Popular in Social Sciences

Books

City Time: On Being Sentenced to Rikers Island

Independent Scholar David A. Campbell

Culture and Society

The City as a Growth Machine: Toward a Political Economy of Place

New York University Harvey Moltoch

Economics

Gender and Economics

University of California Santa Barbara Shelly Lundberg

American Politics

Keeping the Republic

Boston College Marc Landy

Banks and Policy Organisations

Greening the financial sector

European Investment Bank Matteo Ferrazzi

British Politics

Entrepreneurial thinking: the politics and practice of policy impact

University of Sheffield Matthew Flinders

Culture and Society

Carceral Entrapments: Views from the Prison/Street Interface in India

Indian Institute of Technology Delhi Mahuya Bandyopadhya

The Productivity of Professions

View more insights in Social Sciences - Economics - Health Economics

View more figures from University of Illinois at Chicago

View the page of Yiqun Chen

Read the Study

Yiqun Chen discusses the productivity of nurse practitioners (NPs) and physicians, two professions performing overlapping tasks but with stark differences in background, training, and pay.

Image courtesy of interviewee. December 7, 2023

Generating Workspace

Export as PDF

The paper examines the productivity of physicians and nurse practitioners (NPs) in the U.S. healthcare system, where there is rising demand for healthcare services but a relatively fixed supply of physicians. This video video context has increased the reliance on NPs as substitutes for physicians, particularly in emergency department settings. The study utilizes data from the Veterans Health Administration, which in 2016 allowed NPs to practice independently.

The findings reveal that NPs generally treat healthier patients compared to physicians, which complicates direct comparisons of their outcomes. To address this, the research employs quasi-random variation in the number of NPs on duty to assess treatment outcomes and resource use. The results indicate that while NPs incur higher medical costs for patients, they often achieve poorer treatment outcomes on average. Nonetheless, there is considerable variability in productivity within both occupations, with some NPs outperforming average physicians.

The paper suggests that experience may contribute to this productivity gap, as NPs with more experience tend to perform better. The findings recommend that healthcare systems should focus on optimal patient assignments between physicians and NPs, especially for less complex cases where NPs can perform comparably or even better than physicians. Additionally, it emphasizes exploring collaboration between the two roles to maximize patient care efficiency. Overall, the study contributes to ongoing discussions around the roles of NPs in healthcare delivery and the impact of occupational licensing on professional quality.

Language:

Export as PDF

X-TIMESTAMP-MAP:LOCAL=00:00:00.000,MPEGTS=900000

00:00.105 --> 00:03.925
In this paper, we look at the productivity of,

00:04.055 --> 00:07.125
physicians and, nurse practitioners.

00:07.665 --> 00:11.245
Um, in the US there's a increasing demand for healthcare,

00:11.905 --> 00:15.285
but physician supply is relatively inelastic.

00:15.785 --> 00:19.965
So this leads to the increased use of nurse practitioners

00:20.185 --> 00:23.885
as substitute for physicians, nurse practitioners,

00:23.885 --> 00:26.605
they are type of mid-level practitioners.

00:27.195 --> 00:29.645
They are trained to diagnose patients

00:29.825 --> 00:31.445
and also treat patients.

00:32.025 --> 00:34.045
And in this paper we look at,

00:34.145 --> 00:36.365
in the emergency department setting,

00:36.495 --> 00:38.165
which is an important setting,

00:38.195 --> 00:41.485
that is nurse practitioners whether there are differences

00:41.635 --> 00:42.925
between physicians

00:43.425 --> 00:47.765
and nurse practitioners in the, medical resource use

00:47.825 --> 00:49.325
and patient treatment outcomes,

00:49.915 --> 00:52.085
whether there are difference in these dimensions.

00:52.665 --> 00:54.725
So we look at this question at the

00:55.285 --> 00:59.445
Veterans Health Administration in the us, the VA,

00:59.495 --> 01:03.645
which is often just called va, the va, allow,

01:03.645 --> 01:07.125
they have a policy, they passed a policy in 2016

01:07.125 --> 01:09.045
that allowed nurse practitioners

01:09.425 --> 01:13.525
to practice independently without physician supervision.

01:14.055 --> 01:17.245
Nurse practitioners can, for example, diagnose patients,

01:17.295 --> 01:18.325
order treatments

01:18.425 --> 01:21.205
and prescribe drugs as independently

01:21.265 --> 01:23.085
as physicians at the va.

01:23.465 --> 01:25.405
The data, we are using the VA's

01:25.405 --> 01:26.925
Administrative Health records.

01:27.545 --> 01:30.205
So basically it record the patients we the

01:30.465 --> 01:31.645
who treat the patient,

01:31.985 --> 01:34.205
and, all the treatments the patients

01:34.605 --> 01:35.765
received during the visit.

01:36.385 --> 01:39.725
In the data, we can track whether the pa the provider treats

01:39.725 --> 01:41.325
the patient, is a nurse practitioner

01:41.425 --> 01:43.885
or physician, what, what type of,

01:43.885 --> 01:45.805
treatments the patient, received.

01:46.105 --> 01:48.725
And we can also track patient outcomes like,

01:48.725 --> 01:51.325
their mortality and hospital readmissions.

01:51.465 --> 01:54.845
Uh, our identification is the instrumental variable.

01:55.725 --> 01:58.685
H we use the, number of,

01:58.695 --> 02:01.845
nurse practitioners on duty at the ED on a day.

02:01.945 --> 02:04.365
The patient visited as an instrument

02:04.545 --> 02:06.565
for whether the patient is treated

02:06.665 --> 02:07.965
by a nurse practitioner or not.

02:08.545 --> 02:11.685
So if we, compared the average characteristics

02:11.825 --> 02:13.085
of patients treated

02:13.145 --> 02:16.485
by nurse practitioners versus the average characteristics

02:16.545 --> 02:19.365
of patients treated by physicians, we see

02:19.365 --> 02:22.165
that nurse practitioners on average treat healthier

02:22.445 --> 02:23.485
patients than physicians.

02:24.145 --> 02:26.285
For example, patients treated

02:26.345 --> 02:28.365
by nurse practitioners are younger

02:28.665 --> 02:30.325
and have fewer comorbidities.

02:31.035 --> 02:34.725
This suggests that we cannot simply compare, patient

02:35.405 --> 02:36.525
outcomes of patients treated

02:36.585 --> 02:38.765
by physicians versus nurse practitioners

02:38.945 --> 02:40.565
to see their productivity differences.

02:41.225 --> 02:44.725
We therefore uses quasi random variation in the number

02:44.745 --> 02:47.645
of nurse petitioners on duty at ed,

02:47.705 --> 02:49.325
on a data patient visited.

02:50.465 --> 02:54.445
So the, the eds in our setting, they use both physicians

02:54.625 --> 02:56.925
and nurse practitioners to treat patient,

02:57.465 --> 02:59.125
and they use physicians

02:59.305 --> 03:01.925
And nurse practitioners as substitutes.

03:02.625 --> 03:04.685
On days when there are more nurse practitioners,

03:04.685 --> 03:05.885
there are fewer physicians.

03:06.065 --> 03:07.885
And when there are fewer nurse practitioners,

03:07.885 --> 03:08.885
there are more physicians.

03:09.795 --> 03:12.365
Also, because of the institutional features

03:12.505 --> 03:14.445
of providers on duty scheduling,

03:14.695 --> 03:17.845
there is quasi random variation in a number

03:17.905 --> 03:19.565
of nurse petitioners on duty.

03:20.425 --> 03:23.405
But patient probability of being treated

03:23.505 --> 03:25.965
by nurse petitioners increases with the number

03:25.985 --> 03:27.405
of nurse practitioner on duty.

03:28.145 --> 03:31.045
So we use the number of nurse petitioners on duty

03:31.225 --> 03:34.845
as the ED on the day the patient visited as instrument

03:35.025 --> 03:38.045
for whether the patient is treated by a nurse practitioner.

03:38.745 --> 03:42.885
We show that conditional on the baseline controls like hos,

03:42.885 --> 03:47.485
ed by day of the week, ed by month, and ed by year.

03:47.485 --> 03:48.845
Fixed effects,

03:48.845 --> 03:51.925
patient eristics are very well balanced across the number

03:51.925 --> 03:53.605
of nurse practitioners on duty.

03:54.225 --> 03:56.805
We found that on average,

03:56.895 --> 04:00.725
there are productivity differences between physicians

04:00.865 --> 04:02.165
and nurse practitioners.

04:02.905 --> 04:05.005
On average nurse practitioners,

04:05.835 --> 04:08.045
make patients spend more on healthcare

04:09.015 --> 04:11.275
but get worse treatment outcomes.

04:12.735 --> 04:13.955
But we also find

04:13.985 --> 04:17.195
that there's very large productivity variation within

04:17.385 --> 04:18.795
each, each profession.

04:19.845 --> 04:22.875
Among the physicians, there are also low quality physicians,

04:23.015 --> 04:24.635
and among the nurse practitioners,

04:24.635 --> 04:26.835
there are also high quality nurse practitioners.

04:27.185 --> 04:29.355
There's very large productivity overlap

04:29.355 --> 04:30.755
between the two professions,

04:31.015 --> 04:34.235
and some nurse practitioners can perform even better than

04:34.235 --> 04:35.315
an average physician.

04:36.455 --> 04:39.195
Uh, for the mechanisms, we found some,

04:40.045 --> 04:43.395
mechanisms suggesting differences in human capital

04:43.665 --> 04:45.115
between the two professions.

04:45.895 --> 04:47.875
On average, nurse practitioners,

04:47.905 --> 04:50.395
have shown lower productivity than physicians,

04:51.055 --> 04:54.315
and we found that, experience might play a role here.

04:54.985 --> 04:58.955
When providers experience that, is higher, the nurse,

04:59.015 --> 05:02.035
nurse practitioners, physicians' performance gap become,

05:02.135 --> 05:04.075
lower in some outcomes.

05:05.455 --> 05:08.515
So we found that nurse practitioners tend to be more likely

05:08.655 --> 05:12.915
to use, treatments like, dag diagnostic tests

05:13.535 --> 05:16.115
and also like consults, which may help

05:16.145 --> 05:18.155
with their treatments and a diagnosis.

05:18.705 --> 05:21.675
This may also contribute to the higher medical spending

05:21.815 --> 05:22.875
of nurse practitioners.

05:23.895 --> 05:26.875
Um, we also find that the performance difference

05:26.875 --> 05:28.035
between physicians

05:28.095 --> 05:29.915
and nurse practitioners,

05:29.975 --> 05:31.875
become higher when the patient's

05:31.875 --> 05:33.235
conditions are more complex.

05:34.025 --> 05:37.075
This suggests that it could be important to think about

05:37.075 --> 05:39.315
what could be the optimal patient assignments

05:39.315 --> 05:41.515
between physicians and nurse practitioners.

05:41.835 --> 05:44.395
I think there are a couple related res.

05:45.175 --> 05:48.755
One, directly related ation will be the role

05:48.775 --> 05:51.835
of nurse practitioners in, healthcare delivery.

05:52.525 --> 05:55.875
Given that the nurse practitioners are increasingly used in

05:55.875 --> 05:56.915
the US healthcare system,

05:57.175 --> 06:01.205
There is a growing number of papers looking at the role

06:01.225 --> 06:03.205
of nurse petitioners in expanding

06:03.205 --> 06:04.605
healthcare supply capacity.

06:05.745 --> 06:08.205
In addition to looking at the effects

06:08.265 --> 06:11.125
of nurse petitioners on expanding healthcare,

06:11.125 --> 06:14.365
supply capacity, I think it's also very important for us

06:14.425 --> 06:18.285
to understand the productivity of nurse petitioners.

06:19.925 --> 06:23.525
A second related literature is occupational licensing.

06:23.875 --> 06:26.965
This is quite an important literature in labor economics.

06:27.585 --> 06:31.645
Uh, the general findings of this literature is that,

06:31.865 --> 06:33.165
in modern contests,

06:33.565 --> 06:37.165
occupational licensing typically has little effect on

06:37.165 --> 06:38.765
improving workers', quality,

06:39.565 --> 06:41.125
although it increased workers' wage.

06:41.945 --> 06:44.245
But there are a couple, papers found

06:44.245 --> 06:46.205
that in early error, a def formation

06:46.425 --> 06:47.805
of professional standards.

06:47.925 --> 06:51.125
Occupational li occupational licensing has some improving

06:51.125 --> 06:52.725
effects on workers' quality.

06:53.705 --> 06:57.125
Uh, a third related literature, I think it will be the,

06:57.125 --> 06:59.245
worker human capital and productivity.

07:00.225 --> 07:04.805
Uh, this literature mainly focus on productivity variation

07:04.805 --> 07:06.005
within professions.

07:06.545 --> 07:09.605
In our paper, we also look at productivity differences

07:09.835 --> 07:11.045
between professions.

07:11.785 --> 07:13.885
One interesting finding of this paper is

07:13.885 --> 07:17.285
that there could be important productivity difference

07:17.435 --> 07:20.125
between, professions, in our case,

07:20.215 --> 07:23.325
nurse practitioners versus physicians training.

07:23.505 --> 07:27.485
And the selectivity of a, of a profession could be po

07:28.245 --> 07:30.685
positively correlated with its productivity.

07:31.625 --> 07:34.445
On the other hand, our results also show

07:34.445 --> 07:37.925
that there could be very large productivity variation within

07:38.075 --> 07:41.925
each profession, and a large overlap bet in productivity

07:41.995 --> 07:43.125
between professions.

07:43.585 --> 07:44.885
So this also suggests

07:44.885 --> 07:47.885
that profession may be a highly imperfect

07:48.095 --> 07:49.485
proxy for productivity.

07:50.785 --> 07:53.565
And in terms of health policy implications,

07:53.885 --> 07:56.965
I think the question may not be as simple

07:57.105 --> 07:59.845
as whether we should use nurse practitioners or not.

08:00.245 --> 08:02.805
I think here the more natural question could be like,

08:03.475 --> 08:06.205
what might be the opt optimal patient assignments

08:06.315 --> 08:08.805
between physicians and nurse practitioners?

08:08.955 --> 08:11.125
Because for less complex patients,

08:11.175 --> 08:13.805
nurse practitioners may perform as good

08:13.825 --> 08:15.405
as physicians or even better.

08:16.225 --> 08:17.605
And also I think,

08:18.115 --> 08:20.245
another question we could think about would be

08:20.545 --> 08:22.005
how we can make physicians

08:22.005 --> 08:24.725
and nurse practitioners collaborate for the,

08:24.785 --> 08:26.485
for even patient and

08:26.485 --> 08:29.045
what might be the optimal task assignments

08:29.145 --> 08:30.805
to each different types of workers.

08:31.105 --> 08:33.005
So they, there could be a good, skill,

08:33.375 --> 08:35.685
there could be a good, skill task matching.

00:00 - In this paper, we look at the productivity of, physicians and, nurse practitioners. in the US there's a increasing demand for healthcare, but physician supply is relatively inelastic. So this leads to the increased use of nurse practitioners as substitute for physicians, nurse practitioners,

00:23 - they are type of mid-level practitioners. They are trained to diagnose patients and also treat patients. And in this paper we look at, in the emergency department setting, which is an important setting, that is nurse practitioners whether there are differences between physicians and nurse practitioners in the, medical resource use and patient treatment outcomes,

00:49 - whether there are difference in these dimensions. So we look at this question at the Veterans Health Administration in the us, the VA, which is often just called va, the va, allow, they have a policy, they passed a policy in 2016 that allowed nurse practitioners to practice independently without physician supervision. Nurse practitioners can, for example, diagnose patients, order treatments and prescribe drugs as independently

01:21 - as physicians at the va. The data, we are using the VA's Administrative Health records. So basically it record the patients we the who treat the patient, and, all the treatments the patients received during the visit. In the data, we can track whether the pa the provider treats the patient, is a nurse practitioner or physician, what, what type of,

01:43 - treatments the patient, received. And we can also track patient outcomes like, their mortality and hospital readmissions. our identification is the instrumental variable. H we use the, number of, nurse practitioners on duty at the ED on a day. The patient visited as an instrument for whether the patient is treated by a nurse practitioner or not. So if we, compared the average characteristics

02:11 - of patients treated by nurse practitioners versus the average characteristics of patients treated by physicians, we see that nurse practitioners on average treat healthier patients than physicians. For example, patients treated by nurse practitioners are younger and have fewer comorbidities. This suggests that we cannot simply compare, patient outcomes of patients treated

02:36 - by physicians versus nurse practitioners to see their productivity differences. We therefore uses quasi random variation in the number of nurse petitioners on duty at ed, on a data patient visited. So the, the eds in our setting, they use both physicians and nurse practitioners to treat patient, and they use physicians And nurse practitioners as substitutes. On days when there are more nurse practitioners,

03:04 - there are fewer physicians. And when there are fewer nurse practitioners, there are more physicians. Also, because of the institutional features of providers on duty scheduling, there is quasi random variation in a number of nurse petitioners on duty. But patient probability of being treated by nurse petitioners increases with the number of nurse practitioner on duty.

03:28 - So we use the number of nurse petitioners on duty as the ED on the day the patient visited as instrument for whether the patient is treated by a nurse practitioner. We show that conditional on the baseline controls like hos, ed by day of the week, ed by month, and ed by year. Fixed effects, patient eristics are very well balanced across the number of nurse practitioners on duty. We found that on average, there are productivity differences between physicians

04:00 - and nurse practitioners. On average nurse practitioners, make patients spend more on healthcare but get worse treatment outcomes. But we also find that there's very large productivity variation within each, each profession. Among the physicians, there are also low quality physicians, and among the nurse practitioners, there are also high quality nurse practitioners.

04:27 - There's very large productivity overlap between the two professions, and some nurse practitioners can perform even better than an average physician. for the mechanisms, we found some, mechanisms suggesting differences in human capital between the two professions. On average, nurse practitioners, have shown lower productivity than physicians, and we found that, experience might play a role here.

04:54 - When providers experience that, is higher, the nurse, nurse practitioners, physicians' performance gap become, lower in some outcomes. So we found that nurse practitioners tend to be more likely to use, treatments like, dag diagnostic tests and also like consults, which may help with their treatments and a diagnosis. This may also contribute to the higher medical spending of nurse practitioners. we also find that the performance difference

05:26 - between physicians and nurse practitioners, become higher when the patient's conditions are more complex. This suggests that it could be important to think about what could be the optimal patient assignments between physicians and nurse practitioners. I think there are a couple related res. One, directly related ation will be the role of nurse practitioners in, healthcare delivery.

05:52 - Given that the nurse practitioners are increasingly used in the US healthcare system, There is a growing number of papers looking at the role of nurse petitioners in expanding healthcare supply capacity. In addition to looking at the effects of nurse petitioners on expanding healthcare, supply capacity, I think it's also very important for us to understand the productivity of nurse petitioners. A second related literature is occupational licensing.

06:23 - This is quite an important literature in labor economics. the general findings of this literature is that, in modern contests, occupational licensing typically has little effect on improving workers', quality, although it increased workers' wage. But there are a couple, papers found that in early error, a def formation of professional standards. Occupational li occupational licensing has some improving

06:51 - effects on workers' quality. a third related literature, I think it will be the, worker human capital and productivity. this literature mainly focus on productivity variation within professions. In our paper, we also look at productivity differences between professions. One interesting finding of this paper is that there could be important productivity difference between, professions, in our case,

07:20 - nurse practitioners versus physicians training. And the selectivity of a, of a profession could be po positively correlated with its productivity. On the other hand, our results also show that there could be very large productivity variation within each profession, and a large overlap bet in productivity between professions. So this also suggests that profession may be a highly imperfect proxy for productivity.

07:50 - And in terms of health policy implications, I think the question may not be as simple as whether we should use nurse practitioners or not. I think here the more natural question could be like, what might be the opt optimal patient assignments between physicians and nurse practitioners? Because for less complex patients, nurse practitioners may perform as good as physicians or even better. And also I think,

08:18 - another question we could think about would be how we can make physicians and nurse practitioners collaborate for the, for even patient and what might be the optimal task assignments to each different types of workers. So they, there could be a good, skill, there could be a good, skill task matching.

Faculti AI can make mistakes