because this measure of opioid prescriptions is most precise.³⁰ By converting all opioid prescriptions into a standardized measure, total annual MMEs accounts for both the volume of prescriptions as well as the strength of those prescriptions. Simply counting the number of prescriptions or number of pills may obscure actual prescribing behavior because certain medications are much more potent than others and because prescriptions can vary in duration.

However, to further test the association between NP independence and opioid prescriptions, several supplementary analyses, which are reported in the Online Appendix, focused on alternative measures of prescriptions. These measures include the following variables defined at the provider‐year level: the number of unique patients receiving opioids, the number of prescriptions written, the total days supply of opioids, an indicator for whether the provider prescribed any opioids, and a measure of variability in opioid prescribing. Analyses of these alternative variables were conducted using the same study design as for the primary outcome of interest and generally yield similar results as the primary analysis.

Study Design

To examine the effect of NP SOP laws on opioid prescriptions, a series of difference‐in‐differences models was estimated to exploit the staggered adoption of these laws over time. These models compare the change in the trends of opioid prescriptions in states that changed their laws to allow NPs to practice independently with the trends of opioid prescriptions in states that did not change their laws. Thus, these models compare states “treated” with a change in NP SOP laws to “control” states that experienced no such legal change.

The dependent variable in all models was the natural logarithm of total annual MMEs. The independent variable of interest in each model was an indicator for whether a provider practiced in a state that allowed independent NP practice. Each model also included separate indicator variables for whether a state maintained a mandatory‐access PDMP,²⁹, ³¹ allowed access to recreational cannabis,³², ³³, ³⁴ allowed access to medical cannabis,³⁰ and had passed legislation regulating pain clinics.³⁵ Prior work has found that laws governing cannabis access, PDMPs, and pain clinics meaningfully impact opioid prescriptions and opioid‐related deaths, so the primary models include controls for these policies. Models focusing on prescriptions to Medicaid beneficiaries also included an indicator variable for whether the relevant state had expanded Medicaid.

Every model included a full set of provider and year fixed effects—that is, indicator variables for each individual provider and each individual year. Provider fixed effects controlled for observed and unobserved characteristics of providers, and year fixed effects controlled for linear and nonlinear trends in opioid prescriptions over time. The provider fixed effects absorbed much of the heterogeneity present in opioid prescribing and allowed the models to isolate the role of NP independence from any idiosyncratic factors present at the provider level. For example, if the medical (or nursing) school a provider attended influences the provider's opioid prescribing behavior, the inclusion of provider fixed effects will control for this influence.³⁶ Similarly, the inclusion of provider fixed effects can control for the role of provider specialty. Controlling for provider specialty may be particularly important because physicians and NPs work in many different specialties with varying use of prescription opioids. The inclusion of fixed effects can net out the effect of specialty at the individual level and allow the models to estimate the general association between NP independence and opioid prescriptions across all specialties. In general, including provider fixed effects can control for any time‐invariant factor at the level of the individual provider that may influence opioid prescribing patterns. The year fixed effects can control for time‐varying policies that affect all providers. For example, these fixed effects controlled for the introduction of the Affordable Care Act (and the changes in federal regulation related to that act) in 2012 as well as the Centers for Disease Control and Prevention's introduction of its opioid prescribing guideline in 2016.

All models were restricted to physicians and NPs only, who were identified by NPPES taxonomy codes. For each payer category, three primary models were estimated: one including all providers, one including NPs only, and one including physicians only. The all‐provider models estimated the net association between NP SOP laws and opioid prescriptions across physicians and NPs. The models restricted to NPs and the models restricted to physicians estimated the associations specific to these provider types. Throughout the analysis, standard errors were clustered at the state level, and p‐values less than 0.05 were considered statistically significant. As described in detail in the Appendix, the data were tested for parallel trends, and the null hypothesis of parallel trends was not rejected, which supports the use of difference‐in‐differences models.

The Online Appendix also reports additional model specifications. These include models that examine the intensive and extensive margins of change. Because the primary outcome of interest is the number of MMEs prescribed, the primary models do not distinguish between a change in opioid prescriptions because some providers choose not to prescribe any opioids (the extensive margin) or because some providers choose to prescribe a different number or strength of opioids (the intensive margin). Models in the Online Appendix consider whether fewer providers prescribed any opioids following a grant of NP independence and whether, among those providers that prescribed at least one opioid, the amount of opioids prescribed changed.

All analysis was performed in SQL Server 2017 and Stata 14.2. This study was exempt from review by the institutional review board.

Results

Figure 1 reports the mean annual MMEs for NPs and physicians across six payer categories. More MMEs were prescribed to commercially insured and Medicare patients than others, and across all categories, physicians prescribed more MMEs than NPs. Further disaggregating prescribing patterns, Table 1 reports mean annual MMEs for NPs and physicians across payer categories and NP SOP law regimes. These regimes include states that restricted the practices of NPs throughout the entire study period, states that allowed independent practice throughout the entire study period, and states that switched from restricted to independent practice.

Figure 1.

Mean Annual Morphine Milligram Equivalents (MMEs) by Provider Type and Payer

Each bar represents the mean annual morphine milligram equivalents prescribed by either physicians or nurse practitioners to individuals whose prescription medications were covered by the given payer. Means are calculated across all years 2011‐2018. The commercial category includes all commercial insurance plans. The cash category includes all prescriptions that were paid for in cash (and not covered by one of the other payers). Error bars represent the 95% confidence interval for each calculated mean.

Table 1.

Mean Annual Morphine Milligram Equivalents (MME) by Provider Type, Payer, and Scope‐of‐Practice Law

	Assistance MME (SD)	Cash MME (SD)	Commercial MME (SD)	Medicaid MME (SD)	Medicare MME (SD)	All Payers MME (SD)
Restricted Practice
All Providers	512	490	3,917	739	3,121	7,162
	(5,634)	(2,637)	(21,890)	(4,630)	(17,224)	(39,731)
Physicians	562	536	4,319	788	3,386	7,834
	(6,072)	(2,800)	(23,185)	(4,687)	(18,099)	(41,902)
Nurse Practitioners	250	230	1,700	466	1,630	3,478
	(2,236)	(1,370)	(12,297)	(4,289)	(10,922)	(24,309)
Independent Practice
All Providers	544	470	3,832	759	2,869	6,946
	(4,582)	(3,170)	(25,225)	(4,415)	(14,408)	(41,088)
Physicians	550	493	3,949	753	2,835	7,017
	(4,752)	(3,306)	(25,926)	(4,187)	(13,746)	(41,003)
Nurse Practitioners	513	353	3,232	787	3,045	6,580
	(3,647)	(2,344)	(21,268)	(5,393)	(17,460)	(41,516)
Switched to Independence
All Providers	454	420	3,623	890	2,609	6,396
	(4,805)	(2,407)	(21,836)	(7,237)	(14,464)	(37,476)
Physicians	486	453	3,889	913	2,706	6,780
	(5,140)	(2,557)	(22,450)	(7,271)	(14,321)	(37,960)
Nurse Practitioners	297	251	2,285	771	2,106	4,486
	(2,573)	(1,389)	(18,381)	(7,066)	(15,170)	(34,905)

Each cell reports the mean annual morphine milligram equivalents (MMEs) prescribed by the type of provider listed in the left column to patients whose medications were covered by the payer listed across the top. The standard deviation of the mean is provided in parentheses. “All Providers” includes both nurse practitioners (NPs) and physicians but not other provider types. Mean annual MMEs are separately reported for providers who (1) restricted the practices of NPs for all years between 2011 and 2018, (2) allowed NPs to practice independently for all years between 2011 and 2018, and (3) switched from restricted to independent practice at some point between 2011 and 2018. Means are calculated across all years 2011–2018. The commercial category includes all commercial insurance plans. The cash category includes all prescriptions that were paid for in cash (and not covered by one of the other payers).

Physicians prescribe more MMEs than NPs across all payer categories in states that restrict NP practice and states that switched from restricted to independent practice. In states allowing independent practice, physicians similarly prescribe more MMEs than NPs across all payers, except Medicaid. Comparing prescribing patterns across NP SOP laws, physicians prescribe fewer and NPs prescribe more MMEs in states that allow NP independence relative to those that restrict NP practice.

To explore the association between NP SOP laws and the opioid prescribing patterns of NPs and physicians in more depth, Table 2 reports results from a series of regression models. Each row corresponds to a separate regression model for the indicated payer category and provider group. In the interest of succinctness, Table 2 reports only the coefficient on the indicator variable for NP independence. The final column reports the percentage change in MMEs associated with the relevant coefficient to provide clearer insight into the results. Full regression results are available in the Online Appendix.

Table 2.

Regression Results for the Association Between Nurse Practitioner (NP) Scope‐of‐Practice Laws and Annual Morphine Milligram Equivalents

	Coefficient	95% Confidence Interval		P‐Value	N	Percentage Change
Commercial
All Providers	−0.063	(−0.076, −0.050)		<0.001	7,442,076	−6.140
NPs	0.012	(−0.025, 0.049)		0.535	1,123,852	1.176
Physicians	−0.078	(−0.092, −0.064)		<0.001	6,318,224	−7.475
Cash‐Paying
All Providers	−0.020	(−0.031, −0.008)		0.001	4,992,657	−1.959
NPs	0.044	(0.011, 0.078)		0.009	737,225	4.548
Physicians	−0.032	(−0.044, −0.019)		<0.001	4,255,432	−3.116
Medicaid
All Providers	−0.014	(−0.028, 0.001)		0.074	3,778,756	−1.345
NPs	0.001	(−0.042, 0.044)		0.965	568,308	0.095
Physicians	−0.015	(−0.031, 0.001)		0.065	3,210,448	−1.472
Medicare
All Providers	−0.030	(−0.044, −0.016)		<0.001	5,713,794	−2.971
NPs	0.064	(0.023, 0.104)		0.002	840,954	6.557
Physicians	−0.046	(−0.061, −0.032)		<0.001	4,872,840	−4.507
Assistance
All Providers	−0.076	(−0.090, −0.062)		<0.001	3,656,153	−7.313
NPs	−0.016	(−0.054, 0.022)		0.414	572,676	−1.569
Physicians	−0.086	(−0.101, −0.072)		<0.001	3,083,477	−8.257
All Payers
All Providers	−0.048	(−0.062, −0.034)		<0.001	8,147,149	−4.688
NPs	0.044	(0.005, 0.082)		0.027	1,237,038	4.468
Physicians	−0.066	(−0.080, −0.051)		<0.001	6,910,111	−6.352

Each row reports results from a separate regression model. Only the coefficient on the indicator variable NP independence is reported. The 95% confidence intervals and p‐values are calculated based on standard errors clustered at the state level. The dependent variable in all models is the natural logarithm of total annual MMEs. Each regression model is limited to the provider type listed in the left column (with “All Providers” including physicians and nurse practitioners only), and each model includes only prescriptions written by that provider type that were covered by the payer listed across the top. Because all regression models are log‐linear with a dependent variable defined as the natural logarithm of total MMEs plus one, the coefficients are not directly interpretable. The final column reports the percentage change in MMEs associated with NP independence from each model.

Across all providers, Table 2 offers no statistically significant evidence that NP independence is associated with an increase in opioid prescriptions as measured by annual MMEs. Instead, for all payers except Medicaid, the results evince a statistically significant decline in opioid prescriptions associated with NP independence. Focusing first on the results for prescriptions to commercially insured patients, NP independence is associated with a 6.14% decrease in annual MMEs across all providers. The results also indicate a positive, but statistically insignificant, association between NP independence and NP‐prescribed opioids and a negative association between NP independence and physician‐prescribed opioids. The results for cash‐paying patients are similar. NP independence is associated with an overall decline of 1.96% in opioid prescriptions. NPs increase and physicians decrease their opioid prescriptions by 4.55% and 3.12%, respectively, when NPs are granted independence.

Turning to the government payers, prescriptions to Medicare and government‐assistance patients are 2.97% and 7.31% lower, respectively, across all providers when laws allowing NPs to practice independently are in place. For both payer categories, NP independence is associated with a statistically significant decrease in physician‐prescribed opioids. NP independence is associated with a statistically significant increase in NP‐prescribed opioids for Medicare patients but has no statistically significant association with NP‐prescribed opioids for government‐assistance patients.

NP independence is not associated with a statistically significant change in opioid prescriptions to Medicaid patients. The pattern of coefficients in the Medicaid‐specific models, however, is consistent with earlier results; annual MMEs decrease across all providers, increase for NPs, and decrease for physicians.

Across all payers, NP independence is associated with a decline in opioid prescriptions for all providers, an increase in opioid prescriptions for NPs, and a decrease in opioid prescriptions for physicians. All of these changes are statistically significant. Importantly, the models focusing on all payers can account for patients “substituting” between types of providers (i.e., changing from one provider type to the other) for some types of payers. The results for all payers demonstrate that, to the extent patients substitute in this way, NP independence remains associated with the same pattern of results seen for specific payers.

The Online Appendix reports additional results that are generally consistent with those from the primary models. Models replacing the natural logarithm of MMEs with the inverse hyperbolic sine of MMEs follow the same pattern observed in the primary models, with some small changes in coefficient magnitude and statistical significance. Models that can specifically accommodate count data (such as MMEs) also reveal a similar pattern of results. However, these models suggest a positive and statistically significant increase in opioid prescriptions to Medicaid patients following NP independence, consistent with the means reported in Table 1. Analysis of alternative measures of opioid prescriptions likewise yield similar results as those reported in the main analysis, with NP independence being associated with a statistically significant increase in opioid prescriptions to Medicaid patients for some alternative measures.

Finally, the Online Appendix reports models that focus explicitly on the extensive margin (whether providers prescribe any opioids) and intensive margin (the amount of opioids prescribed conditional on prescribing any opioids). These models demonstrate that NP independence is associated with changes in both margins and that these changes largely track the changes observed in the primary models. With some limited exceptions (including the Medicaid population), physicians are less likely to prescribe any opioids and prescribe fewer opioids following NP independence. NPs are both more likely to prescribe opioids and prescribe more opioids following a grant of independence. Across all providers, the probability of prescribing opioids declines slightly, and the amount of opioids prescribed (conditional on prescribing any opioids) decreases.

Discussion

Overall, the primary analysis revealed no evidence that opioid prescriptions, as measured by annual MMEs, increase when states allow NPs to practice independently. Instead, allowing independent NP practice is associated with a statistically significant decrease in opioid prescriptions across physicians and NPs, except among Medicaid beneficiaries. This net decrease stems from a combination of changes in NP and physician opioid prescribing. In independent‐practice states, NPs generally prescribe more opioids than in restricted‐practice states. Meanwhile, physicians prescribe more opioids in restricted‐practice states and less in independence states.

These prescribing patterns are broadly consistent with expected changes in health care delivery following a law allowing independent NP practice.²⁵ Prior work has categorized these expected changes into two general effects: the “substitution effect” and the “access effect.”²⁵ When NPs can practice independently, they are better able to compete with physicians and serve the needs of patients, which can result in both substitutions between providers and a general increase in access to care.¹¹, ¹³

The substitution effect describes a scenario in which patients choose to seek care from NPs instead of physicians when NPs can practice independently. As some patients move to NP‐supplied care, NPs should prescribe more opioids to meet patient needs. As patients substitute NPs for physicians once a state allows NP independence, physicians should prescribe relatively fewer MMEs each year. If NPs and physicians prescribe opioids in the same way, the net change in overall opioid prescriptions should be zero. If, as prior work has found, however, NPs treat similar patients less intensively by performing fewer procedures and prescribing fewer medications,⁸, ⁹, ¹⁰, ³⁷, ³⁸ the substitution of NPs for physicians should result in fewer opioid prescriptions, all else equal. The net result should be fewer opioid prescriptions because the decrease in physician‐prescribed opioids should be larger than the increase in NP‐prescribed opioids.

The access effect refers to a scenario in which patients who previously could not access care are able to do so following a grant of independence.²⁵ They may either obtain care from newly independent NPs or from physicians who have greater capacity as NPs absorb some of their patients via the substitution effect. The access effect suggests that opioid prescriptions should increase across all providers generally because they are treating more patients. Whether the increase in opioid prescriptions stemming from the access effect and the (anticipated) decrease in opioid prescriptions stemming from the substitution effect result in an overall decrease or increase in opioid prescriptions is not clear ex ante.

The evidence developed in the current study, however, suggests that the net change in opioid prescriptions associated with NP independence is negative, except perhaps among Medicaid beneficiaries. In other words, the evidence is consistent with the (negative) magnitude of the substitution effect being larger than the (positive) magnitude of the access effect. The net decrease in opioid prescriptions observed across all NPs and physicians following a grant of NP independence is consistent with prior research, which has found evidence that NPs generally provide less intensive care than physicians.⁸, ⁹, ¹⁰, ³⁷, ³⁸

With respect to Medicaid beneficiaries, the primary results suggest that the substitution and access effects may be in relative equipoise, as the association between NP independence and opioid prescriptions is not statistically significant. Results from alternative models, which are reported in the Online Appendix, suggest that the access effect may be larger than the substitution effect among Medicaid beneficiaries, as some measures of opioid prescriptions increase among Medicaid beneficiaries following a grant of NP independence. A larger access effect among the Medicaid population compared to other populations is consistent with prior work.¹⁴, ³⁹ Multiple studies have found that NPs provide care to underserved populations, including Medicaid beneficiaries, at higher rates than physicians.⁹, ¹⁴, ³⁹ Once granted independence (and the greater ability to provide care that accompanies independence), this tendency of NPs to treat Medicaid beneficiaries at higher rates would naturally translate into a larger access effect with respect to opioid prescriptions to Medicaid patients. As noted in Table 1, NPs prescribe more opioids to Medicaid beneficiaries than do physicians in states that always allowed independent practice, and the results from the primary and supplementary models further suggest that NP independence tends to increase access (and thereby increase opioid prescriptions) among Medicaid patients than among other patient populations. While the results in this study cannot establish this explanation definitively, it is consistent with prior work and warrants investigation in future work.

Overall, the net decrease in opioid prescriptions seen in conjunction with NP independence can generate important public health benefits. It is important, however, to note that opioids remain legal and are the best available and most appropriate treatment for certain conditions. Indeed, policies directed at reducing opioid prescriptions, such as prescription drug monitoring programs, have been criticized for potentially “overcorrecting” by denying patients access to important medications.⁴⁰ Although granting NPs independence is associated with a decrease in opioid prescriptions, the overcorrection problem may not be as relevant as in other policy domains because of the mechanism by which the decrease occurs. As opposed to programs directly aimed at reducing opioid prescriptions, granting NPs independence may result in a more targeted reduction in opioid prescriptions. NP independence can facilitate additional time with providers and allow patients to discuss and explore alternative options for their health care needs instead of facing a blunt reduction in opioid prescriptions.

Additionally, the magnitudes of the coefficients are not consistent with NP independence resulting in overcorrection. Prior work has estimated that providers overprescribe opioids by approximately 20%.⁴¹ The results detailed here demonstrate that NP independence is associated with between a 1% and 8% decrease in opioid prescriptions. These results do not rule out the possibility that some patients may not receive opioids when needed, but they also are not consistent with a problematic overcorrection in general.

The net decrease in overall MMEs does not support the concern that allowing NPs to practice independently will contribute to the opioid epidemic. Accordingly, the study results support granting NPs the authority to practice independently. Recent work has demonstrated that the supply of NPs will continue to grow faster than the supply of physicians in the coming years.⁴², ⁴³, ⁴⁴, ⁴⁵, ⁴⁶ Therefore, understanding how NP independence may impact opioid prescriptions will be important for policymakers as they continue to weigh various regulatory options for NPs. The results of this study can inform those deliberations in the context of a growing NP workforce.

Limitations

This study had several limitations. First, this study lacked information on patient diagnoses and therefore did not examine whether providers prescribed opioids appropriately based on a given patient's condition. Second, and relatedly, the results of this study do not demonstrate that any provider or group of providers over‐ or underprescribes opioids more generally. Such a conclusion would require information on patient diagnoses, which were not available in this study. Third, this study did not account for the potential impact of all possible changes in local, state, insurer, or facility prescribing policies over time.

Fourth, although the models in this study controlled for the effect of provider specialty on opioid prescribing patterns, they did not separately provide estimates of the association between NP independence and opioid prescribing by specialty. Fifth, the study did not consider all possible measures of opioid prescriptions. MMEs is a useful measure of opioid prescriptions because it accounts for both volume of prescriptions (e.g., number of pills) and strength of those prescriptions. However, other measures may also be relevant. The Online Appendix includes results for many other measures of opioid prescriptions, and the results largely track the results for MMEs. However, the study did not consider measures of long‐term opioid use or other potentially relevant opioid measures.

Conclusions

The results of this study suggest that allowing NPs to practice independently is associated with an overall decrease in opioid prescriptions (as measured by total annual MMEs) across physicians and NPs. This evidence does not support concerns that allowing independent NP practice will contribute to increasing opioid prescriptions. Instead, it suggests that NP independence is associated with an opposite change in opioid prescriptions.

Funding/Support: None.

Conflict of Interest Disclosure: The author completed the ICMJE Form for Disclosure of Potential Conflicts of Interest. No conflicts were reported.

Supporting information

Figure S1: Distribution of MMEs (Conditional on MMEs being positive) for Commercially Insured Patients

Figure S2: Parallel Trends for Commercial Insurance

Table S1: Regression Results for the Association Between Nurse‐Practitioner Scope‐of‐Practice Laws and Annual Morphine Milligram Equivalents

Table S2: Regression Results for the Association Between Nurse‐Practitioner Scope‐of‐Practice Laws and Annual Morphine Milligram Equivalents With the Inverse Hyperbolic Sine Transformation

Table S3: Pseudo‐Poisson Maximum Likelihood Results for the Association Between Nurse‐Practitioner Scope‐of‐Practice Laws and Annual Morphine Milligram Equivalents

Table S4: Random Effects Models Results for the Association Between Nurse‐Practitioner Scope‐of‐Practice Laws and Annual Morphine Milligram Equivalents

Table S5: Regression Results for the Association Between Nurse‐Practitioner Scope‐of‐Practice Laws and the Variability of Opioid Prescriptions

Table S6: Regression Results for the Association Between Nurse‐Practitioner Scope‐of‐Practice Laws and Number of Opioid Patients

Table S7: Regression Results for the Association Between Nurse‐Practitioner Scope‐of‐Practice Laws and Number of Opioid Prescriptions

Table S8: Regression Results for the Association Between Nurse‐Practitioner Scope‐of‐Practice Laws and Total Days Supply of Opioids

Table S9: Regression Results for the Association Between Nurse‐Practitioner Scope‐of‐Practice Laws and the Probability a Provider Prescribes Opioids

Table S10: Regression Results for the Association Between Nurse‐Practitioner Scope‐of‐Practice Laws and Annual Morphine Milligram Equivalents Conditional on a Positive Amount Prescribed

Table S11: Regression Results for the Association Between Nurse‐Practitioner Scope‐of‐Practice Laws and Annual Morphine Milligram Equivalents with State‐Specific Linear Time Trends

Table S12: Parallel Trends Tests