Labor market outcomes of granting full professional independence to nurse practitioners

Abstract

Faced with an impending shortage of physicians, the U.S. healthcare industry’s capability and capacity to meet its growing clientele’s needs rely on finding reliable alternative primary care providers. This study examines how professional independence granted to nurse practitioners could change their labor market outcomes and help alleviate the shortage crisis. The study finds that full scope-of-practice laws increase nursing practitioners’ working time allocation, particularly in administrative and consultation tasks. This study’s findings clarify how full scope-of-practice could enhance nurse practitioners’ involvement and professional contributions in healthcare industry, especially benefitting regions and populations that are underserved.

Appendix

1.1 Map of full scope-of-practice law adopting states

See Fig. 4.

Fig. 4

U.S. Map of full and restrictive SOP laws adoption

1.2 Robustness check

This section presents the results of robustness checks.

1.2.1 Difference-in-differences point estimates

Results in Table

Table 5 The results of full scope-of-practice laws on the labor market outcomes of nurse practitioners

5 summarize the labor market outcomes of nurse practitioners after the adoption of full SOP laws. Panels A contain results for the number of hours worked per week. The estimation in column (1) controls for state-specific time trends to address possible correlation arising from coincidences between the introduction of SOP laws and trends in control states that may also affect labor supply decisions of nurse practitioners. In addition, the inclusion of state-specific linear time trend would address concerns that some temporal gaps in the data could be driving the results. Results in column (1) show that the impact of full SOP laws increases nurse practitioners’ worktime allocation by 2.78 h per week. These results are significant at the 1% level. As can be seen from column (2) that excludes state-specific time trend, granting practice and prescription independence to nurse practitioners would increase their working hours by about 3.1 per week. The estimates in both columns are consistent as a t-statistic value of 0.414 indicates that they are not statistically different from each other. Angrist and Pischke (2008) point out that consistent estimates by regressions with and without state-specific time trend suggest robust estimates.

Specification check results in columns (3) and (6) also provide robust evidence supporting the conclusion of an increased worktime trend. Solon et al. (2015) suggest reporting both weighted and unweighted estimates because consistent parameters of these two indicate a correctly specified model. In this analysis, the estimates in columns (3), (5), and (6) obtained from regressions without the survey’s sample weights slightly change compared to the estimates in columns (1) and (2), but they are still broadly in line with the conclusions.

1.2.2 Alternative definition of full SOP laws

We use an alternative definition of full SOP laws suggested by McMichael and Markowitz (2020) to test the consistency of the results using different treatment variables. McMichael and Markowitz (2020) point out that many SOP studies employ different categorizations of the SOP laws and the sources for the laws are also different. As a result, it would cause difficulties in comparing results across studies. To clarify this issue, McMichael and Markowitz (2020) provide a definition of full SOP laws, namely “full practice authority” (FPA), to serve as a reference for relevant studies that examine SOP laws. We redefine the treatment variable of full SOP laws following their suggestion and the results are presented in Table

Table 6 The results of full scope-of-practice laws on the labor market outcomes of nurse practitioners using alternative definition of treatment variable.

6. As the results indicate, the estimates shown in column (1) are similar to our main findings. The estimates in column (2) show the impact of full SOP laws between the years 1998 and 2008 using the alternative treatment. We use 1998 as the sample cut off point as suggested by McMichael and Markowitz (2020), who content that Medicare and many private insurers only reimburse nurse practitioners for health care services incident to physicians until 1998 and as a result, nurse practitioners could be tethered to physicians before that. The results in column (2) show the impact of full SOP laws that were adopted after 1998, which are effectively enacted at the state level. Based on the results, the estimate of 2.86 indicates that full SOP laws increase the nurse practitioners’ working hours per week. One the other hand, we examine the impact of full SOP laws enacted (following the passage of the Balanced Budget Act of 1997 which allows direct reimbursement to nurse practitioners) in 1998 at the state level in column (3). These results are still in line with our main findings.

1.2.3 Bacon decomposition

As the adoption of full SOP laws was gradually realized across states over the years, in addition to the comparison of treated states to untreated states (never adoption states), DID also compares states that adopted full SOP laws in earlier years to these that adopted full SOP laws in later years. For instance, Wyoming adopted full SOP laws in 1993 and it became a treated state since then. As a result, DID compares the labor market outcomes of nurse practitioners in Wyoming to the states that never adopt full SOP laws and also to the states that would adopt the laws in 2000s. On the other hand, the outcomes of states such as Idaho that adopted the laws in 2004 can be compared to the states that adopted full SOP laws in 1990s (these states are therefore regarded as control states). Because the impact of full SOP laws adoption among earlier adopting states may persist for years and also change over the years, a comparison between earlier adopting states as control and later adopting states as treatment may introduce bias to estimation (Goodman-Bacon, 2018). To check if the estimation is biased due to the mentioned issue, Goodman-Bacon (2018) suggest we can examine the sources of model identification and check the weights that are assigned by DID to each type of comparisons (1. Earlier Treated vs. Later Treated; 2. Later Treated vs. Earlier Treated; 3. Ever Treated vs. Never Treated). A higher weight assigned to “Later Treated as treatment vs. Earlier Treated as control” indicate a higher likelihood of biased DID estimates. The estimates of bacon decomposition are presented in Table

Table 7 Bacon decomposition estimates of Difference-in-Difference model

7.

As can be seen, the estimates of full SOP impact are decomposed into three categories for labor market outcomes of nurse practitioners. In general, the category of “Ever Treated as T vs. Never Treated as C” has the highest weight at 0.904 and the other two categories have weights at 0.043 and 0.054, respectively, suggesting little biasing impacts were introduced by the comparison between earlier and later treated states. The results in Table 7 suggest our main findings are consistent. Figure 5 shows graphical evidence of source of identification from different comparison groups. As can be seen that the estimates of comparisons between earlier and later groups (cross signs) are assigned weights that are very close to zero but the estimates of comparisons between ever and never adopted states (triangle signs) in general have high weights. Figure 5 shows that the estimated impact of full SOP is based on nonbiased comparison between ever and never adoption states but not from other two comparison groups.

This study employs the group of registered nurses as another comparison group to conduct a triple-difference estimation. We introduce interaction terms between full SOP laws indicator and nurse practitioners and the specification of the triple-difference model is constructed as follows:

$$ Outcome_{ist} = \alpha + \gamma_{1} SOP_{st} *Nursepractitioner_{ist} + \gamma_{2} Nursepractitioner_{ist} + \gamma_{3} SOP_{st} + \beta_{2} X_{ist}^{^{\prime}} + \beta_{3} \delta_{t} + \beta_{4} \gamma_{s} + \varepsilon_{ist} $$

(3)

$Outcome_{ist}$ is the dependent variables of labor market outcomes. In addition to the covariates that are controlled for in main Eq. (1), $X_{ist}^{^{\prime}}$ includes control variables such as state-nurse practitioner fixed effect and year-nurse practitioner fixed effect. The year- nurse practitioner fixed effect controls for changes that could have different impacts on the nurse practitioner and registered nurses populations. The state- nurse practitioner fixed effect controls for the time-invariant differences among states that could affect labor market outcomes. Robust standard errors are clustered at the state level. The results of triple-difference model are reported in Table

Table 8 Estimates of full SOP impact on registered nurses and triple-difference estimates.

8.

Results in panel A show the impact of full SOP laws on the labor market outcomes of registered nurses. The estimates show that full SOP laws has no impact on registered nurses, which validate the group of registered nurses as another control group.

Estimates in panel B show the triple-difference results for the impact of full SOP laws on nurse practitioners. The results suggest that after the adoption of full SOP laws, nurse practitioners would increase their working hours per week. Meanwhile, there is no change in their wage level in the wake of full SOP adoption. The findings are consistent with and further support the estimates of our main DID findings.

1.2.4 Migration response of nurse practitioners

Adams and Markowitz (2018) suggest that restrictive SOP laws would lead nurse practitioners to move to states with favorable practice environment. Traczynski and Udalova (2018) express similar concerns that they cannot rule out the possibility of nurse practitioners moving into a state allowing expanded SOP. However, Traczynski and Udalova (2018) also point out that such movement may be minimal considering the findings of Timmons (2013) on the low mobility rates among registered nurses even when states relax barriers to practice across states in the wake of the Nurse Licensure Compacts reform.

This study is motivated by two reasons to examine the movement of nurse practitioners across states. First, examining the mobility of nurse practitioner across states could further complement our findings of the labor supply outcomes of nurse practitioners in response to full SOP. Second, if there is indeed a large-scale movement across states, the estimates of working time in this study may be either overestimated or underestimated, depending on the direction of the movement. Therefore, a check on the movement tendencies of nurse practitioners would corroborate our main results.

We use information from the NSSRN pertaining to the survey question asking nurse practitioners about their geographical state of employment during the year of survey and in the previous year. The specification is shown in the following Eq. (4).

$$ Migration_{ist} = \alpha + \beta_{1} Full\_SOP_{ist} + \beta_{2} X_{ist}^{^{\prime}} + \gamma_{t} + \delta_{s} + \varepsilon_{ist} $$

(4)

where $Migration_{ist}$ is a binary outcome that equals one if a nurse practitioner indicate his/her employment state in previous year is different from the current practice state; zero otherwise. $Full\_SOP_{ist}$ equals one if a state grants professional independence to nurse practitioners, zero if otherwise. $X_{ist}^{^{\prime}}$ is a vector of covariates. $\gamma_{t}$ and $\delta_{s}$ are year and state fixed effects, respectively. $\varepsilon_{ist}$ is the error term. Robust standard errors are clustered at the state level. Results from Eq. (4) provide evidence showing if a full SOP state is more likely to have nurse practitioners who migrated from other states.

The linear probability model coefficients of the full SOP indicator indicate the impacts of full SOP on the migration of a nurse practitioner. Moreover, to reveal if full SOP encourage nurse practitioners with high motivations (e.g., nurse practitioners with better ability or those working in rural areas) to migrate to nonrestrictive states, we introduce interaction terms between full SOP laws and rural status. Skill level is measured by the experience. The results are reported in Table 9.

Results in Table 9 for different specifications point to the same conclusion that the adoption of full SOP does not affect the migration likelihood of nurse practitioners. The results indicate that there is no statistically significant migration tendency among nurse practitioners to migrate to adopting states after the adoption of full SOP law. Moreover, we also find little migratory responses among nurse practitioners who are more experienced or serving in rural areas. The summary statistics show that over 70% of the nurse practitioners are married, thus inter-state migration decisions are usually made as a family decision and determined by other factors not solely linked to the favorability of the practice environment in full SOP adopting states. However, we would like to mention that due to data limitations, we cannot observe nurse practitioners who migrated into full SOP states beyond the survey’s time period. Moreover, we could neither observe the long-term migratory responses of nurse practitioners nor identify where nurse practitioners migrated from. This could be delegated to future research when relevant data are available.

1.2.5 Exclude partial SOP adopting states

We include controls for partial SOP laws in the regressions and find the results to be consistent. This robustness check conducts a sample test by excluding the states that adopted partial SOP laws such as granting only practice or prescription independence from the control group. The results are presented in Table

Table 9 The impact of full SOP laws on the migration decisions of nurse practitioners

10 and as can be seen that the robustness check results are consistent with the main findings that the full SOP laws would increase the working time of nurse practitioners but not their wages.

1.3 Event study estimates

Please see Figs. 6, 7, 8 and 9.

Fig. 5

Estimates of Bacon decomposition

Fig. 6

The heterogenous event study estimates of full SOP laws impact on working hours of nurse practitioners

Fig. 7

The heterogenous event study estimates of full SOP laws impact on hourly wage of nurse practitioners

Fig. 8

The event study estimates of full SOP laws impact on employment settings of nurse practitioners

Fig. 9

The event study estimates of full SOP laws impact on task choices of nurse practitioners

1.4 Model specifications of employment setting and task choices

1.4.1 Employment setting

The specification of the regression examining the impact of full SOP laws on employment setting choices of nurse practitioners is constructed as follows:

$$ Emp\_setting_{ist} = \alpha + \beta_{1} Full\_SOP_{ist} + \beta_{2} X_{ist}^{^{\prime}} + \gamma_{t} + \delta_{s} + \varepsilon_{ist} $$

(5)

where $Emp\_setting_{ist}$ is a categorical outcome shows employment settings of nurse practitioners including hospital, community health center, ambulatory care agency, and other agencies. $Full\_SOP_{ist}$ equals one if a state grants professional independence to nurse practitioners without requiring physicians’ supervision or collaboration, zero if otherwise. $X_{ist}^{^{\prime}}$ is a vector of covariates. $\gamma_{t}$ and $\delta_{s}$ are year and state fixed effects, respectively. $\varepsilon_{ist}$ is the error term. Robust standard errors are clustered at the state level. Due to the categorical nature of the employment setting variable, we use multinomial logit regression combining with DID model.

1.5 Task

We ran a regression for the number of hours of each task using the model specified as follows:

$$ Task\_hours_{ist} = \alpha + \beta_{1} Full\_SOP_{ist} + \beta_{2} X_{ist}^{^{\prime}} + \gamma_{t} + \delta_{s} + \varepsilon_{ist} $$

(6)

where $Task\_hours_{ist}$ is continuous outcomes showing the number of hours devoted to each task per week by a nurse practitioner. $Full\_SOP_{ist}$ equals one if a state grants professional independence to nurse practitioners, zero if otherwise. $X_{ist}^{^{\prime}}$ is a vector of covariates. $\gamma_{t}$ and $\delta_{s}$ are year and state fixed effects, respectively. $\varepsilon_{ist}$ is the error term. Robust standard errors are clustered at the state level. Because the assignment of time to different tasks by a nurse practitioner could be correlated, we employ seemingly unrelated regressions to explore the decisions of hours assigned to each task by nurse practitioner.